Remove param from docstring that does not exist in the append_file (#4060)

Regression fixed

.devcontainer/README.MD

@@ -0,0 +1 @@
+The files in this directory configure a development container for GitHub Codespaces.

.devcontainer/devcontainer.json

@@ -0,0 +1,15 @@
+{
+	"name": "OpenHands Codespaces",
+	"image": "mcr.microsoft.com/devcontainers/universal",
+	"customizations":{
+        "vscode":{
+            "extensions": [
+                "ms-python.python"
+            ]
+        }
+    },
+	"onCreateCommand": "sh ./.devcontainer/on_create.sh",
+	"postCreateCommand": "make build",
+	"postStartCommand": "USE_HOST_NETWORK=True nohup bash -c 'make run &'"
+
+}

.devcontainer/on_create.sh

@@ -0,0 +1,8 @@
+#!/usr/bin/env bash
+sudo apt update
+sudo apt install -y netcat
+sudo add-apt-repository -y ppa:deadsnakes/ppa
+sudo apt install -y python3.11
+curl -sSL https://install.python-poetry.org | python3.11 -
+# chromadb requires SQLite > 3.35 but SQLite in Python3.11.9 comes with 3.31.1
+sudo cp /opt/conda/lib/libsqlite3.so.0 /lib/x86_64-linux-gnu/libsqlite3.so.0

.github/.codecov.yml

@@ -1,6 +1,9 @@
 codecov:
   notify:
     wait_for_ci: true
+    # our project is large, so 6 builds are typically uploaded. this waits till 5/6
+    # See https://docs.codecov.com/docs/notifications#section-preventing-notifications-until-after-n-builds
+    after_n_builds: 5
 
 coverage:
   status:
@@ -14,6 +17,3 @@ coverage:
 comment: false
 github_checks:
     annotations: false
-
-ignore:
-  - "agenthub/SWE_agent/**" # SWE agent is deprecated

.github/ISSUE_TEMPLATE/bug_template.yml

@@ -1,5 +1,5 @@
 name: Bug
-description: Report a problem with OpenDevin
+description: Report a problem with OpenHands
 title: '[Bug]: '
 labels: ['bug']
 body:
@@ -12,7 +12,7 @@ body:
       label: Is there an existing issue for the same bug?
       description: Please check if an issue already exists for the bug you encountered.
       options:
-      - label: I have checked the troubleshooting document at https://opendevin.github.io/OpenDevin/modules/usage/troubleshooting
+      - label: I have checked the troubleshooting document at https://docs.all-hands.dev/modules/usage/troubleshooting
         required: true
       - label: I have checked the existing issues.
         required: true
@@ -28,8 +28,8 @@ body:
   - type: textarea
     id: current-version
     attributes:
-      label: Current OpenDevin version
-      description: What version of OpenDevin are you using? If you're running in docker, tell us the tag you're using (e.g. ghcr.io/opendevin/opendevin:0.3.1).
+      label: Current OpenHands version
+      description: What version of OpenHands are you using? If you're running in docker, tell us the tag you're using (e.g. ghcr.io/all-hands-ai/openhands:0.3.1).
       render: bash
     validations:
       required: true
@@ -72,4 +72,4 @@ body:
     id: additional-context
     attributes:
       label: Logs, Errors, Screenshots, and Additional Context
-      description: LLM logs will be stored in the `logs/llm/default` folder. Please add any additional context about the problem here.
+      description: If you want to share the chat history you can click the thumbs-down (👎) button above the input field and you will get a shareable link (you can also click thumbs up when things are going well of course!). LLM logs will be stored in the `logs/llm/default` folder. Please add any additional context about the problem here.

.github/ISSUE_TEMPLATE/feature_request.md

@@ -1,6 +1,6 @@
 ---
 name: Feature Request
-about: Suggest an idea for OpenDevin features
+about: Suggest an idea for OpenHands features
 title: ''
 labels: 'enhancement'
 assignees: ''

.github/ISSUE_TEMPLATE/question.md

@@ -1,16 +0,0 @@
----
-name: Question
-about: Use this template to ask a question regarding the project.
-title: ''
-labels: question
-assignees: ''
-
----
-
-## Describe your question
-
-<!--A clear and concise description of what you want to know.-->
-
-## Additional context
-
-<!--Add any other context about the question here, like what you've tried so far.-->

.github/dependabot.yml

@@ -5,11 +5,34 @@
 
 version: 2
 updates:
-  - package-ecosystem: "pip" # See documentation for possible values
-    directory: "/" # Location of package manifests
+  - package-ecosystem: "pip"
+    directory: "/"
     schedule:
       interval: "daily"
-  - package-ecosystem: "npm" # See documentation for possible values
-    directory: "/frontend" # Location of package manifests
+    open-pull-requests-limit: 20
+
+  - package-ecosystem: "npm"
+    directory: "/frontend"
     schedule:
       interval: "daily"
+    open-pull-requests-limit: 20
+    groups:
+      docusaurus:
+        patterns:
+          - "*docusaurus*"
+      eslint:
+        patterns:
+          - "*eslint*"
+
+  - package-ecosystem: "npm"
+    directory: "/docs"
+    schedule:
+      interval: "daily"
+    open-pull-requests-limit: 20
+    groups:
+      docusaurus:
+        patterns:
+          - "*docusaurus*"
+      eslint:
+        patterns:
+          - "*eslint*"

.github/pull_request_template.md

@@ -0,0 +1,11 @@
+**Short description of the problem this fixes or functionality that this introduces. This may be used for the CHANGELOG**
+
+
+
+---
+**Give a summary of what the PR does, explaining any non-trivial design decisions**
+
+
+
+---
+**Link of any specific issues this addresses**

.github/workflows/clean-up.yml

@@ -0,0 +1,69 @@
+# Workflow that cleans up outdated and old workflows to prevent out of disk issues
+name: Delete old workflow runs
+
+# This workflow is currently only triggered manually
+on:
+  workflow_dispatch:
+    inputs:
+      days:
+        description: 'Days-worth of runs to keep for each workflow'
+        required: true
+        default: '30'
+      minimum_runs:
+        description: 'Minimum runs to keep for each workflow'
+        required: true
+        default: '10'
+      delete_workflow_pattern:
+        description: 'Name or filename of the workflow (if not set, all workflows are targeted)'
+        required: false
+      delete_workflow_by_state_pattern:
+        description: 'Filter workflows by state: active, deleted, disabled_fork, disabled_inactivity, disabled_manually'
+        required: true
+        default: "ALL"
+        type: choice
+        options:
+          - "ALL"
+          - active
+          - deleted
+          - disabled_inactivity
+          - disabled_manually
+      delete_run_by_conclusion_pattern:
+        description: 'Remove runs based on conclusion: action_required, cancelled, failure, skipped, success'
+        required: true
+        default: 'ALL'
+        type: choice
+        options:
+          - 'ALL'
+          - 'Unsuccessful: action_required,cancelled,failure,skipped'
+          - action_required
+          - cancelled
+          - failure
+          - skipped
+          - success
+      dry_run:
+        description: 'Logs simulated changes, no deletions are performed'
+        required: false
+
+jobs:
+  del_runs:
+    runs-on: ubuntu-latest
+    permissions:
+      actions: write
+      contents: read
+    steps:
+      - name: Delete workflow runs
+        uses: Mattraks/delete-workflow-runs@v2
+        with:
+          token: ${{ github.token }}
+          repository: ${{ github.repository }}
+          retain_days: ${{ github.event.inputs.days }}
+          keep_minimum_runs: ${{ github.event.inputs.minimum_runs }}
+          delete_workflow_pattern: ${{ github.event.inputs.delete_workflow_pattern }}
+          delete_workflow_by_state_pattern: ${{ github.event.inputs.delete_workflow_by_state_pattern }}
+          delete_run_by_conclusion_pattern: >-
+            ${{
+              startsWith(github.event.inputs.delete_run_by_conclusion_pattern, 'Unsuccessful:')
+              && 'action_required,cancelled,failure,skipped'
+              || github.event.inputs.delete_run_by_conclusion_pattern
+            }}
+          dry_run: ${{ github.event.inputs.dry_run }}

.github/workflows/deploy-docs.yml

@@ -1,18 +1,25 @@
+# Workflow that builds and deploys the documentation website
 name: Deploy Docs to GitHub Pages
 
+# * Always run on "main"
+# * Run on PRs that target the "main" branch and have changes in the "docs" folder or this workflow
 on:
   push:
     branches:
       - main
   pull_request:
+    paths:
+      - 'docs/**'
+      - '.github/workflows/deploy-docs.yml'
     branches:
       - main
 
 jobs:
+  # Build the documentation website
   build:
+    if: github.repository == 'All-Hands-AI/OpenHands'
     name: Build Docusaurus
     runs-on: ubuntu-latest
-    if: github.repository == 'OpenDevin/OpenDevin'
     steps:
       - uses: actions/checkout@v4
         with:
@@ -25,25 +32,29 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v5
         with:
-          python-version: "3.11"
-
+          python-version: '3.11'
       - name: Generate Python Docs
         run: rm -rf docs/modules/python && pip install pydoc-markdown && pydoc-markdown
       - name: Install dependencies
         run: cd docs && npm ci
       - name: Build website
         run: cd docs && npm run build
-
       - name: Upload Build Artifact
         if: github.ref == 'refs/heads/main'
         uses: actions/upload-pages-artifact@v3
         with:
           path: docs/build
 
+  # Deploy the documentation website
   deploy:
+    if: github.ref == 'refs/heads/main' && github.repository == 'All-Hands-AI/OpenHands'
     name: Deploy to GitHub Pages
+    runs-on: ubuntu-latest
+    # This job only runs on "main" so only run one of these jobs at a time
+    # otherwise it will fail if one is already running
+    concurrency:
+      group: ${{ github.workflow }}-${{ github.ref }}
     needs: build
-    if: github.ref == 'refs/heads/main' && github.repository == 'OpenDevin/OpenDevin'
     # Grant GITHUB_TOKEN the permissions required to make a Pages deployment
     permissions:
       pages: write # to deploy to Pages
@@ -52,7 +63,6 @@ jobs:
     environment:
       name: github-pages
       url: ${{ steps.deployment.outputs.page_url }}
-    runs-on: ubuntu-latest
     steps:
       - name: Deploy to GitHub Pages
         id: deployment

.github/workflows/dummy-agent-test.yml

@@ -1,9 +1,8 @@
-name: Run e2e test with dummy agent
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+# Workflow that uses the DummyAgent to run a simple task
+name: Run E2E test with dummy agent
 
+# Always run on "main"
+# Always run on PRs
 on:
   push:
     branches:
@@ -15,16 +14,43 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
+      - name: Free Disk Space (Ubuntu)
+        uses: jlumbroso/free-disk-space@main
+        with:
+          # this might remove tools that are actually needed,
+          # if set to "true" but frees about 6 GB
+          tool-cache: true
+          # all of these default to true, but feel free to set to
+          # "false" if necessary for your workflow
+          android: true
+          dotnet: true
+          haskell: true
+          large-packages: true
+          docker-images: false
+          swap-storage: true
+      - name: Set up Docker Buildx
+        id: buildx
+        uses: docker/setup-buildx-action@v3
+      - name: Install poetry via pipx
+        run: pipx install poetry
       - name: Set up Python
         uses: actions/setup-python@v5
         with:
           python-version: '3.11'
-      - name: Set up environment
-        run: |
-          curl -sSL https://install.python-poetry.org | python3 -
-          poetry install --without evaluation
-          poetry run playwright install --with-deps chromium
-          wget https://huggingface.co/BAAI/bge-small-en-v1.5/raw/main/1_Pooling/config.json -P /tmp/llama_index/models--BAAI--bge-small-en-v1.5/snapshots/5c38ec7c405ec4b44b94cc5a9bb96e735b38267a/1_Pooling/
+          cache: 'poetry'
+      - name: Install Python dependencies using Poetry
+        run: poetry install --without evaluation,llama-index
+      - name: Build Environment
+        run: make build
       - name: Run tests
         run: |
-          poetry run python opendevin/core/main.py -t "do a flip" -m ollama/not-a-model -d ./workspace/ -c DummyAgent
+          set -e
+          poetry run python3 openhands/core/main.py -t "do a flip" -d ./workspace/ -c DummyAgent
+      - name: Check exit code
+        run: |
+          if [ $? -ne 0 ]; then
+            echo "Test failed"
+            exit 1
+          else
+            echo "Test passed"
+          fi

.github/workflows/fe-unit-tests.yml

@@ -0,0 +1,39 @@
+# Workflow that runs frontend unit tests
+name: Run Frontend Unit Tests
+
+# * Always run on "main"
+# * Run on PRs that have changes in the "frontend" folder or this workflow
+on:
+  push:
+    branches:
+      - main
+  pull_request:
+    paths:
+      - 'frontend/**'
+      -  '.github/workflows/fe-unit-tests.yml'
+
+jobs:
+  # Run frontend unit tests
+  fe-test:
+    name: FE Unit Tests
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        node-version: [20]
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Set up Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: ${{ matrix.node-version }}
+      - name: Install dependencies
+        working-directory: ./frontend
+        run: npm ci
+      - name: Run tests and collect coverage
+        working-directory: ./frontend
+        run: npm run test:coverage
+      - name: Upload coverage to Codecov
+        uses: codecov/codecov-action@v4
+        env:
+          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}

.github/workflows/ghcr_app.yml

@@ -1,13 +1,14 @@
-name: Publish Docker Image
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+# Workflow that builds, tests and then pushes the app docker images to the ghcr.io repository
+name: Build and Publish App Image
 
+# Always run on "main"
+# Always run on tags
+# Always run on PRs
+# Can also be triggered manually
 on:
   push:
     branches:
-    - main
+      - main
     tags:
       - '*'
   pull_request:
@@ -19,28 +20,22 @@ on:
         default: ''
 
 jobs:
-  ghcr_build_and_push:
+  # Builds the OpenHands Docker images
+  ghcr_build:
+    name: Build App Image
     runs-on: ubuntu-latest
-
     permissions:
       contents: read
       packages: write
-
-    strategy:
-      matrix:
-        image: ["app", "sandbox"]
-
     steps:
-      - name: checkout
+      - name: Checkout
         uses: actions/checkout@v4
-
       - name: Free Disk Space (Ubuntu)
         uses: jlumbroso/free-disk-space@main
         with:
           # this might remove tools that are actually needed,
           # if set to "true" but frees about 6 GB
           tool-cache: true
-
           # all of these default to true, but feel free to set to
           # "false" if necessary for your workflow
           android: true
@@ -49,34 +44,22 @@ jobs:
           large-packages: true
           docker-images: false
           swap-storage: true
-
       - name: Set up QEMU
         uses: docker/setup-qemu-action@v3
-
-      - name: Set up Docker Buildx
-        id: buildx
-        uses: docker/setup-buildx-action@v3
-
-      - name: Login to ghcr
-        uses: docker/login-action@v1
+      - name: Login to GHCR
+        uses: docker/login-action@v3
         with:
           registry: ghcr.io
           username: ${{ github.repository_owner }}
           password: ${{ secrets.GITHUB_TOKEN }}
-
-      - name: Build and push ${{ matrix.image }}
+      - name: Set up Docker Buildx
+        id: buildx
+        uses: docker/setup-buildx-action@v3
+      - name: Build and push app image
         if: "!github.event.pull_request.head.repo.fork"
         run: |
-          ./containers/build.sh ${{ matrix.image }} ${{ github.repository_owner }} --push
-
-      - name: Build ${{ matrix.image }}
+          ./containers/build.sh openhands ${{ github.repository_owner }} --push
+      - name: Build app image
         if: "github.event.pull_request.head.repo.fork"
         run: |
-          ./containers/build.sh ${{ matrix.image }} ${{ github.repository_owner }}
-
-  docker_build_success:
-    name: Docker Build Success
-    runs-on: ubuntu-latest
-    needs: ghcr_build_and_push
-    steps:
-    - run: echo Done!
+          ./containers/build.sh openhands image ${{ github.repository_owner }}

.github/workflows/ghcr_runtime.yml

@@ -0,0 +1,358 @@
+# Workflow that builds, tests and then pushes the runtime docker images to the ghcr.io repository
+name: Build, Test and Publish RT Image
+
+# Only run one workflow of the same group at a time.
+# There can be at most one running and one pending job in a concurrency group at any time.
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+# Always run on "main"
+# Always run on tags
+# Always run on PRs
+# Can also be triggered manually
+on:
+  push:
+    branches:
+      - main
+    tags:
+      - '*'
+  pull_request:
+  workflow_dispatch:
+    inputs:
+      reason:
+        description: 'Reason for manual trigger'
+        required: true
+        default: ''
+
+jobs:
+  # Builds the runtime Docker images
+  ghcr_build_runtime:
+    name: Build Image
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      packages: write
+    strategy:
+      matrix:
+        base_image:
+          - image: 'nikolaik/python-nodejs:python3.11-nodejs22'
+            tag: nikolaik
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Free Disk Space (Ubuntu)
+        uses: jlumbroso/free-disk-space@main
+        with:
+          # this might remove tools that are actually needed,
+          # if set to "true" but frees about 6 GB
+          tool-cache: true
+          # all of these default to true, but feel free to set to
+          # "false" if necessary for your workflow
+          android: true
+          dotnet: true
+          haskell: true
+          large-packages: true
+          docker-images: false
+          swap-storage: true
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@v3
+      - name: Login to GHCR
+        uses: docker/login-action@v3
+        with:
+          registry: ghcr.io
+          username: ${{ github.repository_owner }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+      - name: Set up Docker Buildx
+        id: buildx
+        uses: docker/setup-buildx-action@v3
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+      - name: Cache Poetry dependencies
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cache/pypoetry
+            ~/.virtualenvs
+          key: ${{ runner.os }}-poetry-${{ hashFiles('**/poetry.lock') }}
+          restore-keys: |
+            ${{ runner.os }}-poetry-
+      - name: Install poetry via pipx
+        run: pipx install poetry
+      - name: Install Python dependencies using Poetry
+        run: make install-python-dependencies
+      - name: Create source distribution and Dockerfile
+        run: poetry run python3 openhands/runtime/utils/runtime_build.py --base_image ${{ matrix.base_image.image }} --build_folder containers/runtime --force_rebuild
+      - name: Build and push runtime image ${{ matrix.base_image.image }}
+        if: github.event.pull_request.head.repo.fork != true
+        run: |
+          ./containers/build.sh runtime ${{ github.repository_owner }} --push ${{ matrix.base_image.tag }}
+      # Forked repos can't push to GHCR, so we need to upload the image as an artifact
+      - name: Build runtime image ${{ matrix.base_image.image }} for fork
+        if: github.event.pull_request.head.repo.fork
+        uses: docker/build-push-action@v6
+        with:
+          tags: ghcr.io/all-hands-ai/runtime:${{ github.sha }}-${{ matrix.base_image.tag }}
+          outputs: type=docker,dest=/tmp/runtime-${{ matrix.base_image.tag }}.tar
+          context: containers/runtime
+      - name: Upload runtime image for fork
+        if: github.event.pull_request.head.repo.fork
+        uses: actions/upload-artifact@v4
+        with:
+          name: runtime-${{ matrix.base_image.tag }}
+          path: /tmp/runtime-${{ matrix.base_image.tag }}.tar
+
+  # Run unit tests with the EventStream runtime Docker images as root
+  test_runtime_root:
+    name: RT Unit Tests (Root)
+    needs: [ghcr_build_runtime]
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        base_image: ['nikolaik']
+    steps:
+      - uses: actions/checkout@v4
+      - name: Free Disk Space (Ubuntu)
+        uses: jlumbroso/free-disk-space@main
+        with:
+          # this might remove tools that are actually needed,
+          # if set to "true" but frees about 6 GB
+          tool-cache: true
+          # all of these default to true, but feel free to set to
+          # "false" if necessary for your workflow
+          android: true
+          dotnet: true
+          haskell: true
+          large-packages: true
+          docker-images: false
+          swap-storage: true
+      - name: Set up Docker Buildx
+        id: buildx
+        uses: docker/setup-buildx-action@v3
+      # Forked repos can't push to GHCR, so we need to download the image as an artifact
+      - name: Download runtime image for fork
+        if: github.event.pull_request.head.repo.fork
+        uses: actions/download-artifact@v4
+        with:
+          name: runtime-${{ matrix.base_image }}
+          path: /tmp
+      - name: Load runtime image for fork
+        if: github.event.pull_request.head.repo.fork
+        run: |
+          docker load --input /tmp/runtime-${{ matrix.base_image }}.tar
+      - name: Cache Poetry dependencies
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cache/pypoetry
+            ~/.virtualenvs
+          key: ${{ runner.os }}-poetry-${{ hashFiles('**/poetry.lock') }}
+          restore-keys: |
+            ${{ runner.os }}-poetry-
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+      - name: Install poetry via pipx
+        run: pipx install poetry
+      - name: Install Python dependencies using Poetry
+        run: make install-python-dependencies
+      - name: Run runtime tests
+        run: |
+          # We install pytest-xdist in order to run tests across CPUs
+          poetry run pip install pytest-xdist
+
+          # Install to be able to retry on failures for flaky tests
+          poetry run pip install pytest-rerunfailures
+
+          image_name=ghcr.io/${{ github.repository_owner }}/runtime:${{ github.sha }}-${{ matrix.base_image }}
+          image_name=$(echo $image_name | tr '[:upper:]' '[:lower:]')
+
+          SKIP_CONTAINER_LOGS=true \
+          TEST_RUNTIME=eventstream \
+          SANDBOX_USER_ID=$(id -u) \
+          SANDBOX_RUNTIME_CONTAINER_IMAGE=$image_name \
+          TEST_IN_CI=true \
+          RUN_AS_OPENHANDS=false \
+          poetry run pytest -n 3 -raR --reruns 1 --reruns-delay 3 --cov=agenthub --cov=openhands --cov-report=xml -s ./tests/runtime
+      - name: Upload coverage to Codecov
+        uses: codecov/codecov-action@v4
+        env:
+          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
+
+  # Run unit tests with the EventStream runtime Docker images as openhands user
+  test_runtime_oh:
+    name: RT Unit Tests (openhands)
+    runs-on: ubuntu-latest
+    needs: [ghcr_build_runtime]
+    strategy:
+      matrix:
+        base_image: ['nikolaik']
+    steps:
+      - uses: actions/checkout@v4
+      - name: Free Disk Space (Ubuntu)
+        uses: jlumbroso/free-disk-space@main
+        with:
+          # this might remove tools that are actually needed,
+          # if set to "true" but frees about 6 GB
+          tool-cache: true
+          # all of these default to true, but feel free to set to
+          # "false" if necessary for your workflow
+          android: true
+          dotnet: true
+          haskell: true
+          large-packages: true
+          docker-images: false
+          swap-storage: true
+      - name: Set up Docker Buildx
+        id: buildx
+        uses: docker/setup-buildx-action@v3
+      # Forked repos can't push to GHCR, so we need to download the image as an artifact
+      - name: Download runtime image for fork
+        if: github.event.pull_request.head.repo.fork
+        uses: actions/download-artifact@v4
+        with:
+          name: runtime-${{ matrix.base_image }}
+          path: /tmp
+      - name: Load runtime image for fork
+        if: github.event.pull_request.head.repo.fork
+        run: |
+          docker load --input /tmp/runtime-${{ matrix.base_image }}.tar
+      - name: Cache Poetry dependencies
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cache/pypoetry
+            ~/.virtualenvs
+          key: ${{ runner.os }}-poetry-${{ hashFiles('**/poetry.lock') }}
+          restore-keys: |
+            ${{ runner.os }}-poetry-
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+      - name: Install poetry via pipx
+        run: pipx install poetry
+      - name: Install Python dependencies using Poetry
+        run: make install-python-dependencies
+      - name: Run runtime tests
+        run: |
+          # We install pytest-xdist in order to run tests across CPUs
+          poetry run pip install pytest-xdist
+
+          # Install to be able to retry on failures for flaky tests
+          poetry run pip install pytest-rerunfailures
+
+          image_name=ghcr.io/${{ github.repository_owner }}/runtime:${{ github.sha }}-${{ matrix.base_image }}
+          image_name=$(echo $image_name | tr '[:upper:]' '[:lower:]')
+
+          SKIP_CONTAINER_LOGS=true \
+          TEST_RUNTIME=eventstream \
+          SANDBOX_USER_ID=$(id -u) \
+          SANDBOX_RUNTIME_CONTAINER_IMAGE=$image_name \
+          TEST_IN_CI=true \
+          RUN_AS_OPENHANDS=true \
+          poetry run pytest -n 3 -raR --reruns 1 --reruns-delay 3 --cov=agenthub --cov=openhands --cov-report=xml -s ./tests/runtime
+      - name: Upload coverage to Codecov
+        uses: codecov/codecov-action@v4
+        env:
+          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
+
+  # Run integration tests with the eventstream runtime Docker image
+  runtime_integration_tests_on_linux:
+    name: RT Integration Tests (Linux)
+    runs-on: ubuntu-latest
+    needs: [ghcr_build_runtime]
+    strategy:
+      fail-fast: false
+      matrix:
+        base_image: ['nikolaik']
+    steps:
+      - uses: actions/checkout@v4
+      - name: Free Disk Space (Ubuntu)
+        uses: jlumbroso/free-disk-space@main
+        with:
+          # this might remove tools that are actually needed,
+          # if set to "true" but frees about 6 GB
+          tool-cache: true
+          # all of these default to true, but feel free to set to
+          # "false" if necessary for your workflow
+          android: true
+          dotnet: true
+          haskell: true
+          large-packages: true
+          docker-images: false
+          swap-storage: true
+      - name: Set up Docker Buildx
+        id: buildx
+        uses: docker/setup-buildx-action@v3
+      # Forked repos can't push to GHCR, so we need to download the image as an artifact
+      - name: Download runtime image for fork
+        if: github.event.pull_request.head.repo.fork
+        uses: actions/download-artifact@v4
+        with:
+          name: runtime-${{ matrix.base_image }}
+          path: /tmp
+      - name: Load runtime image for fork
+        if: github.event.pull_request.head.repo.fork
+        run: |
+          docker load --input /tmp/runtime-${{ matrix.base_image }}.tar
+      - name: Cache Poetry dependencies
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cache/pypoetry
+            ~/.virtualenvs
+          key: ${{ runner.os }}-poetry-${{ hashFiles('**/poetry.lock') }}
+          restore-keys: |
+            ${{ runner.os }}-poetry-
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+      - name: Install poetry via pipx
+        run: pipx install poetry
+      - name: Install Python dependencies using Poetry
+        run: make install-python-dependencies
+      - name: Run integration tests
+        run: |
+          image_name=ghcr.io/${{ github.repository_owner }}/runtime:${{ github.sha }}-${{ matrix.base_image }}
+          image_name=$(echo $image_name | tr '[:upper:]' '[:lower:]')
+
+          TEST_RUNTIME=eventstream \
+          SANDBOX_USER_ID=$(id -u) \
+          SANDBOX_RUNTIME_CONTAINER_IMAGE=$image_name \
+          TEST_IN_CI=true \
+          TEST_ONLY=true \
+          ./tests/integration/regenerate.sh
+      - name: Upload coverage to Codecov
+        uses: codecov/codecov-action@v4
+        env:
+          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
+
+  # The two following jobs (named identically) are to check whether all the runtime tests have passed as the
+  # "All Runtime Tests Passed" is a required job for PRs to merge
+  # Due to this bug: https://github.com/actions/runner/issues/2566, we want to create a job that runs when the
+  # prerequisites have been cancelled or failed so merging is disallowed, otherwise Github considers "skipped" as "success"
+  runtime_tests_check_success:
+    name: All Runtime Tests Passed
+    if: ${{ !cancelled() && !contains(needs.*.result, 'failure') && !contains(needs.*.result, 'cancelled') }}
+    runs-on: ubuntu-latest
+    needs: [test_runtime_root, test_runtime_oh, runtime_integration_tests_on_linux]
+    steps:
+      - name: All tests passed
+        run: echo "All runtime tests have passed successfully!"
+
+  runtime_tests_check_fail:
+    name: All Runtime Tests Passed
+    if: ${{ cancelled() || contains(needs.*.result, 'failure') || contains(needs.*.result, 'cancelled') }}
+    runs-on: ubuntu-latest
+    needs: [test_runtime_root, test_runtime_oh, runtime_integration_tests_on_linux]
+    steps:
+      - name: Some tests failed
+        run: |
+          echo "Some runtime tests failed or were cancelled"
+          exit 1

.github/workflows/lint.yml

@@ -1,9 +1,9 @@
+# Workflow that runs lint on the frontend and python code
 name: Lint
 
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
-
+# The jobs in this workflow are required, so they must run at all times
+# Always run on "main"
+# Always run on PRs
 on:
   push:
     branches:
@@ -11,27 +11,26 @@ on:
   pull_request:
 
 jobs:
+  # Run lint on the frontend code
   lint-frontend:
     name: Lint frontend
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
-
       - name: Install Node.js 20
         uses: actions/setup-node@v4
         with:
           node-version: 20
-
       - name: Install dependencies
         run: |
           cd frontend
           npm install --frozen-lockfile
-
       - name: Lint
         run: |
           cd frontend
           npm run lint
 
+  # Run lint on the python code
   lint-python:
     name: Lint python
     runs-on: ubuntu-latest
@@ -47,11 +46,4 @@ jobs:
       - name: Install pre-commit
         run: pip install pre-commit==3.7.0
       - name: Run pre-commit hooks
-        if: github.ref != 'refs/heads/main'
-        run: |
-          git fetch https://github.com/OpenDevin/OpenDevin.git main:main && \
-          pre-commit run \
-            --files \
-            $(git diff --name-only $(git merge-base main $(git branch --show-current)) $(git branch --show-current) | tr '\n' ' ') \
-            --show-diff-on-failure \
-            --config ./dev_config/python/.pre-commit-config.yaml
+        run: pre-commit run --files openhands/**/* agenthub/**/* evaluation/**/* tests/**/* --show-diff-on-failure --config ./dev_config/python/.pre-commit-config.yaml

.github/workflows/py-unit-tests.yml

@@ -0,0 +1,132 @@
+# Workflow that runs python unit tests
+name: Run Python Unit Tests
+
+# The jobs in this workflow are required, so they must run at all times
+# * Always run on "main"
+# * Always run on PRs
+on:
+  push:
+    branches:
+      - main
+  pull_request:
+
+jobs:
+  # Run python unit tests on macOS
+  test-on-macos:
+    name: Python Unit Tests on macOS
+    runs-on: macos-12
+    env:
+      INSTALL_DOCKER: '1' # Set to '0' to skip Docker installation
+    strategy:
+      matrix:
+        python-version: ['3.11']
+    steps:
+      - uses: actions/checkout@v4
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Cache Poetry dependencies
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cache/pypoetry
+            ~/.virtualenvs
+          key: ${{ runner.os }}-poetry-${{ hashFiles('**/poetry.lock') }}
+          restore-keys: |
+            ${{ runner.os }}-poetry-
+      - name: Install poetry via pipx
+        run: pipx install poetry
+      - name: Install Python dependencies using Poetry
+        run: poetry install --without evaluation,llama-index
+      - name: Install & Start Docker
+        if: env.INSTALL_DOCKER == '1'
+        run: |
+          INSTANCE_NAME="colima-${GITHUB_RUN_ID}"
+
+          # Uninstall colima to upgrade to the latest version
+          if brew list colima &>/dev/null; then
+            brew uninstall colima
+            # unlinking colima dependency: go
+            brew uninstall go@1.21
+          fi
+          rm -rf ~/.colima ~/.lima
+          brew install --HEAD colima
+          brew install docker
+
+          start_colima() {
+            # Find a free port in the range 10000-20000
+            RANDOM_PORT=$((RANDOM % 10001 + 10000))
+
+            # Original line:
+            if ! colima start --network-address --arch x86_64 --cpu=1 --memory=1 --verbose --ssh-port $RANDOM_PORT; then
+              echo "Failed to start Colima."
+              return 1
+            fi
+            return 0
+          }
+
+          # Attempt to start Colima for 5 total attempts:
+          ATTEMPT_LIMIT=5
+          for ((i=1; i<=ATTEMPT_LIMIT; i++)); do
+
+            if start_colima; then
+              echo "Colima started successfully."
+              break
+            else
+              colima stop -f
+              sleep 10
+              colima delete -f
+              if [ $i -eq $ATTEMPT_LIMIT ]; then
+                exit 1
+              fi
+              sleep 10
+            fi
+          done
+
+          # For testcontainers to find the Colima socket
+          # https://github.com/abiosoft/colima/blob/main/docs/FAQ.md#cannot-connect-to-the-docker-daemon-at-unixvarrundockersock-is-the-docker-daemon-running
+          sudo ln -sf $HOME/.colima/default/docker.sock /var/run/docker.sock
+      - name: Build Environment
+        run: make build
+      - name: Set up Docker Buildx
+        id: buildx
+        uses: docker/setup-buildx-action@v3
+      - name: Run Tests
+        run: poetry run pytest --forked --cov=agenthub --cov=openhands --cov-report=xml ./tests/unit
+      - name: Upload coverage to Codecov
+        uses: codecov/codecov-action@v4
+        env:
+          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
+
+  # Run python unit tests on Linux
+  test-on-linux:
+    name: Python Unit Tests on Linux
+    runs-on: ubuntu-latest
+    env:
+      INSTALL_DOCKER: '0' # Set to '0' to skip Docker installation
+    strategy:
+      matrix:
+        python-version: ['3.11']
+    steps:
+      - uses: actions/checkout@v4
+      - name: Set up Docker Buildx
+        id: buildx
+        uses: docker/setup-buildx-action@v3
+      - name: Install poetry via pipx
+        run: pipx install poetry
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: 'poetry'
+      - name: Install Python dependencies using Poetry
+        run: poetry install --without evaluation,llama-index
+      - name: Build Environment
+        run: make build
+      - name: Run Tests
+        run: poetry run pytest --forked --cov=agenthub --cov=openhands --cov-report=xml -svv ./tests/unit
+      - name: Upload coverage to Codecov
+        uses: codecov/codecov-action@v4
+        env:
+          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}

.github/workflows/pypi-release.yml

@@ -0,0 +1,31 @@
+# Publishes the OpenHands PyPi package
+name: Publish PyPi Package
+
+# Triggered manually
+on:
+  workflow_dispatch:
+    inputs:
+      reason:
+        description: 'Reason for manual trigger'
+        required: true
+        default: ''
+
+jobs:
+  release:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: 3.11
+      - name: Install Poetry
+        uses: snok/install-poetry@v1.4.1
+        with:
+          virtualenvs-in-project: true
+          virtualenvs-path: ~/.virtualenvs
+      - name: Install Poetry Dependencies
+        run: poetry install --no-interaction --no-root
+      - name: Build poetry project
+        run: poetry build -v
+      - name: publish
+        run: poetry publish -u __token__ -p ${{ secrets.PYPI_TOKEN }}

.github/workflows/regenerate_integration_tests.yml

@@ -0,0 +1,73 @@
+
+name: Regenerate Integration Tests
+
+on:
+  workflow_dispatch:
+    inputs:
+      debug:
+        description: 'Enable debug mode'
+        type: boolean
+        default: true
+      log_to_file:
+        description: 'Enable logging to file'
+        type: boolean
+        default: true
+      force_regenerate_tests:
+        description: 'Force regeneration of tests'
+        type: boolean
+        default: false
+      force_use_llm:
+        description: 'Force use of LLM'
+        type: boolean
+        default: false
+
+jobs:
+  regenerate_integration_tests:
+    if: github.ref != 'refs/heads/main'
+    runs-on: ubuntu-latest
+
+    steps:
+    - name: Checkout repository
+      uses: actions/checkout@v4
+    - name: Set up Docker Buildx
+      id: buildx
+      uses: docker/setup-buildx-action@v3
+    - name: Set up Python
+      uses: actions/setup-python@v5
+      with:
+        python-version: "3.11"
+    - name: Cache Poetry dependencies
+      uses: actions/cache@v4
+      with:
+        path: |
+          ~/.cache/pypoetry
+          ~/.virtualenvs
+        key: ${{ runner.os }}-poetry-${{ hashFiles('**/poetry.lock') }}
+        restore-keys: |
+          ${{ runner.os }}-poetry-
+    - name: Install poetry via pipx
+      run: pipx install poetry
+    - name: Install Python dependencies using Poetry
+      run: make install-python-dependencies
+    - name: Build Environment
+      run: make build
+    - name: Regenerate integration tests
+      run: |
+        DEBUG=${{ inputs.debug }} \
+        LOG_TO_FILE=${{ inputs.log_to_file }} \
+        FORCE_REGENERATE_TESTS=${{ inputs.force_regenerate_tests }} \
+        FORCE_USE_LLM=${{ inputs.force_use_llm }} \
+        ./tests/integration/regenerate.sh
+    - name: Commit changes
+      run: |
+        if git diff --quiet --exit-code; then
+          echo "No changes to commit"
+          exit 0
+        fi
+
+        git config --global user.name 'github-actions[bot]'
+        git config --global user.email 'github-actions[bot]@users.noreply.github.com'
+        git add .
+        # run it twice in case pre-commit makes changes
+        git commit -am "Regenerate integration tests" || git commit -am "Regenerate integration tests"
+        git push

.github/workflows/review-pr.yml

@@ -1,4 +1,5 @@
-name: Use OpenDevin to Review Pull Request
+# Workflow that uses OpenHands to review a pull request. PR must be labeled 'review-this'
+name: Use OpenHands to Review Pull Request
 
 on:
   pull_request:
@@ -12,29 +13,31 @@ jobs:
   dogfood:
     if: contains(github.event.pull_request.labels.*.name, 'review-this')
     runs-on: ubuntu-latest
-    container:
-      image: ghcr.io/opendevin/opendevin
-      volumes:
-        - /var/run/docker.sock:/var/run/docker.sock
-
     steps:
+    - uses: actions/checkout@v4
+    - name: Set up Docker Buildx
+      id: buildx
+      uses: docker/setup-buildx-action@v3
+    - name: Set up Python
+      uses: actions/setup-python@v5
+      with:
+        python-version: '3.11'
     - name: install git, github cli
       run: |
-        apt-get install -y git gh
+        sudo apt-get install -y git gh
         git config --global --add safe.directory $PWD
-
     - name: Checkout Repository
       uses: actions/checkout@v4
       with:
         ref: ${{ github.event.pull_request.base.ref }} # check out the target branch
-
     - name: Download Diff
       run: |
         curl -O "${{ github.event.pull_request.diff_url }}" -L
-
     - name: Write Task File
       run: |
-        echo "Your coworker wants to apply a pull request to this project. Read and review ${{ github.event.pull_request.number }}.diff file. Create a review-${{ github.event.pull_request.number }}.txt and write your concise comments and suggestions there." > task.txt
+        echo "Your coworker wants to apply a pull request to this project." > task.txt
+        echo "Read and review ${{ github.event.pull_request.number }}.diff file. Create a review-${{ github.event.pull_request.number }}.txt and write your concise comments and suggestions there." >> task.txt
+        echo "Do not ask me for confirmation at any point." >> task.txt
         echo "" >> task.txt
         echo "Title" >> task.txt
         echo "${{ github.event.pull_request.title }}" >> task.txt
@@ -43,15 +46,25 @@ jobs:
         echo "${{ github.event.pull_request.body }}" >> task.txt
         echo "" >> task.txt
         echo "Diff file is: ${{ github.event.pull_request.number }}.diff" >> task.txt
-
-    - name: Run OpenDevin
-      env:
-        LLM_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-        SANDBOX_TYPE: exec
+    - name: Set up environment
       run: |
-        WORKSPACE_MOUNT_PATH=$GITHUB_WORKSPACE python ./opendevin/core/main.py -i 50 -f task.txt -d $GITHUB_WORKSPACE
+        curl -sSL https://install.python-poetry.org | python3 -
+        export PATH="/github/home/.local/bin:$PATH"
+        poetry install --without evaluation,llama-index
+        poetry run playwright install --with-deps chromium
+    - name: Run OpenHands
+      env:
+        LLM_API_KEY: ${{ secrets.LLM_API_KEY }}
+        LLM_MODEL: ${{ vars.LLM_MODEL }}
+      run: |
+        # Append path to launch poetry
+        export PATH="/github/home/.local/bin:$PATH"
+        # Append path to correctly import package, note: must set pwd at first
+        export PYTHONPATH=$(pwd):$PYTHONPATH
+        export WORKSPACE_MOUNT_PATH=$GITHUB_WORKSPACE
+        export WORKSPACE_BASE=$GITHUB_WORKSPACE
+        echo -e "/exit\n" | poetry run python openhands/core/main.py -i 50 -f task.txt
         rm task.txt
-
     - name: Check if review file is non-empty
       id: check_file
       run: |
@@ -60,7 +73,6 @@ jobs:
           echo "non_empty=true" >> $GITHUB_OUTPUT
         fi
       shell: bash
-
     - name: Create PR review if file is non-empty
       env:
         GH_TOKEN: ${{ github.token }}

.github/workflows/run-integration-tests.yml

@@ -1,101 +0,0 @@
-name: Run Integration Tests
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
-
-on:
-  push:
-    branches:
-      - main
-    paths-ignore:
-      - '**/*.md'
-      - 'frontend/**'
-      - 'docs/**'
-      - 'evaluation/**'
-  pull_request:
-
-jobs:
-  integration-tests-on-linux:
-    name: Integration Tests on Linux
-    runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
-      matrix:
-        python-version: ["3.11"]
-        sandbox: ["ssh", "exec", "local"]
-    steps:
-      - uses: actions/checkout@v4
-
-      - name: Install poetry via pipx
-        run: pipx install poetry
-
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: ${{ matrix.python-version }}
-          cache: 'poetry'
-
-      - name: Install Python dependencies using Poetry
-        run: poetry install
-
-      - name: Build Environment
-        run: make build
-
-      - name: Run Integration Tests
-        env:
-          SANDBOX_TYPE: ${{ matrix.sandbox }}
-        run: |
-          TEST_IN_CI=true TEST_ONLY=true ./tests/integration/regenerate.sh
-
-      - name: Upload coverage to Codecov
-        uses: codecov/codecov-action@v4
-        env:
-          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
-
-  integration-tests-on-mac:
-    name: Integration Tests on MacOS
-    runs-on: macos-13
-    if: contains(github.event.pull_request.title, 'mac') || contains(github.event.pull_request.title, 'Mac')
-    strategy:
-      fail-fast: false
-      matrix:
-        python-version: ["3.11"]
-        sandbox: ["ssh"]
-    steps:
-      - uses: actions/checkout@v4
-
-      - name: Install poetry via pipx
-        run: pipx install poetry
-
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: ${{ matrix.python-version }}
-          cache: 'poetry'
-
-      - name: Install Python dependencies using Poetry
-        run: poetry install
-
-      - name: Install & Start Docker
-        run: |
-          brew install colima docker
-          colima start
-
-          # For testcontainers to find the Colima socket
-          # https://github.com/abiosoft/colima/blob/main/docs/FAQ.md#cannot-connect-to-the-docker-daemon-at-unixvarrundockersock-is-the-docker-daemon-running
-          sudo ln -sf $HOME/.colima/default/docker.sock /var/run/docker.sock
-
-      - name: Build Environment
-        run: make build
-
-      - name: Run Integration Tests
-        env:
-          SANDBOX_TYPE: ${{ matrix.sandbox }}
-        run: |
-          TEST_IN_CI=true TEST_ONLY=true ./tests/integration/regenerate.sh
-
-      - name: Upload coverage to Codecov
-        uses: codecov/codecov-action@v4
-        env:
-          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}

.github/workflows/run-unit-tests.yml

@@ -1,126 +0,0 @@
-name: Run Unit Tests
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
-
-on:
-  push:
-    branches:
-      - main
-    paths-ignore:
-      - '**/*.md'
-      - 'frontend/**'
-      - 'docs/**'
-      - 'evaluation/**'
-  pull_request:
-
-jobs:
-  test-on-macos:
-    name: Test on macOS
-    runs-on: macos-13
-    env:
-      INSTALL_DOCKER: "0" # Set to '0' to skip Docker installation
-    strategy:
-      matrix:
-        python-version: ["3.11"]
-
-    steps:
-      - uses: actions/checkout@v4
-
-      - name: Install poetry via pipx
-        run: pipx install poetry
-
-      - name: Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v5
-        with:
-          python-version: ${{ matrix.python-version }}
-          cache: "poetry"
-
-      - name: Install Python dependencies using Poetry
-        run: poetry install
-
-      - name: Install & Start Docker
-        if: env.INSTALL_DOCKER == '1'
-        run: |
-          brew install colima docker
-          colima start
-
-          # For testcontainers to find the Colima socket
-          # https://github.com/abiosoft/colima/blob/main/docs/FAQ.md#cannot-connect-to-the-docker-daemon-at-unixvarrundockersock-is-the-docker-daemon-running
-          sudo ln -sf $HOME/.colima/default/docker.sock /var/run/docker.sock
-
-      - name: Build Environment
-        run: make build
-
-      - name: Run Tests
-        run: poetry run pytest --forked --cov=agenthub --cov=opendevin --cov-report=xml ./tests/unit -k "not test_sandbox"
-
-      - name: Upload coverage to Codecov
-        uses: codecov/codecov-action@v4
-        env:
-          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
-  test-on-linux:
-    name: Test on Linux
-    runs-on: ubuntu-latest
-    env:
-      INSTALL_DOCKER: "0" # Set to '0' to skip Docker installation
-    strategy:
-      matrix:
-        python-version: ["3.11"]
-
-    steps:
-      - uses: actions/checkout@v4
-
-      - name: Install poetry via pipx
-        run: pipx install poetry
-
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: ${{ matrix.python-version }}
-          cache: "poetry"
-
-      - name: Install Python dependencies using Poetry
-        run: poetry install --without evaluation
-
-      - name: Build Environment
-        run: make build
-
-      - name: Run Tests
-        run: poetry run pytest --forked --cov=agenthub --cov=opendevin --cov-report=xml ./tests/unit -k "not test_sandbox"
-
-      - name: Upload coverage to Codecov
-        uses: codecov/codecov-action@v4
-        env:
-          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
-
-  test-for-sandbox:
-    name: Test for Sandbox
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-
-      - name: Install poetry via pipx
-        run: pipx install poetry
-
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: "3.11"
-          cache: "poetry"
-
-      - name: Install Python dependencies using Poetry
-        run: poetry install
-
-      - name: Build Environment
-        run: make build
-
-      - name: Run Integration Test for Sandbox
-        run: |
-          poetry run pytest --cov=agenthub --cov=opendevin --cov-report=xml -s ./tests/unit/test_sandbox.py
-
-      - name: Upload coverage to Codecov
-        uses: codecov/codecov-action@v4
-        env:
-          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}

.github/workflows/solve-issue.yml

@@ -1,4 +1,5 @@
-name: Use OpenDevin to Resolve GitHub Issue
+# Workflow that uses OpenHands to resolve a GitHub issue. Issue must be labeled 'solve-this'
+name: Use OpenHands to Resolve GitHub Issue
 
 on:
   issues:
@@ -14,17 +15,17 @@ jobs:
     if: github.event.label.name == 'solve-this'
     runs-on: ubuntu-latest
     container:
-      image: ghcr.io/opendevin/opendevin
+      image: ghcr.io/all-hands-ai/openhands
       volumes:
         - /var/run/docker.sock:/var/run/docker.sock
-
     steps:
     - name: install git, github cli
       run: apt-get install -y git gh
-
+    - name: Set up Docker Buildx
+      id: buildx
+      uses: docker/setup-buildx-action@v3
     - name: Checkout Repository
       uses: actions/checkout@v4
-
     - name: Write Task File
       env:
         ISSUE_TITLE: ${{ github.event.issue.title }}
@@ -35,23 +36,31 @@ jobs:
         echo "" >> task.txt
         echo "BODY:" >> task.txt
         echo "${ISSUE_BODY}" >> task.txt
-
-    - name: Run OpenDevin
+    - name: Set up environment
+      run: |
+        curl -sSL https://install.python-poetry.org | python3 -
+        export PATH="/github/home/.local/bin:$PATH"
+        poetry install --without evaluation,llama-index
+        poetry run playwright install --with-deps chromium
+    - name: Run OpenHands
       env:
         ISSUE_TITLE: ${{ github.event.issue.title }}
         ISSUE_BODY: ${{ github.event.issue.body }}
         LLM_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-        SANDBOX_TYPE: exec
+        OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
       run: |
-        WORKSPACE_MOUNT_PATH=$GITHUB_WORKSPACE python ./opendevin/core/main.py -i 50 -f task.txt -d $GITHUB_WORKSPACE
+        # Append path to launch poetry
+        export PATH="/github/home/.local/bin:$PATH"
+        # Append path to correctly import package, note: must set pwd at first
+        export PYTHONPATH=$(pwd):$PYTHONPATH
+        WORKSPACE_MOUNT_PATH=$GITHUB_WORKSPACE poetry run python ./openhands/core/main.py -i 50 -f task.txt -d $GITHUB_WORKSPACE
         rm task.txt
-
     - name: Setup Git, Create Branch, and Commit Changes
       run: |
         # Setup Git configuration
         git config --global --add safe.directory $PWD
-        git config --global user.name 'OpenDevin'
-        git config --global user.email 'OpenDevin@users.noreply.github.com'
+        git config --global user.name 'OpenHands'
+        git config --global user.email 'OpenHands@users.noreply.github.com'
 
         # Create a unique branch name with a timestamp
         BRANCH_NAME="fix/${{ github.event.issue.number }}-$(date +%Y%m%d%H%M%S)"
@@ -63,7 +72,7 @@ jobs:
         git add --all -- ':!task.txt'
 
         # Commit the changes, if any
-        git commit -m "OpenDevin: Resolve Issue #${{ github.event.issue.number }}"
+        git commit -m "OpenHands: Resolve Issue #${{ github.event.issue.number }}"
         if [ $? -ne 0 ]; then
           echo "No changes to commit."
           exit 0
@@ -71,7 +80,6 @@ jobs:
 
         # Push changes
         git push --set-upstream origin $BRANCH_NAME
-
     - name: Fetch Default Branch
       env:
         GH_TOKEN: ${{ github.token }}
@@ -80,16 +88,15 @@ jobs:
         DEFAULT_BRANCH=$(gh repo view --json defaultBranchRef --jq .defaultBranchRef.name)
         echo "Default branch is $DEFAULT_BRANCH"
         echo "DEFAULT_BRANCH=$DEFAULT_BRANCH" >> $GITHUB_ENV
-
     - name: Generate PR
       env:
         GH_TOKEN: ${{ github.token }}
       run: |
         # Create PR and capture URL
         PR_URL=$(gh pr create \
-          --title "OpenDevin: Resolve Issue #2" \
-          --body "This PR was generated by OpenDevin to resolve issue #2" \
-          --repo "foragerr/OpenDevin" \
+          --title "OpenHands: Resolve Issue #2" \
+          --body "This PR was generated by OpenHands to resolve issue #2" \
+          --repo "foragerr/OpenHands" \
           --head "${{ github.head_ref }}" \
           --base "${{ env.DEFAULT_BRANCH }}" \
           | grep -o 'https://github.com/[^ ]*')
@@ -106,4 +113,4 @@ jobs:
         GH_TOKEN: ${{ github.token }}
       run: |
         gh issue comment ${{ github.event.issue.number }} \
-          -b "OpenDevin raised [PR #${{ env.PR_NUMBER }}](${{ env.PR_URL }}) to resolve this issue."
+          -b "OpenHands raised [PR #${{ env.PR_NUMBER }}](${{ env.PR_URL }}) to resolve this issue."

.github/workflows/stale.yml

@@ -1,4 +1,7 @@
+# Workflow that marks issues and PRs with no activity for 30 days with "Stale" and closes them after 7 more days of no activity
 name: 'Close stale issues'
+
+# Runs every day at 01:30
 on:
   schedule:
     - cron: '30 1 * * *'
@@ -9,21 +12,10 @@ jobs:
     steps:
       - uses: actions/stale@v9
         with:
-          # Aggressively close issues that have been explicitly labeled `age-out`
-          any-of-labels: age-out
-          stale-issue-message: 'This issue is stale because it has been open for 7 days with no activity. Remove stale label or comment or this will be closed in 1 day.'
-          close-issue-message: 'This issue was closed because it has been stalled for over 7 days with no activity.'
-          stale-pr-message: 'This PR is stale because it has been open for 7 days with no activity. Remove stale label or comment or this will be closed in 1 days.'
-          close-pr-message: 'This PR was closed because it has been stalled for over 7 days with no activity.'
-          days-before-stale: 7
-          days-before-close: 1
-
-      - uses: actions/stale@v9
-        with:
-          # Be more lenient with other issues
           stale-issue-message: 'This issue is stale because it has been open for 30 days with no activity. Remove stale label or comment or this will be closed in 7 days.'
-          close-issue-message: 'This issue was closed because it has been stalled for over 30 days with no activity.'
           stale-pr-message: 'This PR is stale because it has been open for 30 days with no activity. Remove stale label or comment or this will be closed in 7 days.'
-          close-pr-message: 'This PR was closed because it has been stalled for over 30 days with no activity.'
           days-before-stale: 30
+          exempt-issue-labels: 'tracked'
+          close-issue-message: 'This issue was closed because it has been stalled for over 30 days with no activity.'
+          close-pr-message: 'This PR was closed because it has been stalled for over 30 days with no activity.'
           days-before-close: 7

.gitignore

@@ -161,9 +161,18 @@ cython_debug/
 #  option (not recommended) you can uncomment the following to ignore the entire idea folder.
 .idea/
 .vscode/
+.cursorignore
 
 # evaluation
+evaluation/evaluation_outputs
+evaluation/outputs
+evaluation/swe_bench/eval_workspace*
 evaluation/SWE-bench/data
+evaluation/webarena/scripts/webarena_env.sh
+evaluation/bird/data
+evaluation/gaia/data
+evaluation/gorilla/data
+evaluation/toolqa/data
 
 # frontend
 
@@ -176,6 +185,8 @@ frontend/yarn.lock
 
 # testing
 frontend/coverage
+test_results*
+/_test_files_tmp/
 
 # production
 frontend/build
@@ -203,9 +214,18 @@ cache
 
 # configuration
 config.toml
+config.toml_
 config.toml.bak
-evaluation/swe_bench/eval_workspace*
-evaluation/outputs
-evaluation/evaluation_outputs
-test_results*
-/_test_files_tmp/
+
+containers/agnostic_sandbox
+
+# swe-bench-eval
+image_build_logs
+run_instance_logs
+
+runtime_*.tar
+
+# docker build
+containers/runtime/Dockerfile
+containers/runtime/project.tar.gz
+containers/runtime/code

CODE_OF_CONDUCT.md

@@ -61,7 +61,7 @@ representative at an online or offline event.
 
 Instances of abusive, harassing, or otherwise unacceptable behavior may be
 reported to the community leaders responsible for enforcement at
-contact@rbren.io
+contact@all-hands.dev
 All complaints will be reviewed and investigated promptly and fairly.
 
 All community leaders are obligated to respect the privacy and security of the

CONTRIBUTING.md

@@ -1,36 +1,57 @@
 # Contributing
 
-Thanks for your interest in contributing to OpenDevin! We welcome and appreciate contributions.
-To report bugs, create a [GitHub issue](https://github.com/OpenDevin/OpenDevin/issues/new/choose).
+Thanks for your interest in contributing to OpenHands! We welcome and appreciate contributions.
+
+## How Can I Contribute?
+
+There are many ways that you can contribute:
+
+1. **Download and use** OpenHands, and send [issues](https://github.com/All-Hands-AI/OpenHands/issues) when you encounter something that isn't working or a feature that you'd like to see.
+2. **Send feedback** after each session by [clicking the thumbs-up thumbs-down buttons](https://docs.all-hands.dev/modules/usage/feedback), so we can see where things are working and failing, and also build an open dataset for training code agents.
+3. **Improve the Codebase** by sending PRs (see details below). In particular, we have some [good first issue](https://github.com/All-Hands-AI/OpenHands/labels/good%20first%20issue) issues that may be ones to start on.
+
+## Understanding OpenHands's CodeBase
+
+To understand the codebase, please refer to the README in each module:
+- [frontend](./frontend/README.md)
+- [agenthub](./agenthub/README.md)
+- [evaluation](./evaluation/README.md)
+- [openhands](./openhands/README.md)
+    - [server](./openhands/server/README.md)
+
+When you write code, it is also good to write tests. Please navigate to the `tests` folder to see existing test suites.
+At the moment, we have two kinds of tests: `unit` and `integration`. Please refer to the README for each test suite. These tests also run on GitHub's continuous integration to ensure quality of the project.
+
+## Sending Pull Requests to OpenHands
 
-## Contribution Guide
 ### 1. Fork the Official Repository
-
-Fork [OpenDevin repository](https://github.com/OpenDevin/OpenDevin) into your own account.
-Clone your own forked repository into your local environment.
+Fork the [OpenHands repository](https://github.com/All-Hands-AI/OpenHands) into your own account.
+Clone your own forked repository into your local environment:
 
 ```shell
-git clone git@github.com:<YOUR-USERNAME>/OpenDevin.git
+git clone git@github.com:<YOUR-USERNAME>/OpenHands.git
 ```
 
 ### 2. Configure Git
 
 Set the official repository as your [upstream](https://www.atlassian.com/git/tutorials/git-forks-and-upstreams) to synchronize with the latest update in the official repository.
-Add the original repository as upstream
+Add the original repository as upstream:
 
 ```shell
-cd OpenDevin
-git remote add upstream git@github.com:OpenDevin/OpenDevin.git
+cd OpenHands
+git remote add upstream git@github.com:All-Hands-AI/OpenHands.git
 ```
 
-Verify that the remote is set.
+Verify that the remote is set:
+
 ```shell
 git remote -v
 ```
+
 You should see both `origin` and `upstream` in the output.
 
 ### 3. Synchronize with Official Repository
-Synchronize latest commit with official repository before coding.
+Synchronize latest commit with official repository before coding:
 
 ```shell
 git fetch upstream
@@ -39,25 +60,48 @@ git merge upstream/main
 git push origin main
 ```
 
-### 4. Create a New Branch And Open a Pull Request
-After you finish implementation, open forked repository. The source branch is your new branch, and the target branch is `OpenDevin/OpenDevin` `main` branch. Then PR should appears in [OpenDevin PRs](https://github.com/OpenDevin/OpenDevin/pulls).
+### 4. Set up the Development Environment
 
-Then OpenDevin team will review your code.
+We have a separate doc [Development.md](https://github.com/All-Hands-AI/OpenHands/blob/main/Development.md) that tells you how to set up a development workflow.
+
+### 5. Write Code and Commit It
+
+Once you have done this, you can write code, test it, and commit it to a branch (replace `my_branch` with an appropriate name):
+
+```shell
+git checkout -b my_branch
+git add .
+git commit
+git push origin my_branch
+```
+
+### 6. Open a Pull Request
+
+* On GitHub, go to the page of your forked repository, and create a Pull Request:
+   - Click on `Branches`
+   - Click on the `...` beside your branch and click on `New pull request`
+   - Set `base repository` to `All-Hands-AI/OpenHands`
+   - Set `base` to `main`
+   - Click `Create pull request`
+
+The PR should appear in [OpenHands PRs](https://github.com/All-Hands-AI/OpenHands/pulls).
+
+Then the OpenHands team will review your code.
 
 ## PR Rules
 
 ### 1. Pull Request title
-
-As described in [here](https://github.com/commitizen/conventional-commit-types/blob/master/index.json), a valid PR title should begin with one of the following prefixes:
+As described [here](https://github.com/commitizen/conventional-commit-types/blob/master/index.json), a valid PR title should begin with one of the following prefixes:
 
 - `feat`: A new feature
 - `fix`: A bug fix
-- `doc`: Documentation only changes
+- `docs`: Documentation only changes
+- `style`: Changes that do not affect the meaning of the code (white space, formatting, missing semicolons, etc.)
 - `refactor`: A code change that neither fixes a bug nor adds a feature
-- `style`: A refactoring that improves code style
 - `perf`: A code change that improves performance
 - `test`: Adding missing tests or correcting existing tests
-- `ci`: Changes to CI configuration files and scripts (example scopes: `.github`, `ci` (Buildkite))
+- `build`: Changes that affect the build system or external dependencies (example scopes: gulp, broccoli, npm)
+- `ci`: Changes to our CI configuration files and scripts (example scopes: Travis, Circle, BrowserStack, SauceLabs)
 - `chore`: Other changes that don't modify src or test files
 - `revert`: Reverts a previous commit
 
@@ -65,26 +109,8 @@ For example, a PR title could be:
 - `refactor: modify package path`
 - `feat(frontend): xxxx`, where `(frontend)` means that this PR mainly focuses on the frontend component.
 
-You may also check out previous PRs in the [PR list](https://github.com/OpenDevin/OpenDevin/pulls).
-
-As described in [here](https://github.com/OpenDevin/OpenDevin/labels), we create several labels. Every PR should be tagged with the corresponding labels.
+You may also check out previous PRs in the [PR list](https://github.com/All-Hands-AI/OpenHands/pulls).
 
 ### 2. Pull Request description
-
 - If your PR is small (such as a typo fix), you can go brief.
-- If it is large and you have changed a lot, it's better to write more details.
-
-
-## How to begin
-Please refer to the README in each module:
-- [frontend](./frontend/README.md)
-- [agenthub](./agenthub/README.md)
-- [evaluation](./evaluation/README.md)
-- [opendevin](./opendevin/README.md)
-    - [server](./opendevin/server/README.md)
-    - [mock server](./opendevin/mock/README.md)
-
-## Tests
-Please navigate to `tests` folder to see existing test suites.
-At the moment, we have two kinds of tests: `unit` and `integration`. Please refer to the README for each test suite. These tests also run on CI to ensure quality of
-the project.
+- If it contains a lot of changes, it's better to write more details.

CREDITS.md

@@ -0,0 +1,312 @@
+# Credits
+
+## Contributors
+
+We would like to thank all the [contributors](https://github.com/All-Hands-AI/OpenHands/graphs/contributors) who have helped make OpenHands possible. Your dedication and hard work are greatly appreciated.
+
+## Open Source Projects
+
+OpenHands includes and adapts the following open source projects. We are grateful for their contributions to the open source community:
+
+#### [SWE Agent](https://github.com/princeton-nlp/swe-agent)
+   - License: MIT License
+   - Description: Adapted for use in OpenHands's agenthub
+
+#### [Aider](https://github.com/paul-gauthier/aider)
+   - License: Apache License 2.0
+   - Description: AI pair programming tool. OpenHands has adapted and integrated its linter module for code-related tasks in [`agentskills utilities`](https://github.com/All-Hands-AI/OpenHands/tree/main/openhands/runtime/plugins/agent_skills/utils/aider)
+
+#### [BrowserGym](https://github.com/ServiceNow/BrowserGym)
+   - License: Apache License 2.0
+   - Description: Adapted in implementing the browsing agent
+
+
+### Reference Implementations for Evaluation Benchmarks
+OpenHands integrates code of the reference implementations for the following agent evaluation benchmarks:
+
+#### [HumanEval](https://github.com/openai/human-eval)
+   - License: MIT License
+
+#### [DSP](https://github.com/microsoft/DataScienceProblems)
+   - License: MIT License
+
+#### [HumanEvalPack](https://github.com/bigcode-project/bigcode-evaluation-harness)
+   - License: Apache License 2.0
+
+#### [AgentBench](https://github.com/THUDM/AgentBench)
+   - License: Apache License 2.0
+
+#### [SWE-Bench](https://github.com/princeton-nlp/SWE-bench)
+   - License: MIT License
+
+#### [BIRD](https://bird-bench.github.io/)
+   - License: MIT License
+   - Dataset: CC-BY-SA 4.0
+
+#### [Gorilla APIBench](https://github.com/ShishirPatil/gorilla)
+   - License: Apache License 2.0
+
+#### [GPQA](https://github.com/idavidrein/gpqa)
+   - License: MIT License
+
+#### [ProntoQA](https://github.com/asaparov/prontoqa)
+   - License: Apache License 2.0
+
+
+## Open Source licenses
+
+### MIT License
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+### BSD 3-Clause License
+
+Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+
+1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+
+3. Neither the name of the copyright holder nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+### Apache License 2.0
+
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+
+
+### Non-Open Source Reference Implementations:
+
+#### [MultiPL-E](https://github.com/nuprl/MultiPL-E)
+   - License: BSD 3-Clause License with Machine Learning Restriction
+
+BSD 3-Clause License with Machine Learning Restriction
+
+Copyright (c) 2022, Northeastern University, Oberlin College, Roblox Inc,
+Stevens Institute of Technology, University of Massachusetts Amherst, and
+Wellesley College.
+
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+1. Redistributions of source code must retain the above copyright notice, this
+   list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright notice,
+   this list of conditions and the following disclaimer in the documentation
+   and/or other materials provided with the distribution.
+
+3. Neither the name of the copyright holder nor the names of its
+   contributors may be used to endorse or promote products derived from
+   this software without specific prior written permission.
+
+4.  The contents of this repository may not be used as training data for any
+    machine learning model, including but not limited to neural networks.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

Development.md

@@ -1,14 +1,16 @@
 # Development Guide
-This guide is for people working on OpenDevin and editing the source code.
+This guide is for people working on OpenHands and editing the source code.
+If you wish to contribute your changes, check out the [CONTRIBUTING.md](https://github.com/All-Hands-AI/OpenHands/blob/main/CONTRIBUTING.md) on how to clone and setup the project initially before moving on.
+Otherwise, you can clone the OpenHands project directly.
 
 ## Start the server for development
-
 ### 1. Requirements
-* Linux, Mac OS, or [WSL on Windows](https://learn.microsoft.com/en-us/windows/wsl/install)
-* [Docker](https://docs.docker.com/engine/install/)(For those on MacOS, make sure to allow the default Docker socket to be used from advanced settings!)
-* [Python](https://www.python.org/downloads/) >= 3.11
+* Linux, Mac OS, or [WSL on Windows](https://learn.microsoft.com/en-us/windows/wsl/install)  [ Ubuntu <= 22.04]
+* [Docker](https://docs.docker.com/engine/install/) (For those on MacOS, make sure to allow the default Docker socket to be used from advanced settings!)
+* [Python](https://www.python.org/downloads/) = 3.11
 * [NodeJS](https://nodejs.org/en/download/package-manager) >= 18.17.1
 * [Poetry](https://python-poetry.org/docs/#installing-with-the-official-installer) >= 1.8
+* netcat => sudo apt-get install netcat
 
 Make sure you have all these dependencies installed before moving on to `make build`.
 
@@ -27,41 +29,40 @@ mamba install conda-forge::poetry
 ```
 
 ### 2. Build and Setup The Environment
+Begin by building the project which includes setting up the environment and installing dependencies. This step ensures that OpenHands is ready to run on your system:
 
-- **Build the Project:** Begin by building the project, which includes setting up the environment and installing dependencies. This step ensures that OpenDevin is ready to run smoothly on your system.
-    ```bash
-    make build
-    ```
+```bash
+make build
+```
 
 ### 3. Configuring the Language Model
+OpenHands supports a diverse array of Language Models (LMs) through the powerful [litellm](https://docs.litellm.ai) library. By default, we've chosen the mighty GPT-4 from OpenAI as our go-to model, but the world is your oyster! You can unleash the potential of Anthropic's suave Claude, the enigmatic Llama, or any other LM that piques your interest.
 
-OpenDevin supports a diverse array of Language Models (LMs) through the powerful [litellm](https://docs.litellm.ai) library. By default, we've chosen the mighty GPT-4 from OpenAI as our go-to model, but the world is your oyster! You can unleash the potential of Anthropic's suave Claude, the enigmatic Llama, or any other LM that piques your interest.
+To configure the LM of your choice, run:
 
-To configure the LM of your choice, follow these steps:
-
-1. **Using the Makefile: The Effortless Approach**
-   With a single command, you can have a smooth LM setup for your OpenDevin experience. Simply run:
    ```bash
    make setup-config
    ```
-   This command will prompt you to enter the LLM API key, model name, and other variables ensuring that OpenDevin is tailored to your specific needs. Note that the model name will apply only when you run headless. If you use the UI, please set the model in the UI.
+
+   This command will prompt you to enter the LLM API key, model name, and other variables ensuring that OpenHands is tailored to your specific needs. Note that the model name will apply only when you run headless. If you use the UI, please set the model in the UI.
+
+   Note: If you have previously run OpenHands using the docker command, you may have already set some environmental variables in your terminal. The final configurations are set from highest to lowest priority:
+   Environment variables > config.toml variables > default variables
 
 **Note on Alternative Models:**
-Some alternative models may prove more challenging to tame than others. Fear not, brave adventurer! We shall soon unveil LLM-specific documentation to guide you on your quest. And if you've already mastered the art of wielding a model other than OpenAI's GPT, we encourage you to [share your setup instructions with us](https://github.com/OpenDevin/OpenDevin/issues/417).
+Some alternative models may prove more challenging to tame than others. Fear not, brave adventurer! We shall soon unveil LLM-specific documentation to guide you on your quest.
+And if you've already mastered the art of wielding a model other than OpenAI's GPT, we encourage you to share your setup instructions with us by creating instructions and adding it [to our documentation](https://github.com/All-Hands-AI/OpenHands/tree/main/docs/modules/usage/llms).
 
 For a full list of the LM providers and models available, please consult the [litellm documentation](https://docs.litellm.ai/docs/providers).
 
-There is also [documentation for running with local models using ollama](./docs/documentation/LOCAL_LLM_GUIDE.md).
-
-### 4. Run the Application
-
-- **Run the Application:** Once the setup is complete, launching OpenDevin is as simple as running a single command. This command starts both the backend and frontend servers seamlessly, allowing you to interact with OpenDevin without any hassle.
-    ```bash
-    make run
-    ```
-
-### 5. Individual Server Startup
+### 4. Running the application
+#### Option A: Run the Full Application
+Once the setup is complete, launching OpenHands is as simple as running a single command. This command starts both the backend and frontend servers seamlessly, allowing you to interact with OpenHands:
+```bash
+make run
+```
 
+#### Option B: Individual Server Startup
 - **Start the Backend Server:** If you prefer, you can start the backend server independently to focus on backend-related tasks or configurations.
     ```bash
     make start-backend
@@ -73,29 +74,51 @@ There is also [documentation for running with local models using ollama](./docs/
     ```
 
 ### 6. LLM Debugging
-
-If you encounter any issues with the Language Model (LM) or you're simply curious, you can inspect the actual LLM prompts and responses. To do so, export DEBUG=1 in the environment and restart the backend. OpenDevin will then log the prompts and responses in the logs/llm/CURRENT_DATE directory, allowing you to identify the causes.
+If you encounter any issues with the Language Model (LM) or you're simply curious, you can inspect the actual LLM prompts and responses. To do so, export DEBUG=1 in the environment and restart the backend.
+OpenHands will then log the prompts and responses in the logs/llm/CURRENT_DATE directory, allowing you to identify the causes.
 
 ### 7. Help
-
-- **Get Some Help:** Need assistance or information on available targets and commands? The help command provides all the necessary guidance to ensure a smooth experience with OpenDevin.
-    ```bash
-    make help
-    ```
+Need assistance or information on available targets and commands? The help command provides all the necessary guidance to ensure a smooth experience with OpenHands.
+```bash
+make help
+ ```
 
 ### 8. Testing
-
+To run tests, refer to the following:
 #### Unit tests
 
 ```bash
-poetry run pytest ./tests/unit/test_sandbox.py
+poetry run pytest ./tests/unit/test_*.py
 ```
 
 #### Integration tests
-
 Please refer to [this README](./tests/integration/README.md) for details.
 
 ### 9. Add or update dependency
+1. Add your dependency in `pyproject.toml` or use `poetry add xxx`
+2. Update the poetry.lock file via `poetry lock --no-update`
 
-1. Add your dependency in `pyproject.toml` or use `peotry add xxx`
-2. Update the poetry.lock file via `poetry lock --no-update`
+## Develop inside Docker container
+
+TL;DR
+
+```bash
+make docker-dev
+```
+
+See more details [here](./containers/dev/README.md)
+
+If you are just interested in running `OpenHands` without installing all the required tools on your host.
+
+```bash
+make docker-run
+```
+
+If you do not have `make` on your host, run:
+
+```bash
+cd ./containers/dev
+./dev.sh
+```
+
+You do need [Docker](https://docs.docker.com/engine/install/) installed on your host though.

ISSUE_TRIAGE.md

@@ -0,0 +1,25 @@
+# Issue Triage
+These are the procedures and guidelines on how issues are triaged in this repo by the maintainers.
+
+## General
+* Most issues must be tagged with **enhancement** or **bug**
+* Issues may be tagged with what it relates to (**backend**, **frontend**, **agent quality**, etc.)
+
+## Severity
+* **Low**: Minor issues, single user report
+* **Medium**: Affecting multiple users
+* **Critical**: Affecting all users or potential security issues
+
+## Effort
+* Issues may be estimated with effort required (**small effort**, **medium effort**, **large effort**)
+
+## Difficulty
+* Issues with low implementation difficulty may be tagged with **good first issue**
+
+## Not Enough Information
+* User is asked to provide more information (logs, how to reproduce, etc.) when the issue is not clear
+* If an issue is unclear and the author does not provide more information or respond to a request, the issue may be closed as **not planned** (Usually after a week)
+
+## Multiple Requests/Fixes in One Issue
+* These issues will be narrowed down to one request/fix so the issue is more easily tracked and fixed
+* Issues may be broken down into multiple issues if required

MANIFEST.in

@@ -0,0 +1,5 @@
+# Exclude all Python bytecode files
+global-exclude *.pyc
+
+# Exclude Python cache directories
+global-exclude __pycache__

Makefile

@@ -1,15 +1,16 @@
 SHELL=/bin/bash
-# Makefile for OpenDevin project
+# Makefile for OpenHands project
 
 # Variables
-DOCKER_IMAGE = ghcr.io/opendevin/sandbox
+BACKEND_HOST ?= "127.0.0.1"
 BACKEND_PORT = 3000
-BACKEND_HOST = "127.0.0.1:$(BACKEND_PORT)"
+BACKEND_HOST_PORT = "$(BACKEND_HOST):$(BACKEND_PORT)"
 FRONTEND_PORT = 3001
 DEFAULT_WORKSPACE_DIR = "./workspace"
-DEFAULT_MODEL = "gpt-3.5-turbo"
+DEFAULT_MODEL = "gpt-4o"
 CONFIG_FILE = config.toml
-PRECOMMIT_CONFIG_PATH = "./dev_config/python/.pre-commit-config.yaml"
+PRE_COMMIT_CONFIG_PATH = "./dev_config/python/.pre-commit-config.yaml"
+PYTHON_VERSION = 3.11
 
 # ANSI color codes
 GREEN=$(shell tput -Txterm setaf 2)
@@ -22,12 +23,9 @@ RESET=$(shell tput -Txterm sgr0)
 build:
 	@echo "$(GREEN)Building project...$(RESET)"
 	@$(MAKE) -s check-dependencies
-ifeq ($(INSTALL_DOCKER),)
-	@$(MAKE) -s pull-docker-image
-endif
 	@$(MAKE) -s install-python-dependencies
 	@$(MAKE) -s install-frontend-dependencies
-	@$(MAKE) -s install-precommit-hooks
+	@$(MAKE) -s install-pre-commit-hooks
 	@$(MAKE) -s build-frontend
 	@echo "$(GREEN)Build completed successfully.$(RESET)"
 
@@ -62,10 +60,10 @@ check-system:
 
 check-python:
 	@echo "$(YELLOW)Checking Python installation...$(RESET)"
-	@if command -v python3.11 > /dev/null; then \
-		echo "$(BLUE)$(shell python3.11 --version) is already installed.$(RESET)"; \
+	@if command -v python$(PYTHON_VERSION) > /dev/null; then \
+		echo "$(BLUE)$(shell python$(PYTHON_VERSION) --version) is already installed.$(RESET)"; \
 	else \
-		echo "$(RED)Python 3.11 is not installed. Please install Python 3.11 to continue.$(RESET)"; \
+		echo "$(RED)Python $(PYTHON_VERSION) is not installed. Please install Python $(PYTHON_VERSION) to continue.$(RESET)"; \
 		exit 1; \
 	fi
 
@@ -112,37 +110,43 @@ check-poetry:
 			echo "$(BLUE)$(shell poetry --version) is already installed.$(RESET)"; \
 		else \
 			echo "$(RED)Poetry 1.8 or later is required. You can install poetry by running the following command, then adding Poetry to your PATH:"; \
-			echo "$(RED) curl -sSL https://install.python-poetry.org | python3 -$(RESET)"; \
+			echo "$(RED) curl -sSL https://install.python-poetry.org | python$(PYTHON_VERSION) -$(RESET)"; \
 			echo "$(RED)More detail here: https://python-poetry.org/docs/#installing-with-the-official-installer$(RESET)"; \
 			exit 1; \
 		fi; \
 	else \
 		echo "$(RED)Poetry is not installed. You can install poetry by running the following command, then adding Poetry to your PATH:"; \
-		echo "$(RED) curl -sSL https://install.python-poetry.org | python3.11 -$(RESET)"; \
+		echo "$(RED) curl -sSL https://install.python-poetry.org | python$(PYTHON_VERSION) -$(RESET)"; \
 		echo "$(RED)More detail here: https://python-poetry.org/docs/#installing-with-the-official-installer$(RESET)"; \
 		exit 1; \
 	fi
 
-pull-docker-image:
-	@echo "$(YELLOW)Pulling Docker image...$(RESET)"
-	@docker pull $(DOCKER_IMAGE)
-	@echo "$(GREEN)Docker image pulled successfully.$(RESET)"
-
 install-python-dependencies:
 	@echo "$(GREEN)Installing Python dependencies...$(RESET)"
-	poetry env use python3.11
+	@if [ -z "${TZ}" ]; then \
+		echo "Defaulting TZ (timezone) to UTC"; \
+		export TZ="UTC"; \
+	fi
+	poetry env use python$(PYTHON_VERSION)
 	@if [ "$(shell uname)" = "Darwin" ]; then \
 		echo "$(BLUE)Installing chroma-hnswlib...$(RESET)"; \
 		export HNSWLIB_NO_NATIVE=1; \
 		poetry run pip install chroma-hnswlib; \
 	fi
-	@poetry install
+	@poetry install --without llama-index
 	@if [ -f "/etc/manjaro-release" ]; then \
 		echo "$(BLUE)Detected Manjaro Linux. Installing Playwright dependencies...$(RESET)"; \
 		poetry run pip install playwright; \
 		poetry run playwright install chromium; \
 	else \
-		poetry run playwright install --with-deps chromium; \
+		if [ ! -f cache/playwright_chromium_is_installed.txt ]; then \
+			echo "Running playwright install --with-deps chromium..."; \
+			poetry run playwright install --with-deps chromium; \
+			mkdir -p cache; \
+			touch cache/playwright_chromium_is_installed.txt; \
+		else \
+			echo "Setup already done. Skipping playwright installation."; \
+		fi \
 	fi
 	@echo "$(GREEN)Python dependencies installed successfully.$(RESET)"
 
@@ -150,22 +154,19 @@ install-frontend-dependencies:
 	@echo "$(YELLOW)Setting up frontend environment...$(RESET)"
 	@echo "$(YELLOW)Detect Node.js version...$(RESET)"
 	@cd frontend && node ./scripts/detect-node-version.js
-	@cd frontend && \
-		echo "$(BLUE)Installing frontend dependencies with npm...$(RESET)" && \
-		npm install && \
-		echo "$(BLUE)Running make-i18n with npm...$(RESET)" && \
-		npm run make-i18n
+	echo "$(BLUE)Installing frontend dependencies with npm...$(RESET)"
+	@cd frontend && npm install
 	@echo "$(GREEN)Frontend dependencies installed successfully.$(RESET)"
 
-install-precommit-hooks:
+install-pre-commit-hooks:
 	@echo "$(YELLOW)Installing pre-commit hooks...$(RESET)"
 	@git config --unset-all core.hooksPath || true
-	@poetry run pre-commit install --config $(PRECOMMIT_CONFIG_PATH)
+	@poetry run pre-commit install --config $(PRE_COMMIT_CONFIG_PATH)
 	@echo "$(GREEN)Pre-commit hooks installed successfully.$(RESET)"
 
 lint-backend:
 	@echo "$(YELLOW)Running linters...$(RESET)"
-	@poetry run pre-commit run --files $$(git diff --name-only $$(git merge-base main $$(git branch --show-current)) $$(git branch --show-current) | tr '\n' ' ') --show-diff-on-failure --config $(PRECOMMIT_CONFIG_PATH)
+	@poetry run pre-commit run --files openhands/**/* agenthub/**/* evaluation/**/* --show-diff-on-failure --config $(PRE_COMMIT_CONFIG_PATH)
 
 lint-frontend:
 	@echo "$(YELLOW)Running linters for frontend...$(RESET)"
@@ -189,29 +190,54 @@ build-frontend:
 # Start backend
 start-backend:
 	@echo "$(YELLOW)Starting backend...$(RESET)"
-	@poetry run uvicorn opendevin.server.listen:app --port $(BACKEND_PORT) --reload --reload-exclude "workspace/*"
+	@poetry run uvicorn openhands.server.listen:app --host $(BACKEND_HOST) --port $(BACKEND_PORT) --reload --reload-exclude "$(shell pwd)/workspace"
 
 # Start frontend
 start-frontend:
 	@echo "$(YELLOW)Starting frontend...$(RESET)"
-	@cd frontend && VITE_BACKEND_HOST=$(BACKEND_HOST) VITE_FRONTEND_PORT=$(FRONTEND_PORT) npm run start
+	@cd frontend && VITE_BACKEND_HOST=$(BACKEND_HOST_PORT) VITE_FRONTEND_PORT=$(FRONTEND_PORT) npm run start
 
-# Run the app
-run:
-	@echo "$(YELLOW)Running the app...$(RESET)"
+# Common setup for running the app (non-callable)
+_run_setup:
 	@if [ "$(OS)" = "Windows_NT" ]; then \
-		echo "$(RED)`make run` is not supported on Windows. Please run `make start-frontend` and `make start-backend` separately.$(RESET)"; \
+		echo "$(RED) Windows is not supported, use WSL instead!$(RESET)"; \
 		exit 1; \
 	fi
 	@mkdir -p logs
 	@echo "$(YELLOW)Starting backend server...$(RESET)"
-	@poetry run uvicorn opendevin.server.listen:app --port $(BACKEND_PORT) &
+	@poetry run uvicorn openhands.server.listen:app --host $(BACKEND_HOST) --port $(BACKEND_PORT) &
 	@echo "$(YELLOW)Waiting for the backend to start...$(RESET)"
 	@until nc -z localhost $(BACKEND_PORT); do sleep 0.1; done
 	@echo "$(GREEN)Backend started successfully.$(RESET)"
+
+# Run the app (standard mode)
+run:
+	@echo "$(YELLOW)Running the app...$(RESET)"
+	@$(MAKE) -s _run_setup
 	@cd frontend && echo "$(BLUE)Starting frontend with npm...$(RESET)" && npm run start -- --port $(FRONTEND_PORT)
 	@echo "$(GREEN)Application started successfully.$(RESET)"
 
+# Run the app (in docker)
+docker-run: WORKSPACE_BASE ?= $(PWD)/workspace
+docker-run:
+	@if [ -f /.dockerenv ]; then \
+		echo "Running inside a Docker container. Exiting..."; \
+		exit 0; \
+	else \
+		echo "$(YELLOW)Running the app in Docker $(OPTIONS)...$(RESET)"; \
+		export WORKSPACE_BASE=${WORKSPACE_BASE}; \
+		export SANDBOX_USER_ID=$(shell id -u); \
+		export DATE=$(shell date +%Y%m%d%H%M%S); \
+		docker compose up $(OPTIONS); \
+	fi
+
+# Run the app (WSL mode)
+run-wsl:
+	@echo "$(YELLOW)Running the app in WSL mode...$(RESET)"
+	@$(MAKE) -s _run_setup
+	@cd frontend && echo "$(BLUE)Starting frontend with npm (WSL mode)...$(RESET)" && npm run dev_wsl -- --port $(FRONTEND_PORT)
+	@echo "$(GREEN)Application started successfully in WSL mode.$(RESET)"
+
 # Setup config.toml
 setup-config:
 	@echo "$(YELLOW)Setting up config.toml...$(RESET)"
@@ -222,7 +248,7 @@ setup-config:
 setup-config-prompts:
 	@echo "[core]" > $(CONFIG_FILE).tmp
 
-	@read -p "Enter your workspace directory [default: $(DEFAULT_WORKSPACE_DIR)]: " workspace_dir; \
+	@read -p "Enter your workspace directory (as absolute path) [default: $(DEFAULT_WORKSPACE_DIR)]: " workspace_dir; \
 	 workspace_dir=$${workspace_dir:-$(DEFAULT_WORKSPACE_DIR)}; \
 	 echo "workspace_base=\"$$workspace_dir\"" >> $(CONFIG_FILE).tmp
 
@@ -249,6 +275,10 @@ setup-config-prompts:
 		echo "    - nomic-embed-text"; \
 		echo "    - all-minilm"; \
 		echo "    - stable-code"; \
+		echo "    - bge-m3"; \
+		echo "    - bge-large"; \
+		echo "    - paraphrase-multilingual"; \
+		echo "    - snowflake-arctic-embed"; \
 		echo "  - Leave blank to default to 'BAAI/bge-small-en-v1.5' via huggingface"; \
 		read -p "> " llm_embedding_model; \
 		echo "embedding_model=\"$$llm_embedding_model\"" >> $(CONFIG_FILE).tmp; \
@@ -265,10 +295,20 @@ setup-config-prompts:
 		fi
 
 
+# Develop in container
+docker-dev:
+	@if [ -f /.dockerenv ]; then \
+		echo "Running inside a Docker container. Exiting..."; \
+		exit 0; \
+	else \
+		echo "$(YELLOW)Build and run in Docker $(OPTIONS)...$(RESET)"; \
+		./containers/dev/dev.sh $(OPTIONS); \
+	fi
+
 # Clean up all caches
 clean:
 	@echo "$(YELLOW)Cleaning up caches...$(RESET)"
-	@rm -rf opendevin/.cache
+	@rm -rf openhands/.cache
 	@echo "$(GREEN)Caches cleaned up successfully.$(RESET)"
 
 # Help
@@ -277,13 +317,16 @@ help:
 	@echo "Targets:"
 	@echo "  $(GREEN)build$(RESET)               - Build project, including environment setup and dependencies."
 	@echo "  $(GREEN)lint$(RESET)                - Run linters on the project."
-	@echo "  $(GREEN)setup-config$(RESET)        - Setup the configuration for OpenDevin by providing LLM API key,"
+	@echo "  $(GREEN)setup-config$(RESET)        - Setup the configuration for OpenHands by providing LLM API key,"
 	@echo "                        LLM Model name, and workspace directory."
-	@echo "  $(GREEN)start-backend$(RESET)       - Start the backend server for the OpenDevin project."
-	@echo "  $(GREEN)start-frontend$(RESET)      - Start the frontend server for the OpenDevin project."
-	@echo "  $(GREEN)run$(RESET)                 - Run the OpenDevin application, starting both backend and frontend servers."
+	@echo "  $(GREEN)start-backend$(RESET)       - Start the backend server for the OpenHands project."
+	@echo "  $(GREEN)start-frontend$(RESET)      - Start the frontend server for the OpenHands project."
+	@echo "  $(GREEN)run$(RESET)                 - Run the OpenHands application, starting both backend and frontend servers."
 	@echo "                        Backend Log file will be stored in the 'logs' directory."
+	@echo "  $(GREEN)docker-dev$(RESET)          - Build and run the OpenHands application in Docker."
+	@echo "  $(GREEN)docker-run$(RESET)          - Run the OpenHands application, starting both backend and frontend servers in Docker."
 	@echo "  $(GREEN)help$(RESET)                - Display this help message, providing information on available targets."
 
 # Phony targets
-.PHONY: build check-dependencies check-python check-npm check-docker check-poetry pull-docker-image install-python-dependencies install-frontend-dependencies install-precommit-hooks lint start-backend start-frontend run setup-config setup-config-prompts help
+.PHONY: build check-dependencies check-python check-npm check-docker check-poetry install-python-dependencies install-frontend-dependencies install-pre-commit-hooks lint start-backend start-frontend run run-wsl setup-config setup-config-prompts help
+.PHONY: docker-dev docker-run

README.md

@@ -1,118 +1,103 @@
 <a name="readme-top"></a>
 
-<!--
-*** Thanks for checking out the Best-README-Template. If you have a suggestion
-*** that would make this better, please fork the repo and create a pull request
-*** or simply open an issue with the tag "enhancement".
-*** Don't forget to give the project a star!
-*** Thanks again! Now go create something AMAZING! :D
--->
+<div align="center">
+  <img src="./docs/static/img/logo.png" alt="Logo" width="200">
+  <h1 align="center">OpenHands: Code Less, Make More</h1>
+</div>
 
-<!-- PROJECT SHIELDS -->
-<!--
-*** I'm using markdown "reference style" links for readability.
-*** Reference links are enclosed in brackets [ ] instead of parentheses ( ).
-*** See the bottom of this document for the declaration of the reference variables
-*** for contributors-url, forks-url, etc. This is an optional, concise syntax you may use.
-*** https://www.markdownguide.org/basic-syntax/#reference-style-links
--->
 
 <div align="center">
-  <a href="https://github.com/OpenDevin/OpenDevin/graphs/contributors"><img src="https://img.shields.io/github/contributors/opendevin/opendevin?style=for-the-badge&color=blue" alt="Contributors"></a>
-  <a href="https://github.com/OpenDevin/OpenDevin/network/members"><img src="https://img.shields.io/github/forks/opendevin/opendevin?style=for-the-badge&color=blue" alt="Forks"></a>
-  <a href="https://github.com/OpenDevin/OpenDevin/stargazers"><img src="https://img.shields.io/github/stars/opendevin/opendevin?style=for-the-badge&color=blue" alt="Stargazers"></a>
-  <a href="https://github.com/OpenDevin/OpenDevin/issues"><img src="https://img.shields.io/github/issues/opendevin/opendevin?style=for-the-badge&color=blue" alt="Issues"></a>
-  <a href="https://github.com/OpenDevin/OpenDevin/blob/main/LICENSE"><img src="https://img.shields.io/github/license/opendevin/opendevin?style=for-the-badge&color=blue" alt="MIT License"></a>
+  <a href="https://github.com/All-Hands-AI/OpenHands/graphs/contributors"><img src="https://img.shields.io/github/contributors/All-Hands-AI/OpenHands?style=for-the-badge&color=blue" alt="Contributors"></a>
+  <a href="https://github.com/All-Hands-AI/OpenHands/stargazers"><img src="https://img.shields.io/github/stars/All-Hands-AI/OpenHands?style=for-the-badge&color=blue" alt="Stargazers"></a>
+  <a href="https://codecov.io/github/All-Hands-AI/OpenHands?branch=main"><img alt="CodeCov" src="https://img.shields.io/codecov/c/github/All-Hands-AI/OpenHands?style=for-the-badge&color=blue"></a>
+  <a href="https://github.com/All-Hands-AI/OpenHands/blob/main/LICENSE"><img src="https://img.shields.io/github/license/All-Hands-AI/OpenHands?style=for-the-badge&color=blue" alt="MIT License"></a>
   <br/>
-  <a href="https://join.slack.com/t/opendevin/shared_invite/zt-2i1iqdag6-bVmvamiPA9EZUu7oCO6KhA"><img src="https://img.shields.io/badge/Slack-Join%20Us-red?logo=slack&logoColor=white&style=for-the-badge" alt="Join our Slack community"></a>
+  <a href="https://join.slack.com/t/opendevin/shared_invite/zt-2oikve2hu-UDxHeo8nsE69y6T7yFX_BA"><img src="https://img.shields.io/badge/Slack-Join%20Us-red?logo=slack&logoColor=white&style=for-the-badge" alt="Join our Slack community"></a>
   <a href="https://discord.gg/ESHStjSjD4"><img src="https://img.shields.io/badge/Discord-Join%20Us-purple?logo=discord&logoColor=white&style=for-the-badge" alt="Join our Discord community"></a>
+  <a href="https://github.com/All-Hands-AI/OpenHands/blob/main/CREDITS.md"><img src="https://img.shields.io/badge/Project-Credits-blue?style=for-the-badge&color=FFE165&logo=github&logoColor=white" alt="Credits"></a>
   <br/>
-  <a href="https://huggingface.co/spaces/OpenDevin/evaluation"><img src="https://img.shields.io/badge/SWE--bench%20Lite-25.0%25-green?style=for-the-badge" alt="SWE-bench "></a>
-  <a href="https://codecov.io/github/opendevin/opendevin?branch=main"><img alt="CodeCov" src="https://img.shields.io/codecov/c/github/opendevin/opendevin?style=for-the-badge"></a>
+  <a href="https://docs.all-hands.dev/modules/usage/getting-started"><img src="https://img.shields.io/badge/Documentation-000?logo=googledocs&logoColor=FFE165&style=for-the-badge" alt="Check out the documentation"></a>
+  <a href="https://arxiv.org/abs/2407.16741"><img src="https://img.shields.io/badge/Paper%20on%20Arxiv-000?logoColor=FFE165&logo=arxiv&style=for-the-badge" alt="Paper on Arxiv"></a>
+  <a href="https://huggingface.co/spaces/OpenHands/evaluation"><img src="https://img.shields.io/badge/Benchmark%20score-000?logoColor=FFE165&logo=huggingface&style=for-the-badge" alt="Evaluation Benchmark Score"></a>
+  <hr>
 </div>
 
-<!-- PROJECT LOGO -->
-<div align="center">
-  <img src="./docs/static/img/logo.png" alt="Logo" width="200" height="200">
-  <h1 align="center">OpenDevin: Code Less, Make More</h1>
-  <a href="https://opendevin.github.io/OpenDevin/"><img src="https://img.shields.io/badge/Documentation-OpenDevin-blue?logo=googledocs&logoColor=white&style=for-the-badge" alt="Check out the documentation"></a>
-</div>
-<hr>
+Welcome to OpenHands (formerly OpenDevin), a platform for software development agents powered by AI.
 
-Welcome to OpenDevin, a platform for autonomous software engineers, powered by AI and LLMs.
+OpenHands agents can do anything a human developer can: modify code, run commands, browse the web,
+call APIs, and yes—even copy code snippets from StackOverflow.
 
-OpenDevin agents collaborate with human developers to write code, fix bugs, and ship features.
+Learn more at [docs.all-hands.dev](https://docs.all-hands.dev), or jump to the [Quick Start](#-quick-start).
 
 ![App screenshot](./docs/static/img/screenshot.png)
 
-## ⚡ Getting Started
-The easiest way to run OpenDevin is inside a Docker container. It works best with the most recent version of Docker, `26.0.0`.
-You must be using Linux, Mac OS, or WSL on Windows.
+## ⚡ Quick Start
 
-To start the app, run these commands, replacing `$(pwd)/workspace` with the directory you want OpenDevin to work with.
+The easiest way to run OpenHands is in Docker. You can change `WORKSPACE_BASE` below to
+point OpenHands to existing code that you'd like to modify.
+
+See the [Getting Started](https://docs.all-hands.dev/modules/usage/getting-started) guide for
+system requirements and more information.
 
 ```bash
-# The directory you want OpenDevin to work with. MUST be an absolute path!
-export WORKSPACE_BASE=$(pwd)/workspace;
-```
+export WORKSPACE_BASE=$(pwd)/workspace
 
-> [!WARNING]  
-> OpenDevin runs bash commands within a Docker sandbox, so it should not affect your machine. 
-> But your workspace directory will be attached to that sandbox, and files in the directory may be modified or deleted.
-
-```bash
-docker run \
-    -it \
-    --pull=always \
+docker run -it --pull=always \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=ghcr.io/all-hands-ai/runtime:0.9-nikolaik \
     -e SANDBOX_USER_ID=$(id -u) \
     -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
     -v $WORKSPACE_BASE:/opt/workspace_base \
     -v /var/run/docker.sock:/var/run/docker.sock \
     -p 3000:3000 \
     --add-host host.docker.internal:host-gateway \
-    ghcr.io/opendevin/opendevin:0.5
+    --name openhands-app-$(date +%Y%m%d%H%M%S) \
+    ghcr.io/all-hands-ai/openhands:0.9
 ```
 
-You'll find OpenDevin running at [http://localhost:3000](http://localhost:3000).
+You'll find OpenHands running at [http://localhost:3000](http://localhost:3000)!
 
-## 🚀 Documentation
+You can also run OpenHands in a scriptable [headless mode](https://docs.all-hands.dev/modules/usage/how-to/headless-mode),
+or as an [interactive CLI](https://docs.all-hands.dev/modules/usage/how-to/cli-mode).
 
-To learn more about the project, and for tips on using OpenDevin,
-**check out our [documentation](https://opendevin.github.io/OpenDevin/)**.
+Visit [Getting Started](https://docs.all-hands.dev/modules/usage/getting-started) for more information and setup instructions.
 
-There you'll find resources on how to use different LLM providers (like ollama and Anthropic's Claude),
+If you want to modify the OpenHands source code, check out [Development.md](https://github.com/All-Hands-AI/OpenHands/blob/main/Development.md).
+
+Having issues? The [Troubleshooting Guide](https://docs.all-hands.dev/modules/usage/troubleshooting) can help.
+
+## 📖 Documentation
+
+To learn more about the project, and for tips on using OpenHands,
+**check out our [documentation](https://docs.all-hands.dev/modules/usage/getting-started)**.
+
+There you'll find resources on how to use different LLM providers,
 troubleshooting resources, and advanced configuration options.
 
 ## 🤝 How to Contribute
 
-OpenDevin is a community-driven project, and we welcome contributions from everyone.
+OpenHands is a community-driven project, and we welcome contributions from everyone.
 Whether you're a developer, a researcher, or simply enthusiastic about advancing the field of
 software engineering with AI, there are many ways to get involved:
 
 - **Code Contributions:** Help us develop new agents, core functionality, the frontend and other interfaces, or sandboxing solutions.
 - **Research and Evaluation:** Contribute to our understanding of LLMs in software engineering, participate in evaluating the models, or suggest improvements.
-- **Feedback and Testing:** Use the OpenDevin toolset, report bugs, suggest features, or provide feedback on usability.
+- **Feedback and Testing:** Use the OpenHands toolset, report bugs, suggest features, or provide feedback on usability.
 
 For details, please check [CONTRIBUTING.md](./CONTRIBUTING.md).
 
 ## 🤖 Join Our Community
 
-Whether you're a developer, a researcher, or simply enthusiastic about OpenDevin, we'd love to have you in our community.
+Whether you're a developer, a researcher, or simply enthusiastic about OpenHands, we'd love to have you in our community.
 Let's make software engineering better together!
 
-- [Slack workspace](https://join.slack.com/t/opendevin/shared_invite/zt-2ggtwn3k5-PvAA2LUmqGHVZ~XzGq~ILw) - Here we talk about research, architecture, and future development.
+- [Slack workspace](https://join.slack.com/t/opendevin/shared_invite/zt-2oikve2hu-UDxHeo8nsE69y6T7yFX_BA) - Here we talk about research, architecture, and future development.
 - [Discord server](https://discord.gg/ESHStjSjD4) - This is a community-run server for general discussion, questions, and feedback.
 
 ## 📈 Progress
-<p align="center">
-    <a href="https://www.swebench.com/lite.html">
-        <img src="/docs/static/img/results.png" alt="SWE-Bench Lite Score" width="500" height="auto">
-    </a>
-</p>
 
 <p align="center">
-  <a href="https://star-history.com/#OpenDevin/OpenDevin&Date">
-    <img src="https://api.star-history.com/svg?repos=OpenDevin/OpenDevin&type=Date" width="500" alt="Star History Chart">
+  <a href="https://star-history.com/#All-Hands-AI/OpenHands&Date">
+    <img src="https://api.star-history.com/svg?repos=All-Hands-AI/OpenHands&type=Date" width="500" alt="Star History Chart">
   </a>
 </p>
 
@@ -120,13 +105,22 @@ Let's make software engineering better together!
 
 Distributed under the MIT License. See [`LICENSE`](./LICENSE) for more information.
 
-[contributors-shield]: https://img.shields.io/github/contributors/opendevin/opendevin?style=for-the-badge
-[contributors-url]: https://github.com/OpenDevin/OpenDevin/graphs/contributors
-[forks-shield]: https://img.shields.io/github/forks/opendevin/opendevin?style=for-the-badge
-[forks-url]: https://github.com/OpenDevin/OpenDevin/network/members
-[stars-shield]: https://img.shields.io/github/stars/opendevin/opendevin?style=for-the-badge
-[stars-url]: https://github.com/OpenDevin/OpenDevin/stargazers
-[issues-shield]: https://img.shields.io/github/issues/opendevin/opendevin?style=for-the-badge
-[issues-url]: https://github.com/OpenDevin/OpenDevin/issues
-[license-shield]: https://img.shields.io/github/license/opendevin/opendevin?style=for-the-badge
-[license-url]: https://github.com/OpenDevin/OpenDevin/blob/main/LICENSE
+## 🙏 Acknowledgements
+
+OpenHands is built by a large number of contributors, and every contribution is greatly appreciated! We also build upon other open source projects, and we are deeply thankful for their work.
+
+For a list of open source projects and licenses used in OpenHands, please see our [CREDITS.md](./CREDITS.md) file.
+
+## 📚 Cite
+
+```
+@misc{opendevin,
+      title={{OpenDevin: An Open Platform for AI Software Developers as Generalist Agents}},
+      author={Xingyao Wang and Boxuan Li and Yufan Song and Frank F. Xu and Xiangru Tang and Mingchen Zhuge and Jiayi Pan and Yueqi Song and Bowen Li and Jaskirat Singh and Hoang H. Tran and Fuqiang Li and Ren Ma and Mingzhang Zheng and Bill Qian and Yanjun Shao and Niklas Muennighoff and Yizhe Zhang and Binyuan Hui and Junyang Lin and Robert Brennan and Hao Peng and Heng Ji and Graham Neubig},
+      year={2024},
+      eprint={2407.16741},
+      archivePrefix={arXiv},
+      primaryClass={cs.SE},
+      url={https://arxiv.org/abs/2407.16741},
+}
+```

agenthub/README.md

@@ -1,66 +1,81 @@
-# Agent Framework Research
+# Agent Hub
 
 In this folder, there may exist multiple implementations of `Agent` that will be used by the framework.
 
-For example, `agenthub/monologue_agent`, `agenthub/metagpt_agent`, `agenthub/codeact_agent`, etc.
+For example, `agenthub/codeact_agent`, etc.
 Contributors from different backgrounds and interests can choose to contribute to any (or all!) of these directions.
 
 ## Constructing an Agent
 
-The abstraction for an agent can be found [here](../opendevin/agent.py).
+The abstraction for an agent can be found [here](../openhands/controller/agent.py).
 
 Agents are run inside of a loop. At each iteration, `agent.step()` is called with a
-[State](../opendevin/state.py) input, and the agent must output an [Action](../opendevin/action).
+[State](../openhands/controller/state/state.py) input, and the agent must output an [Action](../openhands/events/action).
 
 Every agent also has a `self.llm` which it can use to interact with the LLM configured by the user.
 See the [LiteLLM docs for `self.llm.completion`](https://docs.litellm.ai/docs/completion).
 
 ## State
 
-The `state` contains:
+The `state` represents the running state of an agent in the OpenHands system. The class handles saving and restoring the agent session. It is serialized in a pickle.
 
-- A history of actions taken by the agent, as well as any observations (e.g. file content, command output) from those actions
-- A list of actions/observations that have happened since the most recent step
-- A [`root_task`](https://github.com/OpenDevin/OpenDevin/blob/main/opendevin/controller/state/task.py), which contains a plan of action
-  - The agent can add and modify subtasks through the `AddTaskAction` and `ModifyTaskAction`
+The State object stores information about:
+
+* Multi-agent state / delegates:
+  * the 'root task' (conversation between the agent and the user)
+  * the subtask (conversation between an agent and the user or another agent)
+  * global and local iterations
+  * delegate levels for multi-agent interactions
+  * almost stuck state
+* Running state of an agent:
+  * current agent state (e.g., LOADING, RUNNING, PAUSED)
+  * traffic control state for rate limiting
+  * confirmation mode
+  * the last error encountered
+* History:
+  * start and end IDs for events in agent's history. This allows to retrieve the actions taken by the agent, and observations (e.g. file content, command output) from the current or past sessions.
+* Metrics:
+  * global metrics for the current task
+  * local metrics for the current subtask
+* Extra data:
+  * additional task-specific data
+
+The agent can add and modify subtasks through the `AddTaskAction` and `ModifyTaskAction`
 
 ## Actions
 
 Here is a list of available Actions, which can be returned by `agent.step()`:
 
-- [`CmdRunAction`](../opendevin/action/bash.py) - Runs a command inside a sandboxed terminal
-- [`CmdKillAction`](../opendevin/action/bash.py) - Kills a background command
-- [`IPythonRunCellAction`](../opendevin/action/bash.py) - Execute a block of Python code interactively (in Jupyter notebook) and receives `CmdOutputObservation`. Requires setting up `jupyter` [plugin](../opendevin/sandbox/plugins) as a requirement.
-- [`FileReadAction`](../opendevin/action/fileop.py) - Reads the content of a file
-- [`FileWriteAction`](../opendevin/action/fileop.py) - Writes new content to a file
-- [`BrowseURLAction`](../opendevin/action/browse.py) - Gets the content of a URL
-- [`AgentRecallAction`](../opendevin/action/agent.py) - Searches memory (e.g. a vector database)
-- [`AddTaskAction`](../opendevin/action/tasks.py) - Adds a subtask to the plan
-- [`ModifyTaskAction`](../opendevin/action/tasks.py) - Changes the state of a subtask
-- [`AgentThinkAction`](../opendevin/action/agent.py) - A no-op that allows the agent to add plaintext to the history (as well as the chat log)
-- [`AgentTalkAction`](../opendevin/action/agent.py) - A no-op that allows the agent to add plaintext to the history and talk to the user.
-- [`AgentFinishAction`](../opendevin/action/agent.py) - Stops the control loop, allowing the user/delegator agent to enter a new task
-- [`AgentRejectAction`](../opendevin/action/agent.py) - Stops the control loop, allowing the user/delegator agent to enter a new task
-- [`AgentFinishAction`](../opendevin/action/agent.py) - Stops the control loop, allowing the user to enter a new task
-- [`MessageAction`](../opendevin/action/message.py) - Represents a message from an agent or the user
+- [`CmdRunAction`](../openhands/events/action/commands.py) - Runs a command inside a sandboxed terminal
+- [`IPythonRunCellAction`](../openhands/events/action/commands.py) - Execute a block of Python code interactively (in Jupyter notebook) and receives `CmdOutputObservation`. Requires setting up `jupyter` [plugin](../openhands/runtime/plugins) as a requirement.
+- [`FileReadAction`](../openhands/events/action/files.py) - Reads the content of a file
+- [`FileWriteAction`](../openhands/events/action/files.py) - Writes new content to a file
+- [`BrowseURLAction`](../openhands/events/action/browse.py) - Gets the content of a URL
+- [`AddTaskAction`](../openhands/events/action/tasks.py) - Adds a subtask to the plan
+- [`ModifyTaskAction`](../openhands/events/action/tasks.py) - Changes the state of a subtask.
+- [`AgentFinishAction`](../openhands/events/action/agent.py) - Stops the control loop, allowing the user/delegator agent to enter a new task
+- [`AgentRejectAction`](../openhands/events/action/agent.py) - Stops the control loop, allowing the user/delegator agent to enter a new task
+- [`AgentFinishAction`](../openhands/events/action/agent.py) - Stops the control loop, allowing the user to enter a new task
+- [`MessageAction`](../openhands/events/action/message.py) - Represents a message from an agent or the user
 
-You can use `action.to_dict()` and `action_from_dict` to serialize and deserialize actions.
+To serialize and deserialize an action, you can use:
+- `action.to_dict()` to serialize the action to a dictionary to be sent to the UI, including a user-friendly string representation of the message
+- `action.to_memory()` to serialize the action to a dictionary to be sent to the LLM. It may include raw information, such as the underlying exceptions that occurred during the action execution.
+- `action_from_dict(action_dict)` to deserialize the action from a dictionary.
 
 ## Observations
 
 There are also several types of Observations. These are typically available in the step following the corresponding Action.
-But they may also appear as a result of asynchronous events (e.g. a message from the user, logs from a command running
-in the background).
+But they may also appear as a result of asynchronous events (e.g. a message from the user).
 
 Here is a list of available Observations:
 
-- [`CmdOutputObservation`](../opendevin/observation/run.py)
-- [`BrowserOutputObservation`](../opendevin/observation/browse.py)
-- [`FileReadObservation`](../opendevin/observation/files.py)
-- [`FileWriteObservation`](../opendevin/observation/files.py)
-- [`AgentRecallObservation`](../opendevin/observation/recall.py)
-- [`ErrorObservation`](../opendevin/observation/error.py)
-- [`SuccessObservation`](../opendevin/observation/success.py)
+- [`CmdOutputObservation`](../openhands/events/observation/commands.py)
+- [`BrowserOutputObservation`](../openhands/events/observation/browse.py)
+- [`FileReadObservation`](../openhands/events/observation/files.py)
+- [`FileWriteObservation`](../openhands/events/observation/files.py)
+- [`ErrorObservation`](../openhands/events/observation/error.py)
+- [`SuccessObservation`](../openhands/events/observation/success.py)
 
 You can use `observation.to_dict()` and `observation_from_dict` to serialize and deserialize observations.
 
@@ -77,13 +92,51 @@ def step(self, state: "State") -> "Action"
 `step` moves the agent forward one step towards its goal. This probably means
 sending a prompt to the LLM, then parsing the response into an `Action`.
 
-### `search_memory`
+## Agent Delegation
+
+OpenHands is a multi-agentic system. Agents can delegate tasks to other agents, whether
+prompted by the user, or when the agent decides to ask another agent for help. For example,
+the `CodeActAgent` might delegate to the `BrowsingAgent` to answer questions that involve browsing
+the web. The Delegator Agent forwards tasks to micro-agents, such as 'RepoStudyAgent' to study a repo,
+or 'VerifierAgent' to verify a task completion.
+
+### Understanding the terminology
+
+A `task` is an end-to-end conversation between OpenHands (the whole system) and the user,
+which might involve one or more inputs from the user. It starts with an initial input
+(typically a task statement) from the user, and ends with either an `AgentFinishAction`
+initiated by the agent, a stop initiated by the user, or an error.
+
+A `subtask` is an end-to-end conversation between an agent and the user, or
+another agent. If a `task` is conducted by a single agent, then it's also a `subtask`
+itself. Otherwise, a `task` consists of multiple `subtasks`, each executed by
+one agent.
+
+For example, considering a task from the user: `tell me how many GitHub stars
+OpenHands repo has`. Let's assume the default agent is CodeActAgent.
 
 ```
-def search_memory(self, query: str) -> list[str]:
+-- TASK STARTS (SUBTASK 0 STARTS) --
+
+DELEGATE_LEVEL 0, ITERATION 0, LOCAL_ITERATION 0
+CodeActAgent: I should request help from BrowsingAgent
+
+-- DELEGATE STARTS (SUBTASK 1 STARTS) --
+
+DELEGATE_LEVEL 1, ITERATION 1, LOCAL_ITERATION 0
+BrowsingAgent: Let me find the answer on GitHub
+
+DELEGATE_LEVEL 1, ITERATION 2, LOCAL_ITERATION 1
+BrowsingAgent: I found the answer, let me convey the result and finish
+
+-- DELEGATE ENDS (SUBTASK 1 ENDS) --
+
+DELEGATE_LEVEL 0, ITERATION 3, LOCAL_ITERATION 1
+CodeActAgent: I got the answer from BrowsingAgent, let me convey the result
+and finish
+
+-- TASK ENDS (SUBTASK 0 ENDS) --
 ```
 
-`search_memory` should return a list of events that match the query. This will be used
-for the `recall` action.
-
-You can optionally just return `[]` for this method, meaning the agent has no long-term memory.
+Note how ITERATION counter is shared across agents, while LOCAL_ITERATION
+is local to each subtask.

agenthub/SWE_agent/Examples.md

@@ -1,405 +0,0 @@
-# GPT-4 Examples:
-
-### Notes:
-- The final result is not completed as I only gave it 10 steps, but shows very promising use of the tools.
-- This whole run costs: $0.35 using gpt-4-turbo
-  - Input Tokens: 29,734
-  - Completion Tokens: 1,636
-
-Interesting things:
-1. Using comments as an initial placeholder for the next move and then having that output in memory is a very cool behavior.
-2. It makes very small iterative changes due to the prompting.
-3. A lot of the logic for the game is present but not put together well
-4. There are a lot of odd quirks with the code that it has written.
-5. Some odd agent behavior with putting random lines of code in random places?
-6. The structure of the feedback loop needs some improvement so the agent can see the changes it has made better in context.
-7. Might be beneficial to add in some line numbers so the model knows what lines it is looking at.
-
-
-```
-==============
-STEP 0
-
-PLAN:
-Please make conway's game of life in python using the pygame library. Be sure to use the tools in your workspace to help you out.
-Usage(completion_tokens=91, prompt_tokens=2047, total_tokens=2138)
-
-==== RAW OUTPUT ====
-Thoughts:
-To start implementing Conway's Game of Life in Python using the pygame library, I first need to set up a Python script file where I can write the code. I'll create a new Python file named `game_of_life.py`. This file will serve as the base for writing the game logic and using pygame functionalities.
-
-Action:
-"""
-write game_of_life.py # This is a placeholder for the game of life implementation using pygame.
-"""
-
-==== END RAW ====
-
-
-ACTION:
-FileWriteAction(path='game_of_life.py', content='# This is a placeholder for the game of life implementation using pygame.', start=0, end=-1, thoughts="Thoughts:\nTo start implementing Conway's Game of Life in Python using the pygame library, I first need to set up a Python script file where I can write the code. I'll create a new Python file named `game_of_life.py`. This file will serve as the base for writing the game logic and using pygame functionalities.\n\nAction:", action=<ActionType.WRITE: 'write'>)
-
-OBSERVATION:
-WRITE OPERATION:
-You have written to "game_of_life.py" on these lines: 0:-1.
-
-
-==============
-STEP 1
-
-PLAN:
-Please make conway's game of life in python using the pygame library. Be sure to use the tools in your workspace to help you out.
-Usage(completion_tokens=101, prompt_tokens=2351, total_tokens=2452)
-
-==== RAW OUTPUT ====
-Thoughts:
-The placeholder for the game of life implementation using pygame has been set up in the file `game_of_life.py`. The next step is to start implementing the basic structure of the game. This includes importing necessary libraries, initializing the pygame module, setting up the display window, and defining the main game loop. I will begin by importing pygame and initializing it, which are the foundational steps for any pygame application.
-
-Action:
-"""
-edit 0 0 import pygame
-
-pygame.init()
-"""
-
-==== END RAW ====
-
-
-ACTION:
-FileWriteAction(path='game_of_life.py', content='import pygame\n\npygame.init()', start=0, end=0, thoughts='Thoughts:\nThe placeholder for the game of life implementation using pygame has been set up in the file `game_of_life.py`. The next step is to start implementing the basic structure of the game. This includes importing necessary libraries, initializing the pygame module, setting up the display window, and defining the main game loop. I will begin by importing pygame and initializing it, which are the foundational steps for any pygame application.\n\nAction:', action=<ActionType.WRITE: 'write'>)
-
-OBSERVATION:
-WRITE OPERATION:
-You have written to "game_of_life.py" on these lines: 0:0.
-
-
-==============
-STEP 2
-
-PLAN:
-Please make conway's game of life in python using the pygame library. Be sure to use the tools in your workspace to help you out.
-Usage(completion_tokens=123, prompt_tokens=2552, total_tokens=2675)
-
-==== RAW OUTPUT ====
-Thoughts:
-Previously, I initialized pygame in the `game_of_life.py` file. The next step is to set up the display window for the game. This involves defining the size of the window and setting the title. This is crucial as it provides the visual interface for the game. I will also add a comment to indicate that this is the setup for the display window.
-
-Action:
-"""
-edit 2 2
-# Setup the display window
-screen_size = (800, 600)
-screen = pygame.display.set_mode(screen_size)
-pygame.display.set_caption("Conway's Game of Life")
-"""
-==== END RAW ====
-
-
-ACTION:
-FileWriteAction(path='game_of_life.py', content='# Setup the display window\nscreen_size = (800, 600)\nscreen = pygame.display.set_mode(screen_size)\npygame.display.set_caption("Conway\'s Game of Life")', start=2, end=2, thoughts='Thoughts:\nPreviously, I initialized pygame in the `game_of_life.py` file. The next step is to set up the display window for the game. This involves defining the size of the window and setting the title. This is crucial as it provides the visual interface for the game. I will also add a comment to indicate that this is the setup for the display window.\n\nAction:', action=<ActionType.WRITE: 'write'>)
-
-OBSERVATION:
-WRITE OPERATION:
-You have written to "game_of_life.py" on these lines: 2:2.
-
-
-==============
-STEP 3
-
-PLAN:
-Please make conway's game of life in python using the pygame library. Be sure to use the tools in your workspace to help you out.
-Usage(completion_tokens=144, prompt_tokens=2777, total_tokens=2921)
-
-==== RAW OUTPUT ====
-Thoughts:
-So far, I have set up the basic structure for Conway's Game of Life using pygame, including initializing pygame and setting up the display window. The next step is to define the main game loop where the game's logic will be continuously updated and rendered. This loop will handle events, update the game state, and redraw the screen. I will start by defining the main loop structure and a basic event handling mechanism to allow for quitting the game when the user presses the close button.
-
-Action:
-"""
-edit 5 5 # Main game loop
-running = True
-while running:
-    for event in pygame.event.get():
-        if event.type == pygame.QUIT:
-            running = False
-"""
-
-==== END RAW ====
-
-
-ACTION:
-FileWriteAction(path='game_of_life.py', content='# Main game loop\nrunning = True\nwhile running:\n    for event in pygame.event.get():\n        if event.type == pygame.QUIT:\n            running = False', start=5, end=5, thoughts="Thoughts:\nSo far, I have set up the basic structure for Conway's Game of Life using pygame, including initializing pygame and setting up the display window. The next step is to define the main game loop where the game's logic will be continuously updated and rendered. This loop will handle events, update the game state, and redraw the screen. I will start by defining the main loop structure and a basic event handling mechanism to allow for quitting the game when the user presses the close button.\n\nAction:", action=<ActionType.WRITE: 'write'>)
-
-OBSERVATION:
-WRITE OPERATION:
-You have written to "game_of_life.py" on these lines: 5:5.
-
-
-==============
-STEP 4
-
-PLAN:
-Please make conway's game of life in python using the pygame library. Be sure to use the tools in your workspace to help you out.
-Usage(completion_tokens=222, prompt_tokens=3022, total_tokens=3244)
-
-==== RAW OUTPUT ====
-Thoughts:
-So far, I have set up the basic structure for Conway's Game of Life using pygame, including initializing pygame, setting up the display window, and defining the main game loop. The next step is to implement the logic for Conway's Game of Life. This involves creating a grid to represent the game state, initializing it with a random pattern, and defining the rules for cell survival, birth, and death. I will start by defining the grid size and initializing the grid with random states.
-
-Action:
-"""
-edit 7 7 # Define grid size and initialize grid
-grid_size = (50, 50)  # Grid size as a tuple (rows, columns)
-grid = [[0 for x in range(grid_size[1])] for y in range(grid_size[0])]  # Initialize grid with zeros
-import random
-for y in range(grid_size[0]):
-    for x in range(grid_size[1]):
-        grid[y][x] = random.randint(0, 1)  # Randomly assign cells as alive (1) or dead (0)
-"""
-==== END RAW ====
-
-
-ACTION:
-FileWriteAction(path='game_of_life.py', content='# Define grid size and initialize grid\ngrid_size = (50, 50)  # Grid size as a tuple (rows, columns)\ngrid = [[0 for x in range(grid_size[1])] for y in range(grid_size[0])]  # Initialize grid with zeros\nimport random\nfor y in range(grid_size[0]):\n    for x in range(grid_size[1]):\n        grid[y][x] = random.randint(0, 1)  # Randomly assign cells as alive (1) or dead (0)', start=7, end=7, thoughts="Thoughts:\nSo far, I have set up the basic structure for Conway's Game of Life using pygame, including initializing pygame, setting up the display window, and defining the main game loop. The next step is to implement the logic for Conway's Game of Life. This involves creating a grid to represent the game state, initializing it with a random pattern, and defining the rules for cell survival, birth, and death. I will start by defining the grid size and initializing the grid with random states.\n\nAction:", action=<ActionType.WRITE: 'write'>)
-
-OBSERVATION:
-WRITE OPERATION:
-You have written to "game_of_life.py" on these lines: 7:7.
-
-
-==============
-STEP 5
-
-PLAN:
-Please make conway's game of life in python using the pygame library. Be sure to use the tools in your workspace to help you out.
-Usage(completion_tokens=234, prompt_tokens=3158, total_tokens=3392)
-
-==== RAW OUTPUT ====
-Thoughts:
-So far, I have set up the basic structure for Conway's Game of Life using pygame, including initializing pygame, setting up the display window, defining the main game loop, and initializing the grid with random states. The next step is to implement the rules for Conway's Game of Life. This involves checking each cell's neighbors to determine its next state based on the number of alive neighbors. I will start by defining a function to count the alive neighbors for a given cell.
-
-Action:
-"""
-edit 9 9 def count_alive_neighbors(grid, x, y):
-    directions = [(-1, -1), (-1, 0), (-1, 1), (0, -1), (0, 1), (1, -1), (1, 0), (1, 1)]
-    count = 0
-    for dx, dy in directions:
-        nx, ny = x + dx, y + dy
-        if 0 <= nx < grid_size[1] and 0 <= ny < grid_size[0]:
-            count += grid[ny][nx]
-    return count
-"""
-
-==== END RAW ====
-
-
-ACTION:
-FileWriteAction(path='game_of_life.py', content='def count_alive_neighbors(grid, x, y):\n    directions = [(-1, -1), (-1, 0), (-1, 1), (0, -1), (0, 1), (1, -1), (1, 0), (1, 1)]\n    count = 0\n    for dx, dy in directions:\n        nx, ny = x + dx, y + dy\n        if 0 <= nx < grid_size[1] and 0 <= ny < grid_size[0]:\n            count += grid[ny][nx]\n    return count', start=9, end=9, thoughts="Thoughts:\nSo far, I have set up the basic structure for Conway's Game of Life using pygame, including initializing pygame, setting up the display window, defining the main game loop, and initializing the grid with random states. The next step is to implement the rules for Conway's Game of Life. This involves checking each cell's neighbors to determine its next state based on the number of alive neighbors. I will start by defining a function to count the alive neighbors for a given cell.\n\nAction:", action=<ActionType.WRITE: 'write'>)
-
-OBSERVATION:
-WRITE OPERATION:
-You have written to "game_of_life.py" on these lines: 9:9.
-
-
-==============
-STEP 6
-
-PLAN:
-Please make conway's game of life in python using the pygame library. Be sure to use the tools in your workspace to help you out.
-Usage(completion_tokens=257, prompt_tokens=3290, total_tokens=3547)
-
-==== RAW OUTPUT ====
-Thoughts:
-So far, I have set up the basic structure for Conway's Game of Life using pygame, including initializing pygame, setting up the display window, defining the main game loop, initializing the grid with random states, and creating a function to count alive neighbors. The next step is to implement the rules for Conway's Game of Life. This involves updating the grid state based on the number of alive neighbors each cell has. I will define a function `update_grid` that applies the rules of the game to update the grid for the next generation.
-
-Action:
-"""
-write game_of_life.py def update_grid(grid):
-    new_grid = [[0 for x in range(grid_size[1])] for y in range(grid_size[0])]
-    for y in range(grid_size[0]):
-        for x in range(grid_size[1]):
-            alive_neighbors = count_alive_neighbors(grid, x, y)
-            if grid[y][x] == 1 and (alive_neighbors == 2 or alive_neighbors == 3):
-                new_grid[y][x] = 1
-            elif grid[y][x] == 0 and alive_neighbors == 3:
-                new_grid[y][x] = 1
-    return new_grid
- 11
- 11
-"""
-==== END RAW ====
-
-
-ACTION:
-FileWriteAction(path='game_of_life.py', content='def update_grid(grid):\n    new_grid = [[0 for x in range(grid_size[1])] for y in range(grid_size[0])]\n    for y in range(grid_size[0]):\n        for x in range(grid_size[1]):\n            alive_neighbors = count_alive_neighbors(grid, x, y)\n            if grid[y][x] == 1 and (alive_neighbors == 2 or alive_neighbors == 3):\n                new_grid[y][x] = 1\n            elif grid[y][x] == 0 and alive_neighbors == 3:\n                new_grid[y][x] = 1\n    return new_grid', start=11, end=11, thoughts="Thoughts:\nSo far, I have set up the basic structure for Conway's Game of Life using pygame, including initializing pygame, setting up the display window, defining the main game loop, initializing the grid with random states, and creating a function to count alive neighbors. The next step is to implement the rules for Conway's Game of Life. This involves updating the grid state based on the number of alive neighbors each cell has. I will define a function `update_grid` that applies the rules of the game to update the grid for the next generation.\n\nAction:", action=<ActionType.WRITE: 'write'>)
-
-OBSERVATION:
-WRITE OPERATION:
-You have written to "game_of_life.py" on these lines: 11:11.
-
-
-==============
-STEP 7
-
-PLAN:
-Please make conway's game of life in python using the pygame library. Be sure to use the tools in your workspace to help you out.
-Usage(completion_tokens=124, prompt_tokens=3416, total_tokens=3540)
-
-==== RAW OUTPUT ====
-Thoughts:
-So far, I have set up the basic structure for Conway's Game of Life using pygame, including initializing pygame, setting up the display window, defining the main game loop, initializing the grid with random states, and creating functions to count alive neighbors and update the grid. The next step is to integrate the `update_grid` function into the main game loop to continuously update the state of the grid based on the game rules. This will allow the game to evolve over time, showing different generations of the cell population.
-
-Action:
-"""
-edit 6 6
-    grid = update_grid(grid)
-"""
-==== END RAW ====
-
-
-ACTION:
-FileWriteAction(path='game_of_life.py', content='grid = update_grid(grid)', start=6, end=6, thoughts="Thoughts:\nSo far, I have set up the basic structure for Conway's Game of Life using pygame, including initializing pygame, setting up the display window, defining the main game loop, initializing the grid with random states, and creating functions to count alive neighbors and update the grid. The next step is to integrate the `update_grid` function into the main game loop to continuously update the state of the grid based on the game rules. This will allow the game to evolve over time, showing different generations of the cell population.\n\nAction:", action=<ActionType.WRITE: 'write'>)
-
-OBSERVATION:
-WRITE OPERATION:
-You have written to "game_of_life.py" on these lines: 6:6.
-
-
-==============
-STEP 8
-
-PLAN:
-Please make conway's game of life in python using the pygame library. Be sure to use the tools in your workspace to help you out.
-Usage(completion_tokens=166, prompt_tokens=3393, total_tokens=3559)
-
-==== RAW OUTPUT ====
-Thoughts:
-So far, I have set up the basic structure for Conway's Game of Life using pygame, including initializing pygame, setting up the display window, defining the main game loop, initializing the grid with random states, and creating functions to count alive neighbors and update the grid. The next step is to integrate the `update_grid` function into the main game loop to continuously update the state of the grid based on the game rules. This will allow the game to evolve over time, showing different generations of the cell population. I need to ensure that the `update_grid` function is called within the main game loop and that the display is updated accordingly. I will start by checking the current content of the main game loop to see where to integrate the `update_grid` function.
-
-Action:
-"""
-goto 6
-"""
-==== END RAW ====
-
-
-ACTION:
-FileReadAction(path='game_of_life.py', start=6, end=106, thoughts="Thoughts:\nSo far, I have set up the basic structure for Conway's Game of Life using pygame, including initializing pygame, setting up the display window, defining the main game loop, initializing the grid with random states, and creating functions to count alive neighbors and update the grid. The next step is to integrate the `update_grid` function into the main game loop to continuously update the state of the grid based on the game rules. This will allow the game to evolve over time, showing different generations of the cell population. I need to ensure that the `update_grid` function is called within the main game loop and that the display is updated accordingly. I will start by checking the current content of the main game loop to see where to integrate the `update_grid` function.\n\nAction:", action=<ActionType.READ: 'read'>)
-
-OBSERVATION:
-grid = update_grid(grid)
-running = True
-# Define grid size and initialize grid
-grid_size = (50, 50)  # Grid size as a tuple (rows, columns)
-def count_alive_neighbors(grid, x, y):
-    directions = [(-1, -1), (-1, 0), (-1, 1), (0, -1), (0, 1), (1, -1), (1, 0), (1, 1)]
-def update_grid(grid):
-    new_grid = [[0 for x in range(grid_size[1])] for y in range(grid_size[0])]
-    for y in range(grid_size[0]):
-        for x in range(grid_size[1]):
-            alive_neighbors = count_alive_neighbors(grid, x, y)
-            if grid[y][x] == 1 and (alive_neighbors == 2 or alive_neighbors == 3):
-                new_grid[y][x] = 1
-            elif grid[y][x] == 0 and alive_neighbors == 3:
-                new_grid[y][x] = 1
-    return new_grid
-    count = 0
-    for dx, dy in directions:
-        nx, ny = x + dx, y + dy
-        if 0 <= nx < grid_size[1] and 0 <= ny < grid_size[0]:
-            count += grid[ny][nx]
-    return count
-grid = [[0 for x in range(grid_size[1])] for y in range(grid_size[0])]  # Initialize grid with zeros
-import random
-for y in range(grid_size[0]):
-    for x in range(grid_size[1]):
-        grid[y][x] = random.randint(0, 1)  # Randomly assign cells as alive (1) or dead (0)
-while running:
-    for event in pygame.event.get():
-        if event.type == pygame.QUIT:
-            running = False
-pygame.display.set_caption("Conway's Game of Life")
-pygame.init()
-
-
-
-==============
-STEP 9
-
-PLAN:
-Please make conway's game of life in python using the pygame library. Be sure to use the tools in your workspace to help you out.
-Usage(completion_tokens=174, prompt_tokens=3728, total_tokens=3902)
-
-==== RAW OUTPUT ====
-Thoughts:
-From the previous actions, it seems that the basic structure for Conway's Game of Life using pygame has been set up, including initializing pygame, setting up the display window, defining the main game loop, initializing the grid with random states, and creating functions to count alive neighbors and update the grid. The next step is to ensure that the game loop properly updates the grid and refreshes the display to show the new state of the grid in each iteration. This involves calling `pygame.display.update()` to update the entire screen and using `pygame.time.Clock()` to control the frame rate. I will start by adding these necessary updates to the game loop.
-
-Action:
-"""
-edit 95 95
-    pygame.display.update()
-    clock = pygame.time.Clock()
-    clock.tick(10)  # Control the frame rate (10 frames per second)
-"""
-
-==== END RAW ====
-
-
-ACTION:
-FileWriteAction(path='game_of_life.py', content='pygame.display.update()\n    clock = pygame.time.Clock()\n    clock.tick(10)  # Control the frame rate (10 frames per second)', start=95, end=95, thoughts="Thoughts:\nFrom the previous actions, it seems that the basic structure for Conway's Game of Life using pygame has been set up, including initializing pygame, setting up the display window, defining the main game loop, initializing the grid with random states, and creating functions to count alive neighbors and update the grid. The next step is to ensure that the game loop properly updates the grid and refreshes the display to show the new state of the grid in each iteration. This involves calling `pygame.display.update()` to update the entire screen and using `pygame.time.Clock()` to control the frame rate. I will start by adding these necessary updates to the game loop.\n\nAction:", action=<ActionType.WRITE: 'write'>)
-
-OBSERVATION:
-WRITE OPERATION:
-You have written to "game_of_life.py" on these lines: 95:95.
-16:43:55 - opendevin:INFO: agent_controller.py:125 - Exited before finishing the task.
-
-
-```
-
-## Result:
-
-```
-import pygame
-
-# Setup the display window
-screen_size = (800, 600)
-screen = pygame.display.set_mode(screen_size)
-# Main game loop
-grid = update_grid(grid)
-running = True
-# Define grid size and initialize grid
-grid_size = (50, 50)  # Grid size as a tuple (rows, columns)
-def count_alive_neighbors(grid, x, y):
-    directions = [(-1, -1), (-1, 0), (-1, 1), (0, -1), (0, 1), (1, -1), (1, 0), (1, 1)]
-def update_grid(grid):
-    new_grid = [[0 for x in range(grid_size[1])] for y in range(grid_size[0])]
-    for y in range(grid_size[0]):
-        for x in range(grid_size[1]):
-            alive_neighbors = count_alive_neighbors(grid, x, y)
-            if grid[y][x] == 1 and (alive_neighbors == 2 or alive_neighbors == 3):
-                new_grid[y][x] = 1
-            elif grid[y][x] == 0 and alive_neighbors == 3:
-                new_grid[y][x] = 1
-    return new_grid
-    count = 0
-    for dx, dy in directions:
-        nx, ny = x + dx, y + dy
-        if 0 <= nx < grid_size[1] and 0 <= ny < grid_size[0]:
-            count += grid[ny][nx]
-    return count
-grid = [[0 for x in range(grid_size[1])] for y in range(grid_size[0])]  # Initialize grid with zeros
-import random
-for y in range(grid_size[0]):
-    for x in range(grid_size[1]):
-        grid[y][x] = random.randint(0, 1)  # Randomly assign cells as alive (1) or dead (0)
-while running:
-    for event in pygame.event.get():
-        if event.type == pygame.QUIT:
-            running = False
-pygame.display.set_caption("Conway's Game of Life")
-pygame.init()
-# This is a placeholder for the game of life implementation using pygame.pygame.display.update()
-    clock = pygame.time.Clock()
-    clock.tick(10)  # Control the frame rate (10 frames per second)
-
-```

agenthub/SWE_agent/__init__.py

@@ -1,5 +0,0 @@
-from opendevin.controller.agent import Agent
-
-from .agent import SWEAgent
-
-Agent.register('SWEAgent', SWEAgent)

agenthub/SWE_agent/agent.py

@@ -1,115 +0,0 @@
-from opendevin.controller.agent import Agent
-from opendevin.controller.state.state import State
-from opendevin.events.action import (
-    Action,
-    FileReadAction,
-    FileWriteAction,
-    MessageAction,
-)
-from opendevin.events.serialization.event import event_to_memory
-from opendevin.llm.llm import LLM
-
-from .parser import parse_command
-from .prompts import (
-    CONTEXT_PROMPT,
-    MEMORY_FORMAT,
-    NO_ACTION,
-    STEP_PROMPT,
-    SYSTEM_MESSAGE,
-)
-
-
-class SWEAgent(Agent):
-    VERSION = '1.0'
-    DEPRECATED = True
-    """
-    An attempt to recreate swe_agent with output parsing, prompting style, and Application Computer Interface (ACI).
-
-    SWE-agent includes ACI functions like 'goto', 'search_for', 'edit', 'scroll', 'run'
-    """
-
-    def __init__(self, llm: LLM):
-        super().__init__(llm)
-        self.memory_window = 4
-        self.max_retries = 2
-        self.cur_file: str = ''
-        self.cur_line: int = 0
-
-    def _think_act(self, messages: list[dict]) -> tuple[Action, str]:
-        resp = self.llm.do_completion(
-            messages=messages,
-            temperature=0.05,
-        )
-        action_resp = resp['choices'][0]['message']['content']
-        print(f"\033[1m\033[91m{resp['usage']}\033[0m")
-        print(
-            '\n==== RAW OUTPUT ====',
-            f'\033[96m{action_resp}\033[0m',
-            '==== END RAW ====\n',
-            sep='\n',
-        )
-        return parse_command(action_resp, self.cur_file, self.cur_line)
-
-    def _update(self, action: Action) -> None:
-        if isinstance(action, (FileReadAction, FileWriteAction)):
-            self.cur_file = action.path
-            self.cur_line = action.start
-
-    def step(self, state: State) -> Action:
-        """
-        SWE-Agent step:
-            1. Get context - past actions, custom commands, current step
-            2. Perform think-act - prompt model for action and reasoning
-            3. Catch errors - ensure model takes action (5 attempts max)
-        """
-        # retrieve short term memories from state.history, up to memory_window
-        memory_window = min(self.memory_window, len(state.history))
-        running_memory: list[str] = []
-        for prev_action, obs in state.history[-memory_window:]:
-            running_memory.append(
-                MEMORY_FORMAT(event_to_memory(prev_action), event_to_memory(obs))
-            )
-
-        goal = state.get_current_user_intent()
-
-        # always in the prompt if they exist: file and line
-        prompt = STEP_PROMPT(goal, self.cur_file, self.cur_line)
-
-        # prepare messages
-        msgs = [
-            {'content': SYSTEM_MESSAGE, 'role': 'system'},
-            {'content': prompt, 'role': 'user'},
-        ]
-
-        # insert memories
-        if len(running_memory) > 0:
-            context = CONTEXT_PROMPT(running_memory, self.memory_window)
-            msgs.insert(1, {'content': context, 'role': 'user'})
-        # clrs = [''] * (len(msgs)-2) + ['\033[0;36m', '\033[0;35m']
-        # print('\n\n'.join([c+m['content']+'\033[0m' for c, m in zip(clrs, msgs)]))
-
-        # send it over
-        action, thought = self._think_act(messages=msgs)
-
-        # be robust with malformed responses
-        start_msg_len = len(msgs)
-        while not action and len(msgs) < self.max_retries + start_msg_len:
-            error = NO_ACTION(thought)
-            error_msg = {'content': error, 'role': 'user'}
-            msgs.append(error_msg)
-            action, thought = self._think_act(messages=msgs)
-
-        if not action:
-            action = MessageAction(thought)
-
-        self._update(action)
-        self.latest_action = action
-        return action
-
-    def search_memory(self, query: str) -> list[str]:
-        # return [item for item in self.running_memory if query in item]
-        raise NotImplementedError('Search_memory not implemented currently')
-
-    def reset(self) -> None:
-        """Used to reset the agent"""
-        super().reset()

agenthub/SWE_agent/parser.py

@@ -1,167 +0,0 @@
-import re
-
-from opendevin.events.action import (
-    Action,
-    AgentFinishAction,
-    BrowseURLAction,
-    CmdRunAction,
-    FileReadAction,
-    FileWriteAction,
-    MessageAction,
-)
-
-from .prompts import COMMAND_USAGE, CUSTOM_DOCS
-
-# commands: exit, read, write, browse, kill, search_file, search_dir
-
-no_open_file_error = MessageAction(
-    'You are not currently in a file. You can use the read command to open a file and then use goto to navigate through it.'
-)
-
-
-def invalid_error(cmd, docs):
-    return f"""ERROR:
-Invalid command structure for
-```
-{cmd}
-```
-You may have caused this error by having multiple commands within your command block.
-If so, try again by running only one of the commands:
-
-Try again using this format:
-{COMMAND_USAGE[docs]}
-"""
-
-
-def get_action_from_string(
-    command_string: str, path: str, line: int, thoughts: str = ''
-) -> Action | None:
-    """
-    Parses the command string to find which command the agent wants to run
-    Converts the command into a proper Action and returns
-    """
-    vars = command_string.split(' ')
-    cmd = vars[0]
-    args = [] if len(vars) == 1 else vars[1:]
-
-    if 'exit' == cmd:
-        return AgentFinishAction()
-
-    elif 'think' == cmd:
-        return MessageAction(' '.join(args))
-
-    elif 'scroll_up' == cmd:
-        if not path:
-            return no_open_file_error
-        return FileReadAction(path, line + 100, line + 200, thoughts)
-
-    elif 'scroll_down' == cmd:
-        if not path:
-            return no_open_file_error
-        return FileReadAction(path, line - 100, line, thoughts)
-
-    elif 'goto' == cmd:
-        if not path:
-            return no_open_file_error
-        rex = r'^goto\s+(\d+)$'
-        valid = re.match(rex, command_string)
-        if valid:
-            start = int(valid.group(1))
-            end = start + 100
-            return FileReadAction(path, start, end, thoughts)
-        else:
-            return MessageAction(invalid_error(command_string, 'goto'))
-
-    elif 'edit' == cmd:
-        if not path:
-            return no_open_file_error
-        rex = r'^edit\s+(\d+)\s+(-?\d+)\s+(\S.*)$'
-        valid = re.match(rex, command_string, re.DOTALL)
-        if valid:
-            start = int(valid.group(1))
-            end = int(valid.group(2))
-            change = valid.group(3)
-            if '"' == change[-1] and '"' == change[0]:
-                change = change[1:-1]
-            return FileWriteAction(path, change, start, end, thoughts)
-        else:
-            return MessageAction(invalid_error(command_string, 'edit'))
-
-    elif 'read' == cmd:
-        rex = r'^read\s+(\S+)(?:\s+(\d+))?(?:\s+(-?\d+))?$'
-        valid = re.match(rex, command_string, re.DOTALL)
-        if valid:
-            file = valid.group(1)
-            start_str = valid.group(2)
-            end_str = valid.group(3)
-
-            start = 0 if not start_str else int(start_str)
-            end = -1 if not end_str else int(end_str)
-
-            return FileReadAction(file, start, end, thoughts)
-        else:
-            return MessageAction(invalid_error(command_string, 'read'))
-
-    elif 'write' == cmd:
-        rex = r'^write\s+(\S+)\s+(.*?)\s*(\d+)?\s*(-?\d+)?$'
-        valid = re.match(rex, command_string, re.DOTALL)
-
-        if valid:
-            file = valid.group(1)
-            content = valid.group(2)
-            start_str = valid.group(3)
-            end_str = valid.group(4)
-
-            start = 0 if not start_str else int(start_str)
-            end = -1 if not end_str else int(end_str)
-
-            if '"' == content[-1] and '"' == content[0]:
-                content = content[1:-1]
-
-            return FileWriteAction(file, content, start, end, thoughts)
-        else:
-            return MessageAction(invalid_error(command_string, 'write'))
-
-    elif 'browse' == cmd:
-        return BrowseURLAction(args[0].strip())
-
-    elif cmd in ['search_file', 'search_dir', 'find_file']:
-        rex = r'^(search_file|search_dir|find_file)\s+(\S+)(?:\s+(\S+))?$'
-        valid = re.match(rex, command_string, re.DOTALL)
-        if valid:
-            return CmdRunAction(command_string)
-        else:
-            return MessageAction(
-                f'Invalid command structure for\n ```\n{command_string}\n```.\nTry again using this format:\n{CUSTOM_DOCS}'
-            )
-    else:
-        # check bash command
-        obs = str(CmdRunAction(f'type {cmd}'))
-        if obs.split(':')[-1].strip() == 'not found':
-            # echo not found error for llm
-            return MessageAction(content=obs)
-        else:
-            # run valid command
-            return CmdRunAction(command_string)
-
-
-def parse_command(input_str: str, path: str, line: int):
-    """
-    Parses a given string and separates the command (enclosed in triple backticks) from any accompanying text.
-
-    Args:
-        input_str (str): The input string to be parsed.
-
-    Returns:
-        tuple: A tuple containing the command and the accompanying text (if any).
-    """
-    input_str = input_str.strip()
-    if '```' in input_str:
-        parts = input_str.split('```')
-        command_str = parts[1].strip()
-        ind = 2 if len(parts) > 2 else 1
-        accompanying_text = ''.join(parts[:-ind]).strip()
-        action = get_action_from_string(command_str, path, line, accompanying_text)
-        if action:
-            return action, accompanying_text
-    return None, input_str  # used for retry

agenthub/SWE_agent/prompts.py

@@ -1,208 +0,0 @@
-DEFAULT_COMMANDS_DICT = {
-    'exit': 'Executed when task is complete',
-    'read <file_name> [<start_line>] [<end_line>]': "Shows a given file's contents starting from <start_line> up to <end_line>. Default: start_line = 0, end_line = -1. By default the whole file will be read.",
-    'write <file> <changes> [<start_line>] [<end_line>]': 'Modifies a <file> by replacing the current lines between <start_line> and <end_line> with <changes>. Default start_line = 0 and end_line = -1. Calling this with no line args will replace the whole file.',
-    'browse <url>': 'Returns the text version of any url, this can be useful to look up documentation or finding issues on github',
-    'scroll_up': 'Takes no arguments. This will scroll up and show you the 100 lines above your current lines',
-    'scroll_down': 'Takes no arguments. This will scroll down and show you the 100 lines below your current lines',
-    'edit <start_line> <end_line> <changes>': 'This will modify lines in the currently open file. use start_line and end_line to designate which lines to change and then write the multiline changes. Set end_line to -1 to denote the end of the file',
-    'goto <line_num>': 'This will take you directly to a line and show you the 100 lines below it.',
-    '<bash_command> <args>': 'You can use any bash command you need (cd, ls, rm, grep, dir, mv, wget, git, zip, etc.) with their arguments included',
-    'pip install <package>': 'You can use this to import python packages. Make sure you include the correct package name when using this command.',
-    'ls': 'Use the ls command to view all the files in your current directory, this is a good starting point.',
-    'NOT ALLOWED': 'You cannot use interactive commands like python or node',
-}
-
-COMMAND_USAGE = {
-    'exit': 'Usage:\n```\nexit\n```\nExecuted when task is complete',
-    'read': "Args:\n<file_name> [<start_line>] [<end_line>]\nUsage:\n```\nread file.py\n```\nor\n```\nread example.py <start_line> <end_line>\n```\nShows a given file's contents starting from <start_line> up to <end_line>. Default: start_line = 0, end_line = -1. by default the whole file will be read.",
-    'write': 'Args:\n<file> <changes> [<start_line>] [<end_line>]\nUsage:\n```\nwrite "def main():\n    print("This is line one")" 0 2\n```\nModifies a <file> by replacing the current lines between <start_line> and <end_line> with <changes>. Default start_line = 0 and end_line = -1. Calling this with no line args will replace the whole file.',
-    'edit': 'Args:\n<start_line> <end_line> <changes>\nUsage:\n```\nedit 0 1 import pandas as pd\n```\nThis will modify the current file you are in with the changes you make between the line numbers you designate',
-    'goto': 'Args:\n<line_num>\nUsage:\n```\ngoto <line_num>\n```\nThis will show you the 100 lines below and including the line you specify within your current file.',
-    'scroll_up': 'Usage:\n```\nscroll_up\n```\nThis will return the 100 lines above where you are currently at',
-    'scroll_down': 'Usage:\n```\nscroll_down\n```\nThis will return the 100 line below where you are currently at',
-    'browse': 'Args:\n<url>\nUsage:\n```\nbrowse https://github.com/OpenDevin/OpenDevin\n```\nThis will fetch the Text elements from the given url and show them to you.',
-}
-
-DEFAULT_COMMANDS = '\n'.join([k + ' - ' + v for k, v in DEFAULT_COMMANDS_DICT.items()])
-
-# from opendevin.parse_commands import parse_command_file
-# USE parse_command_file(filepath) to get the custom commands
-CUSTOM_DOCS = None
-
-CUSTOM_COMMANDS = f"""Custom bash commands:
-{CUSTOM_DOCS}
-"""
-
-DOCUMENTATION = f"""DOCUMENTATION:
-It is recommend that you use the commands provided for interacting with files and your directory because they have been specially built for you.
-They will make it much easier for you to look at files and make changes. Using these commands will help you be better at your task.
-You can open an file by using either the read or write operations.
-- If a file already exists you should read it before making any changes. Use the `edit` command to make changes once you have read it.
-- If you are creating a new file use the write command. Use the `edit` command to make changes once you have created the new file.
-
-Commands:
-{DEFAULT_COMMANDS}
-{CUSTOM_COMMANDS}
-
-The following commands require an open file to be used: edit, scroll_up, scroll_down, goto
-To modify the current file use 'edit'. To move through the current file use 'goto' or 'scroll_up'/'scroll_down'
-when using write and edit do not surround the code with any "" just write the code.
-"""
-
-GENERAL_GUIDELINES = """INSTRUCTIONS:
-Now, you're going to solve this issue on your own. You can use any bash commands or custom commands you wish to complete your task. Edit all the files you need to and run any checks or tests that you want.
-Remember, YOU CAN ONLY ENTER ONE COMMAND AT A TIME. You should always wait for feedback after every command.
-When you're satisfied with all of the changes you've made, you can indicate that you are done by running the exit command.
-Note however that you cannot use any interactive session commands (e.g. python, vim, node) in this environment, but you can write scripts and run them. E.g. you can write a python script and then run it with `python <script_name>.py`.
-
-NOTE ABOUT THE write COMMAND: Indentation really matters! When editing a file, make sure to insert appropriate indentation before each line!
-
-IMPORTANT TIPS:
-1. Reproduce the bug: Always start by trying to replicate the bug that the issue discusses. If the issue includes code for reproducing the bug, we recommend that you re-implement that in your environment and run it to ensure you can reproduce the bug. Then, start trying to fix it. When you think you've fixed the bug, re-run the bug reproduction script to make sure that the issue has indeed been resolved.
-   If the bug reproduction script does not print anything when it successfully runs, we recommend adding a print("Script completed successfully, no errors.") command at the end of the file, so that you can be sure the script ran fine all the way through.
-2. Try different commands: If you run a command and it doesn't work, try running a different command. A command that did not work once will not work the second time unless you modify it.
-3. Navigate large files: If you open a file and need to get to an area around a specific line that is not in the first 100 lines, say line 583, you would use the 'read' command like this: 'read <file> 583'. This is a much faster way to read through the file.
-4. Handle input files: If the bug reproduction script requires inputting/reading a specific file, such as 'buggy-input.png', and you'd like to understand how to input that file, conduct a search in the existing repository code to see whether someone else has already done that. Do this by running the command: 'search_dir "buggy-input.png"'. If that doesn't work, use the Linux 'find' command.
-5. Understand your context: Always make sure to look at the currently open file and the current working directory. The currently open file might be in a different directory than the working directory.
-6. Verify your edits: When editing files, it is easy to accidentally specify a wrong line number or to write code with incorrect indentation. Always check the code after you issue an edit to make sure that it reflects what you wanted to accomplish. If it didn't, issue another command to fix it.
-7. Thoroughly test your solution: After making any changes to fix a bug, be sure to thoroughly test your solution to ensure the bug has been resolved. Re-run the bug reproduction script and verify that the issue has been addressed.
-"""
-
-RESPONSE_FORMAT = """RESPONSE FORMAT:
-This is the format of the response you will make in order to solve the current issue.
-You will be given multiple iterations to complete this task so break it into steps and solve them one by one.
-
-Your output must contain the following:
-- First, thoughts about what your next action should be and plan it out.
-    - You will have a memory of your thoughts so you can use this to remember things for the next step.
-    - Use your thoughts to think about what you are currently doing, what you have done on prior steps and how that relates to solving the problem.
-- Second, create a piece of code that will execute your next action based on the thoughts you have.
-    - Remember that you can only have one action for each thought, do not include multiple actions.
-
-Your code MUST be surrounded in triple back ticks EXACTLY like this:
-```
-<code>
-```
-
-Notes:
-- Adhere to the format so that the program loop continues smoothly, it is very important to only give one command per output.
-- DO NOT give more than one command within the triple backticks. This will just throw an error and nothing will happen as a result.
-- Do not give multiple code blocks, if you do only the second one will be captured and run, this might give an error if the first one was necessary.
-- To execute multiple commands you should write them down in your thoughts section so you can remember it on the next step and execute them then.
-- The only commands you are not capable of executing are interactive commands like `python` or `node` by themselves.
-- If you think that you have completed the task that has been given to you based on your previous actions and outputs then use ``` exit ``` as the command to let the system know that you are done.
-- DO NOT make any copies of your previous memories, those will be provided to you at each step, making copies just wastes time and energy. Think smarter not harder.
-- The write and edit commands requires proper indentation in the content section ex. `write hw.py def hello():\n    print(\'Hello World\')` this is how you would have to format your write command.
-    - The white spaces matter as the code changes will be added to the code so they must have proper syntax.
-
-This is a template using the format described above
-Items in <> are suggestions for you, fill them out based on the context of the problem you are solving.
-
-[ FORMAT ]
-Thoughts:
-<Provide clear and concise thoughts on the next step to take, highlighting any important details or context that should be remembered.>
-<You can use multiple lines to express your thoughts>
-
-Action:
-```
-<command> <params>
-```
-[ END FORMAT ]
-
-Do not provide anything extra just your thought and action.
-"""
-
-SYSTEM_MESSAGE = f"""SYSTEM INFO:
-You are an autonomous coding agent, here to provide solutions for coding issues.
-You have been designed to assist with a wide range of programming tasks, from code editing and debugging to testing and deployment.
-You have access to a variety of tools and commands that you can use to help you solve problems efficiently.
-
-{GENERAL_GUIDELINES}
-
-{DOCUMENTATION}
-""".strip()
-
-
-def NO_ACTION(latest):
-    return f"""
-You did not include any action to take in your most recent output:
-
-===== Output ======
-{latest}
-==== End Output ===
-
-Remember these are the custom commands you can use:
-{DOCUMENTATION}
-
-Lets try that again, it is very important that you adhere to the output format
-This time, be sure to use the exact format below, replacing anything in <> with the appropriate value(s):
-{RESPONSE_FORMAT}
-
-It is crucial you use the format provided as the output will be parsed automatically.
-"""
-
-
-def file_info(file: str, line: int):
-    if file:
-        return f"""CURRENT WORKSPACE:
-    Open File: {file} on line {line}
-    You can use these commands with the current file:
-    Navigation: `scroll_up`, `scroll_down`, and `goto <line>`
-    Modification: `edit <start_line> <end_line> <changes>`
-    """
-
-
-def STEP_PROMPT(task, file, line_num):
-    return f"""
-{RESPONSE_FORMAT}
-You are currently trying to complete this task:
-{task}
-
-{file_info(file, line_num)}
-
-Keep all of the guidelines above in mind when you are thinking and making code.
-Please come up with a thought and action based on your current task and latest steps.
-Make sure that you do not repeat the same actions, there will not be any changes in result if you do not changes anything.
-Be very strict about the formatting that you use and make sure you follow the guidelines.
-NEVER output multiple commands. ONLY take ONE STEP at a time.
-When you have completed your task run the "exit" command.
-Begin with your thought about the next step and then come up with an action to perform your thought.
-""".strip()
-
-
-def unpack_dict(data: dict, restrict: list[str] | None = None):
-    lines = []
-    restrict = [] if restrict is None else restrict
-    for key, value in data.items():
-        if key in restrict:
-            continue
-        elif isinstance(value, dict):
-            nested_str = unpack_dict(value, restrict).replace('\n', '\n  ')
-            val = f'{key}:' + '\n  ' + f'{nested_str}'
-            lines.append(val)
-        else:
-            lines.append(f'{key}: {value}')
-    return '\n'.join(lines)
-
-
-def MEMORY_FORMAT(act, obs):
-    return f"""
-Previous Action:
-{unpack_dict(act, ["content"])}
-
-Output from Action:
-{unpack_dict(obs)}
-""".strip()
-
-
-def CONTEXT_PROMPT(memory, window):
-    res = f'These are your past {window} actions:\n'
-    window_size = window if len(memory) > window else len(memory)
-    cur_mems = memory[-window_size:]
-    res += '===== Previous Actions =====\n'
-    for idx, mem in enumerate(cur_mems):
-        res += f'\nMemory {idx}:\n{mem}\n'
-    res += '======= End Actions =======\n'
-    res += 'Use these memories to provide additional context to the problem you are solving.\nRemember that you have already completed these steps so you do not need to perform them again.'
-    return res

agenthub/__init__.py

@@ -1,30 +1,25 @@
 from dotenv import load_dotenv
 
-from opendevin.controller.agent import Agent
-
-from .micro.agent import MicroAgent
-from .micro.registry import all_microagents
+from agenthub.micro.agent import MicroAgent
+from agenthub.micro.registry import all_microagents
+from openhands.controller.agent import Agent
 
 load_dotenv()
 
 
-from . import (  # noqa: E402
-    SWE_agent,
+from agenthub import (  # noqa: E402
     browsing_agent,
     codeact_agent,
     codeact_swe_agent,
     delegator_agent,
     dummy_agent,
-    monologue_agent,
     planner_agent,
 )
 
 __all__ = [
-    'monologue_agent',
     'codeact_agent',
     'codeact_swe_agent',
     'planner_agent',
-    'SWE_agent',
     'delegator_agent',
     'dummy_agent',
     'browsing_agent',

agenthub/browsing_agent/README.md

@@ -8,7 +8,7 @@ This folder implements the basic BrowserGym [demo agent](https://github.com/Serv
 Note that for browsing tasks, GPT-4 is usually a requirement to get reasonable results, due to the complexity of the web page structures.
 
 ```
-poetry run python ./opendevin/core/main.py \
+poetry run python ./openhands/core/main.py \
            -i 10 \
            -t "tell me the usa's president using google search" \
            -c BrowsingAgent \

agenthub/browsing_agent/__init__.py

@@ -1,5 +1,4 @@
-from opendevin.controller.agent import Agent
-
-from .browsing_agent import BrowsingAgent
+from agenthub.browsing_agent.browsing_agent import BrowsingAgent
+from openhands.controller.agent import Agent
 
 Agent.register('BrowsingAgent', BrowsingAgent)

agenthub/browsing_agent/browsing_agent.py

@@ -1,38 +1,94 @@
-import ast
+import os
 
 from browsergym.core.action.highlevel import HighLevelActionSet
 from browsergym.utils.obs import flatten_axtree_to_str
 
-from opendevin.controller.agent import Agent
-from opendevin.controller.state.state import State
-from opendevin.core.logger import opendevin_logger as logger
-from opendevin.events.action import (
+from agenthub.browsing_agent.response_parser import BrowsingResponseParser
+from openhands.controller.agent import Agent
+from openhands.controller.state.state import State
+from openhands.core.config import AgentConfig
+from openhands.core.logger import openhands_logger as logger
+from openhands.core.message import Message, TextContent
+from openhands.events.action import (
     Action,
     AgentFinishAction,
     BrowseInteractiveAction,
     MessageAction,
 )
-from opendevin.events.observation import BrowserOutputObservation
-from opendevin.llm.llm import LLM
-from opendevin.runtime.plugins import (
+from openhands.events.event import EventSource
+from openhands.events.observation import BrowserOutputObservation
+from openhands.events.observation.observation import Observation
+from openhands.llm.llm import LLM
+from openhands.runtime.plugins import (
     PluginRequirement,
 )
 
+USE_NAV = (
+    os.environ.get('USE_NAV', 'true') == 'true'
+)  # only disable NAV actions when running webarena and miniwob benchmarks
+USE_CONCISE_ANSWER = (
+    os.environ.get('USE_CONCISE_ANSWER', 'false') == 'true'
+)  # only return concise answer when running webarena and miniwob benchmarks
 
-def parse_response(response: str) -> Action:
-    if '```' not in response:
-        # unexpected response format, message back to user
-        return MessageAction(response)
-    thought = response.split('```')[0].strip()
-    action_str = response.split('```')[1].strip()
-    # handle send message to user function call in BrowserGym
-    for sub_action in action_str.split('\n'):
-        if 'send_msg_to_user(' in sub_action:
-            tree = ast.parse(sub_action)
-            args = tree.body[0].value.args  # type: ignore
-            return MessageAction(args[0].value)
+if not USE_NAV and USE_CONCISE_ANSWER:
+    EVAL_MODE = True  # disabled NAV actions and only return concise answer, for webarena and miniwob benchmarks\
+else:
+    EVAL_MODE = False
 
-    return BrowseInteractiveAction(browser_actions=action_str, thought=thought)
+
+def get_error_prefix(last_browser_action: str) -> str:
+    return f'IMPORTANT! Last action is incorrect:\n{last_browser_action}\nThink again with the current observation of the page.\n'
+
+
+def get_system_message(goal: str, action_space: str) -> str:
+    return f"""\
+# Instructions
+Review the current state of the page and all other information to find the best
+possible next action to accomplish your goal. Your answer will be interpreted
+and executed by a program, make sure to follow the formatting instructions.
+
+# Goal:
+{goal}
+
+# Action Space
+{action_space}
+"""
+
+
+CONCISE_INSTRUCTION = """\
+
+Here is another example with chain of thought of a valid action when providing a concise answer to user:
+"
+In order to accomplish my goal I need to send the information asked back to the user. This page list the information of HP Inkjet Fax Machine, which is the product identified in the objective. Its price is $279.49. I will send a message back to user with the answer.
+```send_msg_to_user("$279.49")```
+"
+"""
+
+
+def get_prompt(
+    error_prefix: str, cur_url: str, cur_axtree_txt: str, prev_action_str: str
+) -> str:
+    prompt = f"""\
+{error_prefix}
+
+# Current Page URL:
+{cur_url}
+
+# Current Accessibility Tree:
+{cur_axtree_txt}
+
+# Previous Actions
+{prev_action_str}
+
+Here is an example with chain of thought of a valid action when clicking on a button:
+"
+In order to accomplish my goal I need to click on the button with bid 12
+```click("12")```
+"
+""".strip()
+    if USE_CONCISE_ANSWER:
+        prompt += CONCISE_INSTRUCTION
+    return prompt
 
 
 class BrowsingAgent(Agent):
@@ -42,25 +98,26 @@ class BrowsingAgent(Agent):
     """
 
     sandbox_plugins: list[PluginRequirement] = []
+    response_parser = BrowsingResponseParser()
 
     def __init__(
         self,
         llm: LLM,
+        config: AgentConfig,
     ) -> None:
-        """
-        Initializes a new instance of the BrowsingAgent class.
+        """Initializes a new instance of the BrowsingAgent class.
 
         Parameters:
         - llm (LLM): The llm to be used by this agent
         """
-        super().__init__(llm)
+        super().__init__(llm, config)
+        # define a configurable action space, with chat functionality, web navigation, and webpage grounding using accessibility tree and HTML.
+        # see https://github.com/ServiceNow/BrowserGym/blob/main/core/src/browsergym/core/action/highlevel.py for more details
+        action_subsets = ['chat', 'bid']
+        if USE_NAV:
+            action_subsets.append('nav')
         self.action_space = HighLevelActionSet(
-            # see https://github.com/ServiceNow/BrowserGym/blob/main/core/src/browsergym/core/action/highlevel.py for more details
-            subsets=[
-                'chat',
-                'bid',
-                'nav',
-            ],  # define a configurable action space, with chat functionality, web navigation, and webpage grounding using accessibility tree and HTML.
+            subsets=action_subsets,
             strict=False,  # less strict on the parsing of the actions
             multiaction=True,  # enable to agent to take multiple actions at once
         )
@@ -68,15 +125,13 @@ class BrowsingAgent(Agent):
         self.reset()
 
     def reset(self) -> None:
-        """
-        Resets the Browsing Agent.
-        """
+        """Resets the Browsing Agent."""
         super().reset()
         self.cost_accumulator = 0
+        self.error_accumulator = 0
 
     def step(self, state: State) -> Action:
-        """
-        Performs one step using the Browsing Agent.
+        """Performs one step using the Browsing Agent.
         This includes gathering information on previous steps and prompting the model to make a browsing command to execute.
 
         Parameters:
@@ -87,81 +142,83 @@ class BrowsingAgent(Agent):
         - MessageAction(content) - Message action to run (e.g. ask for clarification)
         - AgentFinishAction() - end the interaction
         """
-        goal = state.get_current_user_intent()
-        messages = []
-        prev_actions = ''
+        messages: list[Message] = []
+        prev_actions = []
+        cur_url = ''
         cur_axtree_txt = ''
         error_prefix = ''
         last_obs = None
-        for prev_action, obs in state.history:
-            if isinstance(prev_action, BrowseInteractiveAction):
-                prev_actions += f'{prev_action.browser_actions}\n'
-                last_obs = obs
-            elif (
-                isinstance(prev_action, MessageAction) and prev_action.source != 'user'
-            ):
-                # agent has responded, task finish.
-                return AgentFinishAction()
+        last_action = None
+
+        if EVAL_MODE and len(state.history.get_events_as_list()) == 1:
+            # for webarena and miniwob++ eval, we need to retrieve the initial observation already in browser env
+            # initialize and retrieve the first observation by issuing an noop OP
+            # For non-benchmark browsing, the browser env starts with a blank page, and the agent is expected to first navigate to desired websites
+            return BrowseInteractiveAction(browser_actions='noop()')
+
+        for event in state.history.get_events():
+            if isinstance(event, BrowseInteractiveAction):
+                prev_actions.append(event.browser_actions)
+                last_action = event
+            elif isinstance(event, MessageAction) and event.source == EventSource.AGENT:
+                # agent has responded, task finished.
+                return AgentFinishAction(outputs={'content': event.content})
+            elif isinstance(event, Observation):
+                last_obs = event
+
+        if EVAL_MODE:
+            prev_actions = prev_actions[1:]  # remove the first noop action
+
+        prev_action_str = '\n'.join(prev_actions)
+        # if the final BrowserInteractiveAction exec BrowserGym's send_msg_to_user,
+        # we should also send a message back to the user in OpenHands and call it a day
+        if (
+            isinstance(last_action, BrowseInteractiveAction)
+            and last_action.browsergym_send_msg_to_user
+        ):
+            return MessageAction(last_action.browsergym_send_msg_to_user)
 
         if isinstance(last_obs, BrowserOutputObservation):
             if last_obs.error:
                 # add error recovery prompt prefix
-                error_prefix = f'IMPORTANT! Last action is incorrect:\n{last_obs.last_browser_action}\nThink again with the current observation of the page.\n'
-            cur_axtree_txt = flatten_axtree_to_str(last_obs.axtree_object)
+                error_prefix = get_error_prefix(last_obs.last_browser_action)
+                self.error_accumulator += 1
+                if self.error_accumulator > 5:
+                    return MessageAction('Too many errors encountered. Task failed.')
 
-        system_msg = f"""\
-# Instructions
-Review the current state of the page and all other information to find the best
-possible next action to accomplish your goal. Your answer will be interpreted
-and executed by a program, make sure to follow the formatting instructions.
+            cur_url = last_obs.url
 
-# Goal:
-{goal}
+            try:
+                cur_axtree_txt = flatten_axtree_to_str(
+                    last_obs.axtree_object,
+                    extra_properties=last_obs.extra_element_properties,
+                    with_clickable=True,
+                    filter_visible_only=True,
+                )
+            except Exception as e:
+                logger.error(
+                    'Error when trying to process the accessibility tree: %s', e
+                )
+                return MessageAction('Error encountered when browsing.')
 
-# Action Space
-{self.action_space.describe(with_long_description=False, with_examples=True)}
-"""
+        goal, _ = state.get_current_user_intent()
 
-        messages.append({'role': 'system', 'content': system_msg})
+        if goal is None:
+            goal = state.inputs['task']
 
-        prompt = f"""\
-{error_prefix}
+        system_msg = get_system_message(
+            goal,
+            self.action_space.describe(with_long_description=False, with_examples=True),
+        )
 
-# Current Accessibility Tree:
-{cur_axtree_txt}
+        messages.append(Message(role='system', content=[TextContent(text=system_msg)]))
 
-# Previous Actions
-{prev_actions}
+        prompt = get_prompt(error_prefix, cur_url, cur_axtree_txt, prev_action_str)
+        messages.append(Message(role='user', content=[TextContent(text=prompt)]))
 
-Here is an example with chain of thought of a valid action when clicking on a button:
-"
-In order to accomplish my goal I need to click on the button with bid 12
-```click("12")```
-"
-""".strip()
-        messages.append({'role': 'user', 'content': prompt})
         response = self.llm.completion(
-            messages=messages,
+            messages=self.llm.format_messages_for_llm(messages),
             temperature=0.0,
+            stop=[')```', ')\n```'],
         )
-        self.log_cost(response)
-        action_resp = response['choices'][0]['message']['content']
-        logger.info(prompt)
-        logger.info(action_resp)
-        return parse_response(action_resp)
-
-    def search_memory(self, query: str) -> list[str]:
-        raise NotImplementedError('Implement this abstract method')
-
-    def log_cost(self, response):
-        # TODO: refactor to unified cost tracking
-        try:
-            cur_cost = self.llm.completion_cost(response)
-        except Exception:
-            cur_cost = 0
-        self.cost_accumulator += cur_cost
-        logger.info(
-            'Cost: %.2f USD | Accumulated Cost: %.2f USD',
-            cur_cost,
-            self.cost_accumulator,
-        )
+        return self.response_parser.parse(response)

agenthub/browsing_agent/prompt.py

@@ -12,12 +12,11 @@ from browsergym.core.action.base import AbstractActionSet
 from browsergym.core.action.highlevel import HighLevelActionSet
 from browsergym.core.action.python import PythonActionSet
 
-from opendevin.runtime.browser.browser_env import BrowserEnv
-
-from .utils import (
+from agenthub.browsing_agent.utils import (
     ParseError,
     parse_html_tags_raise,
 )
+from openhands.runtime.browser.browser_env import BrowserEnv
 
 
 @dataclass
@@ -58,7 +57,7 @@ class Flags:
 
     @classmethod
     def from_dict(self, flags_dict):
-        """Helper for JSON serializble requirement."""
+        """Helper for JSON serializable requirement."""
         if isinstance(flags_dict, Flags):
             return flags_dict
 
@@ -75,7 +74,8 @@ class PromptElement:
     Prompt elements are used to build the prompt. Use flags to control which
     prompt elements are visible. We use class attributes as a convenient way
     to implement static prompts, but feel free to override them with instance
-    attributes or @property decorator."""
+    attributes or @property decorator.
+    """
 
     _prompt = ''
     _abstract_ex = ''
@@ -146,7 +146,7 @@ class Shrinkable(PromptElement, abc.ABC):
         """Implement shrinking of this prompt element.
 
         You need to recursively call all shrinkable elements that are part of
-        this prompt. You can also implement a shriking startegy for this prompt.
+        this prompt. You can also implement a shrinking strategy for this prompt.
         Shrinking is can be called multiple times to progressively shrink the
         prompt until it fits max_tokens. Default max shrink iterations is 20.
         """
@@ -161,7 +161,7 @@ class Truncater(Shrinkable):
 
     def __init__(self, visible, shrink_speed=0.3, start_truncate_iteration=10):
         super().__init__(visible=visible)
-        self.shrink_speed = shrink_speed  # the percentage shrinked in each iteration
+        self.shrink_speed = shrink_speed  # the percentage shrunk in each iteration
         self.start_truncate_iteration = (
             start_truncate_iteration  # the iteration to start truncating
         )
@@ -200,11 +200,10 @@ def fit_tokens(
     model_name : str, optional
         The name of the model used when tokenizing.
 
-    Returns
+    Returns:
     -------
     str : the prompt after shrinking.
     """
-
     if max_prompt_chars is None:
         return shrinkable.prompt
 
@@ -355,7 +354,7 @@ and executed by a program, make sure to follow the formatting instructions.
         self._prompt += '\n'.join(
             [
                 f"""\
- - [{msg['role']}] {msg['message']}"""
+ - [{msg['role']}], {msg['message']}"""
                 for msg in chat_messages
             ]
         )
@@ -494,11 +493,13 @@ def _get_action_space(flags: Flags) -> AbstractActionSet:
             action_space = PythonActionSet(strict=flags.is_strict)
             if flags.multi_actions:
                 warn(
-                    f'Flag action_space={repr(flags.action_space)} incompatible with multi_actions={repr(flags.multi_actions)}.'
+                    f'Flag action_space={repr(flags.action_space)} incompatible with multi_actions={repr(flags.multi_actions)}.',
+                    stacklevel=2,
                 )
             if flags.demo_mode != 'off':
                 warn(
-                    f'Flag action_space={repr(flags.action_space)} incompatible with demo_mode={repr(flags.demo_mode)}.'
+                    f'Flag action_space={repr(flags.action_space)} incompatible with demo_mode={repr(flags.demo_mode)}.',
+                    stacklevel=2,
                 )
             return action_space
         case 'bid':
@@ -577,8 +578,8 @@ the form is not visible yet or some fields are disabled. I need to replan.
 def diff(previous, new):
     """Return a string showing the difference between original and new.
 
-    If the difference is above diff_threshold, return the diff string."""
-
+    If the difference is above diff_threshold, return the diff string.
+    """
     if previous == new:
         return 'Identical', []
 

agenthub/browsing_agent/response_parser.py

@@ -0,0 +1,88 @@
+import ast
+
+from openhands.controller.action_parser import ActionParser, ResponseParser
+from openhands.core.logger import openhands_logger as logger
+from openhands.events.action import (
+    Action,
+    BrowseInteractiveAction,
+)
+
+
+class BrowsingResponseParser(ResponseParser):
+    def __init__(self):
+        # Need to pay attention to the item order in self.action_parsers
+        super().__init__()
+        self.action_parsers = [BrowsingActionParserMessage()]
+        self.default_parser = BrowsingActionParserBrowseInteractive()
+
+    def parse(self, response: str) -> Action:
+        action_str = self.parse_response(response)
+        return self.parse_action(action_str)
+
+    def parse_response(self, response) -> str:
+        action_str = response['choices'][0]['message']['content']
+        if action_str is None:
+            return ''
+        action_str = action_str.strip()
+        if action_str and not action_str.endswith('```'):
+            action_str = action_str + ')```'
+        logger.debug(action_str)
+        return action_str
+
+    def parse_action(self, action_str: str) -> Action:
+        for action_parser in self.action_parsers:
+            if action_parser.check_condition(action_str):
+                return action_parser.parse(action_str)
+        return self.default_parser.parse(action_str)
+
+
+class BrowsingActionParserMessage(ActionParser):
+    """Parser action:
+    - BrowseInteractiveAction(browser_actions) - unexpected response format, message back to user
+    """
+
+    def __init__(
+        self,
+    ):
+        pass
+
+    def check_condition(self, action_str: str) -> bool:
+        return '```' not in action_str
+
+    def parse(self, action_str: str) -> Action:
+        msg = f'send_msg_to_user("""{action_str}""")'
+        return BrowseInteractiveAction(
+            browser_actions=msg,
+            thought=action_str,
+            browsergym_send_msg_to_user=action_str,
+        )
+
+
+class BrowsingActionParserBrowseInteractive(ActionParser):
+    """Parser action:
+    - BrowseInteractiveAction(browser_actions) - handle send message to user function call in BrowserGym
+    """
+
+    def __init__(
+        self,
+    ):
+        pass
+
+    def check_condition(self, action_str: str) -> bool:
+        return True
+
+    def parse(self, action_str: str) -> Action:
+        thought = action_str.split('```')[0].strip()
+        action_str = action_str.split('```')[1].strip()
+        msg_content = ''
+        for sub_action in action_str.split('\n'):
+            if 'send_msg_to_user(' in sub_action:
+                tree = ast.parse(sub_action)
+                args = tree.body[0].value.args  # type: ignore
+                msg_content = args[0].value
+
+        return BrowseInteractiveAction(
+            browser_actions=action_str,
+            thought=thought,
+            browsergym_send_msg_to_user=msg_content,
+        )

agenthub/browsing_agent/utils.py

@@ -7,7 +7,6 @@ import yaml
 
 def yaml_parser(message):
     """Parse a yaml message for the retry function."""
-
     # saves gpt-3.5 from some yaml parsing errors
     message = re.sub(r':\s*\n(?=\S|\n)', ': ', message)
 
@@ -16,7 +15,7 @@ def yaml_parser(message):
         valid = True
         retry_message = ''
     except yaml.YAMLError as e:
-        warn(str(e))
+        warn(str(e), stacklevel=2)
         value = {}
         valid = False
         retry_message = "Your response is not a valid yaml. Please try again and be careful to the format. Don't add any apology or comment, just the answer."
@@ -47,7 +46,6 @@ def _compress_chunks(text, identifier, skip_list, split_regex='\n\n+'):
 
 def compress_string(text):
     """Compress a string by replacing redundant paragraphs and lines with identifiers."""
-
     # Perform paragraph-level compression
     def_dict, compressed_text = _compress_chunks(
         text, identifier='§', skip_list=[], split_regex='\n\n+'
@@ -79,12 +77,12 @@ def extract_html_tags(text, keys):
     keys : list of str
         The HTML tags to extract the content from.
 
-    Returns
+    Returns:
     -------
     dict
         A dictionary mapping each key to a list of subset in `text` that match the key.
 
-    Notes
+    Notes:
     -----
     All text and keys will be converted to lowercase before matching.
 
@@ -126,7 +124,7 @@ def parse_html_tags(text, keys=(), optional_keys=(), merge_multiple=False):
     optional_keys : list of str
         The HTML tags to extract the content from, but are optional.
 
-    Returns
+    Returns:
     -------
     dict
         A dictionary mapping each key to subset of `text` that match the key.

agenthub/codeact_agent/README.md

@@ -9,21 +9,4 @@ The conceptual idea is illustrated below. At each turn, the agent can:
    - Execute any valid Linux `bash` command
    - Execute any valid `Python` code with [an interactive Python interpreter](https://ipython.org/). This is simulated through `bash` command, see plugin system below for more details.
 
-![image](https://github.com/OpenDevin/OpenDevin/assets/38853559/92b622e3-72ad-4a61-8f41-8c040b6d5fb3)
-
-## Plugin System
-
-To make the CodeAct agent more powerful with only access to `bash` action space, CodeAct agent leverages OpenDevin's plugin system:
-- [Jupyter plugin](https://github.com/OpenDevin/OpenDevin/tree/main/opendevin/runtime/plugins/jupyter): for IPython execution via bash command
-- [SWE-agent tool plugin](https://github.com/OpenDevin/OpenDevin/tree/main/opendevin/runtime/plugins/swe_agent_commands): Powerful bash command line tools for software development tasks introduced by [swe-agent](https://github.com/princeton-nlp/swe-agent).
-
-## Demo
-
-https://github.com/OpenDevin/OpenDevin/assets/38853559/f592a192-e86c-4f48-ad31-d69282d5f6ac
-
-*Example of CodeActAgent with `gpt-4-turbo-2024-04-09` performing a data science task (linear regression)*
-
-## Work-in-progress & Next step
-
-[] Support web-browsing
-[] Complete the workflow for CodeAct agent to submit Github PRs
+![image](https://github.com/All-Hands-AI/OpenHands/assets/38853559/92b622e3-72ad-4a61-8f41-8c040b6d5fb3)

agenthub/codeact_agent/__init__.py

@@ -1,5 +1,4 @@
-from opendevin.controller.agent import Agent
-
-from .codeact_agent import CodeActAgent
+from agenthub.codeact_agent.codeact_agent import CodeActAgent
+from openhands.controller.agent import Agent
 
 Agent.register('CodeActAgent', CodeActAgent)

agenthub/codeact_agent/action_parser.py

@@ -0,0 +1,177 @@
+import re
+
+from openhands.controller.action_parser import ActionParser, ResponseParser
+from openhands.events.action import (
+    Action,
+    AgentDelegateAction,
+    AgentFinishAction,
+    CmdRunAction,
+    IPythonRunCellAction,
+    MessageAction,
+)
+
+
+class CodeActResponseParser(ResponseParser):
+    """Parser action:
+    - CmdRunAction(command) - bash command to run
+    - IPythonRunCellAction(code) - IPython code to run
+    - AgentDelegateAction(agent, inputs) - delegate action for (sub)task
+    - MessageAction(content) - Message action to run (e.g. ask for clarification)
+    - AgentFinishAction() - end the interaction
+    """
+
+    def __init__(self):
+        # Need pay attention to the item order in self.action_parsers
+        super().__init__()
+        self.action_parsers = [
+            CodeActActionParserFinish(),
+            CodeActActionParserCmdRun(),
+            CodeActActionParserIPythonRunCell(),
+            CodeActActionParserAgentDelegate(),
+        ]
+        self.default_parser = CodeActActionParserMessage()
+
+    def parse(self, response) -> Action:
+        action_str = self.parse_response(response)
+        return self.parse_action(action_str)
+
+    def parse_response(self, response) -> str:
+        action = response.choices[0].message.content
+        if action is None:
+            return ''
+        for lang in ['bash', 'ipython', 'browse']:
+            if f'<execute_{lang}>' in action and f'</execute_{lang}>' not in action:
+                action += f'</execute_{lang}>'
+        return action
+
+    def parse_action(self, action_str: str) -> Action:
+        for action_parser in self.action_parsers:
+            if action_parser.check_condition(action_str):
+                return action_parser.parse(action_str)
+        return self.default_parser.parse(action_str)
+
+
+class CodeActActionParserFinish(ActionParser):
+    """Parser action:
+    - AgentFinishAction() - end the interaction
+    """
+
+    def __init__(
+        self,
+    ):
+        self.finish_command = None
+
+    def check_condition(self, action_str: str) -> bool:
+        self.finish_command = re.search(r'<finish>.*</finish>', action_str, re.DOTALL)
+        return self.finish_command is not None
+
+    def parse(self, action_str: str) -> Action:
+        assert (
+            self.finish_command is not None
+        ), 'self.finish_command should not be None when parse is called'
+        thought = action_str.replace(self.finish_command.group(0), '').strip()
+        return AgentFinishAction(thought=thought)
+
+
+class CodeActActionParserCmdRun(ActionParser):
+    """Parser action:
+    - CmdRunAction(command) - bash command to run
+    - AgentFinishAction() - end the interaction
+    """
+
+    def __init__(
+        self,
+    ):
+        self.bash_command = None
+
+    def check_condition(self, action_str: str) -> bool:
+        self.bash_command = re.search(
+            r'<execute_bash>(.*?)</execute_bash>', action_str, re.DOTALL
+        )
+        return self.bash_command is not None
+
+    def parse(self, action_str: str) -> Action:
+        assert (
+            self.bash_command is not None
+        ), 'self.bash_command should not be None when parse is called'
+        thought = action_str.replace(self.bash_command.group(0), '').strip()
+        # a command was found
+        command_group = self.bash_command.group(1).strip()
+        if command_group.strip() == 'exit':
+            return AgentFinishAction(thought=thought)
+        return CmdRunAction(command=command_group, thought=thought)
+
+
+class CodeActActionParserIPythonRunCell(ActionParser):
+    """Parser action:
+    - IPythonRunCellAction(code) - IPython code to run
+    """
+
+    def __init__(
+        self,
+    ):
+        self.python_code = None
+        self.jupyter_kernel_init_code: str = 'from agentskills import *'
+
+    def check_condition(self, action_str: str) -> bool:
+        self.python_code = re.search(
+            r'<execute_ipython>(.*?)</execute_ipython>', action_str, re.DOTALL
+        )
+        return self.python_code is not None
+
+    def parse(self, action_str: str) -> Action:
+        assert (
+            self.python_code is not None
+        ), 'self.python_code should not be None when parse is called'
+        code_group = self.python_code.group(1).strip()
+        thought = action_str.replace(self.python_code.group(0), '').strip()
+        return IPythonRunCellAction(
+            code=code_group,
+            thought=thought,
+            kernel_init_code=self.jupyter_kernel_init_code,
+        )
+
+
+class CodeActActionParserAgentDelegate(ActionParser):
+    """Parser action:
+    - AgentDelegateAction(agent, inputs) - delegate action for (sub)task
+    """
+
+    def __init__(
+        self,
+    ):
+        self.agent_delegate = None
+
+    def check_condition(self, action_str: str) -> bool:
+        self.agent_delegate = re.search(
+            r'<execute_browse>(.*)</execute_browse>', action_str, re.DOTALL
+        )
+        return self.agent_delegate is not None
+
+    def parse(self, action_str: str) -> Action:
+        assert (
+            self.agent_delegate is not None
+        ), 'self.agent_delegate should not be None when parse is called'
+        thought = action_str.replace(self.agent_delegate.group(0), '').strip()
+        browse_actions = self.agent_delegate.group(1).strip()
+        task = f'{thought}. I should start with: {browse_actions}'
+        return AgentDelegateAction(agent='BrowsingAgent', inputs={'task': task})
+
+
+class CodeActActionParserMessage(ActionParser):
+    """Parser action:
+    - MessageAction(content) - Message action to run (e.g. ask for clarification)
+    """
+
+    def __init__(
+        self,
+    ):
+        pass
+
+    def check_condition(self, action_str: str) -> bool:
+        # We assume the LLM is GOOD enough that when it returns pure natural language
+        # it wants to talk to the user
+        return True
+
+    def parse(self, action_str: str) -> Action:
+        return MessageAction(content=action_str, wait_for_response=True)

agenthub/codeact_agent/codeact_agent.py

@@ -1,131 +1,47 @@
-import re
+import os
+from itertools import islice
 
-from agenthub.codeact_agent.prompt import (
-    COMMAND_DOCS,
-    EXAMPLES,
-    GITHUB_MESSAGE,
-    SYSTEM_PREFIX,
-    SYSTEM_SUFFIX,
-)
-from opendevin.controller.agent import Agent
-from opendevin.controller.state.state import State
-from opendevin.events.action import (
+from agenthub.codeact_agent.action_parser import CodeActResponseParser
+from openhands.controller.agent import Agent
+from openhands.controller.state.state import State
+from openhands.core.config import AgentConfig
+from openhands.core.message import ImageContent, Message, TextContent
+from openhands.events.action import (
     Action,
+    AgentDelegateAction,
     AgentFinishAction,
-    BrowseInteractiveAction,
     CmdRunAction,
     IPythonRunCellAction,
     MessageAction,
 )
-from opendevin.events.observation import (
-    BrowserOutputObservation,
+from openhands.events.observation import (
+    AgentDelegateObservation,
     CmdOutputObservation,
     IPythonRunCellObservation,
+    UserRejectObservation,
 )
-from opendevin.llm.llm import LLM
-from opendevin.runtime.plugins import (
+from openhands.events.observation.error import ErrorObservation
+from openhands.events.observation.observation import Observation
+from openhands.events.serialization.event import truncate_content
+from openhands.llm.llm import LLM
+from openhands.runtime.plugins import (
     AgentSkillsRequirement,
     JupyterRequirement,
     PluginRequirement,
 )
-
-ENABLE_GITHUB = True
-
-
-def parse_response(response) -> str:
-    action = response.choices[0].message.content
-    for lang in ['bash', 'ipython', 'browse']:
-        if f'<execute_{lang}>' in action and f'</execute_{lang}>' not in action:
-            action += f'</execute_{lang}>'
-    return action
-
-
-def action_to_str(action: Action) -> str:
-    if isinstance(action, CmdRunAction):
-        return f'{action.thought}\n<execute_bash>\n{action.command}\n</execute_bash>'
-    elif isinstance(action, IPythonRunCellAction):
-        return f'{action.thought}\n<execute_ipython>\n{action.code}\n</execute_ipython>'
-    elif isinstance(action, BrowseInteractiveAction):
-        return f'{action.thought}\n<execute_browse>\n{action.browser_actions}\n</execute_browse>'
-    elif isinstance(action, MessageAction):
-        return action.content
-    return ''
-
-
-def get_action_message(action: Action) -> dict[str, str] | None:
-    if (
-        isinstance(action, BrowseInteractiveAction)
-        or isinstance(action, CmdRunAction)
-        or isinstance(action, IPythonRunCellAction)
-        or isinstance(action, MessageAction)
-    ):
-        return {
-            'role': 'user' if action.source == 'user' else 'assistant',
-            'content': action_to_str(action),
-        }
-    return None
-
-
-def get_observation_message(obs) -> dict[str, str] | None:
-    if isinstance(obs, CmdOutputObservation):
-        content = 'OBSERVATION:\n' + truncate_observation(obs.content)
-        content += (
-            f'\n[Command {obs.command_id} finished with exit code {obs.exit_code}]]'
-        )
-        return {'role': 'user', 'content': content}
-    elif isinstance(obs, IPythonRunCellObservation):
-        content = 'OBSERVATION:\n' + obs.content
-        # replace base64 images with a placeholder
-        splitted = content.split('\n')
-        for i, line in enumerate(splitted):
-            if '![image](data:image/png;base64,' in line:
-                splitted[i] = (
-                    '![image](data:image/png;base64, ...) already displayed to user'
-                )
-        content = '\n'.join(splitted)
-        content = truncate_observation(content)
-        return {'role': 'user', 'content': content}
-    elif isinstance(obs, BrowserOutputObservation):
-        content = 'OBSERVATION:\n' + truncate_observation(obs.content)
-        return {'role': 'user', 'content': content}
-    return None
-
-
-def truncate_observation(observation: str, max_chars: int = 10_000) -> str:
-    """
-    Truncate the middle of the observation if it is too long.
-    """
-    if len(observation) <= max_chars:
-        return observation
-    half = max_chars // 2
-    return (
-        observation[:half]
-        + '\n[... Observation truncated due to length ...]\n'
-        + observation[-half:]
-    )
-
-
-# FIXME: We can tweak these two settings to create MicroAgents specialized toward different area
-def get_system_message() -> str:
-    if ENABLE_GITHUB:
-        return f'{SYSTEM_PREFIX}\n{GITHUB_MESSAGE}\n\n{COMMAND_DOCS}\n\n{SYSTEM_SUFFIX}'
-    else:
-        return f'{SYSTEM_PREFIX}\n\n{COMMAND_DOCS}\n\n{SYSTEM_SUFFIX}'
-
-
-def get_in_context_example() -> str:
-    return EXAMPLES
+from openhands.utils.microagent import MicroAgent
+from openhands.utils.prompt import PromptManager
 
 
 class CodeActAgent(Agent):
-    VERSION = '1.5'
+    VERSION = '1.9'
     """
     The Code Act Agent is a minimalist agent.
     The agent works by passing the model a list of action-observation pairs and prompting the model to take the next step.
 
     ### Overview
 
-    This agent implements the CodeAct idea ([paper](https://arxiv.org/abs/2402.13463), [tweet](https://twitter.com/xingyaow_/status/1754556835703751087)) that consolidates LLM agents’ **act**ions into a unified **code** action space for both *simplicity* and *performance* (see paper for more details).
+    This agent implements the CodeAct idea ([paper](https://arxiv.org/abs/2402.01030), [tweet](https://twitter.com/xingyaow_/status/1754556835703751087)) that consolidates LLM agents’ **act**ions into a unified **code** action space for both *simplicity* and *performance* (see paper for more details).
 
     The conceptual idea is illustrated below. At each turn, the agent can:
 
@@ -134,148 +50,236 @@ class CodeActAgent(Agent):
     - Execute any valid Linux `bash` command
     - Execute any valid `Python` code with [an interactive Python interpreter](https://ipython.org/). This is simulated through `bash` command, see plugin system below for more details.
 
-    ![image](https://github.com/OpenDevin/OpenDevin/assets/38853559/92b622e3-72ad-4a61-8f41-8c040b6d5fb3)
-
-    ### Plugin System
-
-    To make the CodeAct agent more powerful with only access to `bash` action space, CodeAct agent leverages OpenDevin's plugin system:
-    - [Jupyter plugin](https://github.com/OpenDevin/OpenDevin/tree/main/opendevin/runtime/plugins/jupyter): for IPython execution via bash command
-    - [SWE-agent tool plugin](https://github.com/OpenDevin/OpenDevin/tree/main/opendevin/runtime/plugins/swe_agent_commands): Powerful bash command line tools for software development tasks introduced by [swe-agent](https://github.com/princeton-nlp/swe-agent).
-
-    ### Demo
-
-    https://github.com/OpenDevin/OpenDevin/assets/38853559/f592a192-e86c-4f48-ad31-d69282d5f6ac
-
-    *Example of CodeActAgent with `gpt-4-turbo-2024-04-09` performing a data science task (linear regression)*
-
-    ### Work-in-progress & Next step
-
-    [] Support web-browsing
-    [] Complete the workflow for CodeAct agent to submit Github PRs
+    ![image](https://github.com/All-Hands-AI/OpenHands/assets/38853559/92b622e3-72ad-4a61-8f41-8c040b6d5fb3)
 
     """
 
     sandbox_plugins: list[PluginRequirement] = [
         # NOTE: AgentSkillsRequirement need to go before JupyterRequirement, since
-        # AgentSkillsRequirement provides a lot of Python functions
-        # and it need to be initialized before Jupyter for Jupyter to use those functions.
+        # AgentSkillsRequirement provides a lot of Python functions,
+        # and it needs to be initialized before Jupyter for Jupyter to use those functions.
         AgentSkillsRequirement(),
         JupyterRequirement(),
     ]
-    jupyter_kernel_init_code: str = 'from agentskills import *'
 
-    system_message: str = get_system_message()
-    in_context_example: str = f"Here is an example of how you can interact with the environment for task solving:\n{get_in_context_example()}\n\nNOW, LET'S START!"
+    action_parser = CodeActResponseParser()
 
     def __init__(
         self,
         llm: LLM,
+        config: AgentConfig,
     ) -> None:
-        """
-        Initializes a new instance of the CodeActAgent class.
+        """Initializes a new instance of the CodeActAgent class.
 
         Parameters:
         - llm (LLM): The llm to be used by this agent
         """
-        super().__init__(llm)
+        super().__init__(llm, config)
         self.reset()
 
+        self.micro_agent = (
+            MicroAgent(
+                os.path.join(
+                    os.path.dirname(__file__), 'micro', f'{config.micro_agent_name}.md'
+                )
+            )
+            if config.micro_agent_name
+            else None
+        )
+
+        self.prompt_manager = PromptManager(
+            prompt_dir=os.path.join(os.path.dirname(__file__)),
+            agent_skills_docs=AgentSkillsRequirement.documentation,
+            micro_agent=self.micro_agent,
+        )
+
+    def action_to_str(self, action: Action) -> str:
+        if isinstance(action, CmdRunAction):
+            return (
+                f'{action.thought}\n<execute_bash>\n{action.command}\n</execute_bash>'
+            )
+        elif isinstance(action, IPythonRunCellAction):
+            return f'{action.thought}\n<execute_ipython>\n{action.code}\n</execute_ipython>'
+        elif isinstance(action, AgentDelegateAction):
+            return f'{action.thought}\n<execute_browse>\n{action.inputs["task"]}\n</execute_browse>'
+        elif isinstance(action, MessageAction):
+            return action.content
+        elif isinstance(action, AgentFinishAction) and action.source == 'agent':
+            return action.thought
+        return ''
+
+    def get_action_message(self, action: Action) -> Message | None:
+        if (
+            isinstance(action, AgentDelegateAction)
+            or isinstance(action, CmdRunAction)
+            or isinstance(action, IPythonRunCellAction)
+            or isinstance(action, MessageAction)
+            or (isinstance(action, AgentFinishAction) and action.source == 'agent')
+        ):
+            content = [TextContent(text=self.action_to_str(action))]
+
+            if (
+                self.llm.vision_is_active()
+                and isinstance(action, MessageAction)
+                and action.images_urls
+            ):
+                content.append(ImageContent(image_urls=action.images_urls))
+
+            return Message(
+                role='user' if action.source == 'user' else 'assistant', content=content
+            )
+        return None
+
+    def get_observation_message(self, obs: Observation) -> Message | None:
+        max_message_chars = self.llm.config.max_message_chars
+        obs_prefix = 'OBSERVATION:\n'
+        if isinstance(obs, CmdOutputObservation):
+            text = obs_prefix + truncate_content(obs.content, max_message_chars)
+            text += (
+                f'\n[Command {obs.command_id} finished with exit code {obs.exit_code}]'
+            )
+            return Message(role='user', content=[TextContent(text=text)])
+        elif isinstance(obs, IPythonRunCellObservation):
+            text = obs_prefix + obs.content
+            # replace base64 images with a placeholder
+            splitted = text.split('\n')
+            for i, line in enumerate(splitted):
+                if '![image](data:image/png;base64,' in line:
+                    splitted[i] = (
+                        '![image](data:image/png;base64, ...) already displayed to user'
+                    )
+            text = '\n'.join(splitted)
+            text = truncate_content(text, max_message_chars)
+            return Message(role='user', content=[TextContent(text=text)])
+        elif isinstance(obs, AgentDelegateObservation):
+            text = obs_prefix + truncate_content(
+                obs.outputs['content'] if 'content' in obs.outputs else '',
+                max_message_chars,
+            )
+            return Message(role='user', content=[TextContent(text=text)])
+        elif isinstance(obs, ErrorObservation):
+            text = obs_prefix + truncate_content(obs.content, max_message_chars)
+            text += '\n[Error occurred in processing last action]'
+            return Message(role='user', content=[TextContent(text=text)])
+        elif isinstance(obs, UserRejectObservation):
+            text = 'OBSERVATION:\n' + truncate_content(obs.content, max_message_chars)
+            text += '\n[Last action has been rejected by the user]'
+            return Message(role='user', content=[TextContent(text=text)])
+        else:
+            # If an observation message is not returned, it will cause an error
+            # when the LLM tries to return the next message
+            raise ValueError(f'Unknown observation type: {type(obs)}')
+
     def reset(self) -> None:
-        """
-        Resets the CodeAct Agent.
-        """
+        """Resets the CodeAct Agent."""
         super().reset()
 
     def step(self, state: State) -> Action:
-        """
-        Performs one step using the CodeAct Agent.
+        """Performs one step using the CodeAct Agent.
         This includes gathering info on previous steps and prompting the model to make a command to execute.
 
         Parameters:
-        - state (State): used to get updated info and background commands
+        - state (State): used to get updated info
 
         Returns:
         - CmdRunAction(command) - bash command to run
         - IPythonRunCellAction(code) - IPython code to run
-        - BrowseInteractiveAction(browsergym_command) - BrowserGym commands to run
+        - AgentDelegateAction(agent, inputs) - delegate action for (sub)task
         - MessageAction(content) - Message action to run (e.g. ask for clarification)
         - AgentFinishAction() - end the interaction
         """
-        messages: list[dict[str, str]] = [
-            {'role': 'system', 'content': self.system_message},
-            {'role': 'user', 'content': self.in_context_example},
-        ]
+        # if we're done, go back
+        latest_user_message = state.history.get_last_user_message()
+        if latest_user_message and latest_user_message.strip() == '/exit':
+            return AgentFinishAction()
 
-        for prev_action, obs in state.history:
-            action_message = get_action_message(prev_action)
-            if action_message:
-                messages.append(action_message)
-
-            obs_message = get_observation_message(obs)
-            if obs_message:
-                messages.append(obs_message)
-
-        latest_user_message = [m for m in messages if m['role'] == 'user'][-1]
-        if latest_user_message:
-            if latest_user_message['content'].strip() == '/exit':
-                return AgentFinishAction()
-            latest_user_message['content'] += (
-                f'\n\nENVIRONMENT REMINDER: You have {state.max_iterations - state.iteration} turns left to complete the task.'
-            )
-
-        response = self.llm.do_completion(
-            messages=messages,
-            stop=[
+        # prepare what we want to send to the LLM
+        messages = self._get_messages(state)
+        params = {
+            'messages': self.llm.format_messages_for_llm(messages),
+            'stop': [
                 '</execute_ipython>',
                 '</execute_bash>',
                 '</execute_browse>',
             ],
-            temperature=0.0,
+        }
+
+        if self.llm.is_caching_prompt_active():
+            params['extra_headers'] = {
+                'anthropic-beta': 'prompt-caching-2024-07-31',
+            }
+
+        response = self.llm.completion(**params)
+
+        return self.action_parser.parse(response)
+
+    def _get_messages(self, state: State) -> list[Message]:
+        messages: list[Message] = [
+            Message(
+                role='system',
+                content=[
+                    TextContent(
+                        text=self.prompt_manager.system_message,
+                        cache_prompt=self.llm.is_caching_prompt_active(),  # Cache system prompt
+                    )
+                ],
+            ),
+            Message(
+                role='user',
+                content=[
+                    TextContent(
+                        text=self.prompt_manager.initial_user_message,
+                        cache_prompt=self.llm.is_caching_prompt_active(),  # if the user asks the same query,
+                    )
+                ],
+            ),
+        ]
+
+        for event in state.history.get_events():
+            # create a regular message from an event
+            if isinstance(event, Action):
+                message = self.get_action_message(event)
+            elif isinstance(event, Observation):
+                message = self.get_observation_message(event)
+            else:
+                raise ValueError(f'Unknown event type: {type(event)}')
+
+            # add regular message
+            if message:
+                # handle error if the message is the SAME role as the previous message
+                # litellm.exceptions.BadRequestError: litellm.BadRequestError: OpenAIException - Error code: 400 - {'detail': 'Only supports u/a/u/a/u...'}
+                # there shouldn't be two consecutive messages from the same role
+                if messages and messages[-1].role == message.role:
+                    messages[-1].content.extend(message.content)
+                else:
+                    messages.append(message)
+
+        # Add caching to the last 2 user messages
+        if self.llm.is_caching_prompt_active():
+            user_turns_processed = 0
+            for message in reversed(messages):
+                if message.role == 'user' and user_turns_processed < 2:
+                    message.content[
+                        -1
+                    ].cache_prompt = True  # Last item inside the message content
+                    user_turns_processed += 1
+
+        # The latest user message is important:
+        # we want to remind the agent of the environment constraints
+        latest_user_message = next(
+            islice(
+                (
+                    m
+                    for m in reversed(messages)
+                    if m.role == 'user'
+                    and any(isinstance(c, TextContent) for c in m.content)
+                ),
+                1,
+            ),
+            None,
         )
+        if latest_user_message:
+            reminder_text = f'\n\nENVIRONMENT REMINDER: You have {state.max_iterations - state.iteration} turns left to complete the task. When finished reply with <finish></finish>.'
+            latest_user_message.content.append(TextContent(text=reminder_text))
 
-        action_str: str = parse_response(response)
-        state.num_of_chars += sum(
-            len(message['content']) for message in messages
-        ) + len(action_str)
-
-        if finish_command := re.search(r'<finish>.*</finish>', action_str, re.DOTALL):
-            thought = action_str.replace(finish_command.group(0), '').strip()
-            return AgentFinishAction(thought=thought)
-        if bash_command := re.search(
-            r'<execute_bash>(.*?)</execute_bash>', action_str, re.DOTALL
-        ):
-            # remove the command from the action string to get thought
-            thought = action_str.replace(bash_command.group(0), '').strip()
-            # a command was found
-            command_group = bash_command.group(1).strip()
-
-            if command_group.strip() == 'exit':
-                return AgentFinishAction()
-            return CmdRunAction(command=command_group, thought=thought)
-        elif python_code := re.search(
-            r'<execute_ipython>(.*?)</execute_ipython>', action_str, re.DOTALL
-        ):
-            # a code block was found
-            code_group = python_code.group(1).strip()
-            thought = action_str.replace(python_code.group(0), '').strip()
-            return IPythonRunCellAction(
-                code=code_group,
-                thought=thought,
-                kernel_init_code=self.jupyter_kernel_init_code,
-            )
-        elif browse_command := re.search(
-            r'<execute_browse>(.*)</execute_browse>', action_str, re.DOTALL
-        ):
-            # BrowserGym actions was found
-            browse_actions = browse_command.group(1).strip()
-            thought = action_str.replace(browse_command.group(0), '').strip()
-            return BrowseInteractiveAction(
-                browser_actions=browse_actions, thought=thought
-            )
-        else:
-            # We assume the LLM is GOOD enough that when it returns pure natural language
-            # it want to talk to the user
-            return MessageAction(content=action_str, wait_for_response=True)
-
-    def search_memory(self, query: str) -> list[str]:
-        raise NotImplementedError('Implement this abstract method')
+        return messages

agenthub/codeact_agent/micro/github.md

@@ -0,0 +1,69 @@
+---
+name: github
+agent: CodeActAgent
+require_env_var:
+    SANDBOX_ENV_GITHUB_TOKEN: "Create a GitHub Personal Access Token (https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/managing-your-personal-access-tokens) and set it as SANDBOX_GITHUB_TOKEN in your environment variables."
+---
+
+# How to Interact with Github
+
+## Environment Variable Available
+
+- `GITHUB_TOKEN`: A read-only token for Github.
+
+## Using GitHub's RESTful API
+
+Use `curl` with the `GITHUB_TOKEN` to interact with GitHub's API. Here are some common operations:
+
+Here's a template for API calls:
+
+```sh
+curl -H "Authorization: token $GITHUB_TOKEN" \
+    "https://api.github.com/{endpoint}"
+```
+
+First replace `{endpoint}` with the specific API path. Common operations:
+
+1. View an issue or pull request:
+   - Issues: `/repos/{owner}/{repo}/issues/{issue_number}`
+   - Pull requests: `/repos/{owner}/{repo}/pulls/{pull_request_number}`
+
+2. List repository issues or pull requests:
+   - Issues: `/repos/{owner}/{repo}/issues`
+   - Pull requests: `/repos/{owner}/{repo}/pulls`
+
+3. Search issues or pull requests:
+   - `/search/issues?q=repo:{owner}/{repo}+is:{type}+{search_term}+state:{state}`
+   - Replace `{type}` with `issue` or `pr`
+
+4. List repository branches:
+   `/repos/{owner}/{repo}/branches`
+
+5. Get commit details:
+   `/repos/{owner}/{repo}/commits/{commit_sha}`
+
+6. Get repository details:
+   `/repos/{owner}/{repo}`
+
+7. Get user information:
+   `/user`
+
+8. Search repositories:
+   `/search/repositories?q={query}`
+
+9. Get rate limit status:
+   `/rate_limit`
+
+Replace `{owner}`, `{repo}`, `{commit_sha}`, `{issue_number}`, `{pull_request_number}`,
+`{search_term}`, `{state}`, and `{query}` with appropriate values.
+
+## Important Notes
+
+1. Always use the GitHub API for operations instead of a web browser.
+2. The `GITHUB_TOKEN` is read-only. Avoid operations that require write access.
+3. Git config (username and email) is pre-set. Do not modify.
+4. Edit and test code locally. Never push directly to remote.
+5. Verify correct branch before committing.
+6. Commit changes frequently.
+7. If the issue or task is ambiguous or lacks sufficient detail, always request clarification from the user before proceeding.
+8. You should avoid using command line tools like `sed` for file editing.

agenthub/codeact_agent/system_prompt.j2

@@ -0,0 +1,52 @@
+{% set MINIMAL_SYSTEM_PREFIX %}
+A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed answers to the user's questions.
+The assistant can use a Python environment with <execute_ipython>, e.g.:
+<execute_ipython>
+print("Hello World!")
+</execute_ipython>
+The assistant can execute bash commands wrapped with <execute_bash>, e.g. <execute_bash> ls </execute_bash>.
+If a bash command returns exit code `-1`, this means the process is not yet finished.
+The assistant must then send a second <execute_bash>. The second <execute_bash> can be empty
+(which will retrieve any additional logs), or it can contain text to be sent to STDIN of the running process,
+or it can contain the text `ctrl+c` to interrupt the process.
+
+For commands that may run indefinitely, the output should be redirected to a file and the command run
+in the background, e.g. <execute_bash> python3 app.py > server.log 2>&1 & </execute_bash>
+If a command execution result says "Command timed out. Sending SIGINT to the process",
+the assistant should retry running the command in the background.
+{% endset %}
+{% set BROWSING_PREFIX %}
+The assistant can browse the Internet with <execute_browse> and </execute_browse>.
+For example, <execute_browse> Tell me the usa's president using google search </execute_browse>.
+Or <execute_browse> Tell me what is in http://example.com </execute_browse>.
+{% endset %}
+{% set PIP_INSTALL_PREFIX %}
+The assistant can install Python packages using the %pip magic command in an IPython environment by using the following syntax: <execute_ipython> %pip install [package needed] </execute_ipython> and should always import packages and define variables before starting to use them.
+{% endset %}
+{% set SYSTEM_PREFIX = MINIMAL_SYSTEM_PREFIX + BROWSING_PREFIX + PIP_INSTALL_PREFIX %}
+{% set COMMAND_DOCS %}
+Apart from the standard Python library, the assistant can also use the following functions (already imported) in <execute_ipython> environment:
+{{ agent_skills_docs }}
+IMPORTANT:
+- `open_file` only returns the first 100 lines of the file by default! The assistant MUST use `scroll_down` repeatedly to read the full file BEFORE making edits!
+- The assistant shall adhere to THE `edit_file_by_replace`, `append_file` and `insert_content_at_line` FUNCTIONS REQUIRING PROPER INDENTATION. If the assistant would like to add the line '        print(x)', it must fully write the line out, with all leading spaces before the code!
+- Indentation is important and code that is not indented correctly will fail and require fixing before it can be run.
+- Any code issued should be less than 50 lines to avoid context being cut off!
+- After EVERY `create_file` the method `append_file` shall be used to write the FIRST content!
+- For `edit_file_by_replace` NEVER provide empty parameters!
+- For `edit_file_by_replace` the file must be read fully before any replacements!
+{% endset %}
+{% set SYSTEM_SUFFIX %}
+Responses should be concise.
+The assistant should attempt fewer things at a time instead of putting too many commands OR too much code in one "execute" block.
+Include ONLY ONE <execute_ipython>, <execute_bash>, or <execute_browse> per response, unless the assistant is finished with the task or needs more input or action from the user in order to proceed.
+If the assistant is finished with the task you MUST include <finish></finish> in your response.
+IMPORTANT: Execute code using <execute_ipython>, <execute_bash>, or <execute_browse> whenever possible.
+The assistant should utilize full file paths and the `pwd` command to prevent path-related errors.
+The assistant must avoid apologies and thanks in its responses.
+
+{% endset %}
+{# Combine all parts without newlines between them #}
+{{ SYSTEM_PREFIX -}}
+{{- COMMAND_DOCS -}}
+{{- SYSTEM_SUFFIX }}

agenthub/codeact_agent/user_prompt.j2

@@ -1,50 +1,10 @@
-from opendevin.runtime.plugins import AgentSkillsRequirement
-
-_AGENT_SKILLS_DOCS = AgentSkillsRequirement.documentation
-
-COMMAND_DOCS = (
-    '\nApart from the standard Python library, the assistant can also use the following functions (already imported) in <execute_ipython> environment:\n'
-    f'{_AGENT_SKILLS_DOCS}'
-    "Please note that THE `edit_file` FUNCTION REQUIRES PROPER INDENTATION. If the assistant would like to add the line '        print(x)', it must fully write that out, with all those spaces before the code! Indentation is important and code that is not indented correctly will fail and require fixing before it can be run."
-)
-
-# ======= SYSTEM MESSAGE =======
-MINIMAL_SYSTEM_PREFIX = """A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions.
-The assistant can interact with an interactive Python (Jupyter Notebook) environment and receive the corresponding output when needed. The code should be enclosed using "<execute_ipython>" tag, for example:
-<execute_ipython>
-print("Hello World!")
-</execute_ipython>
-The assistant can execute bash commands on behalf of the user by wrapping them with <execute_bash> and </execute_bash>.
-For example, you can list the files in the current directory by <execute_bash> ls </execute_bash>.
-"""
-
-BROWSING_PREFIX = """The assistant can browse the Internet with commands on behalf of the user by wrapping them with <execute_browse> and </execute_browse>.
-For example, you can browse a given URL by <execute_browse> goto("<URL>") </execute_browse>.
-The assistant should attempt fewer things at a time instead of putting too much commands OR code in one "execute" block.
-"""
-PIP_INSTALL_PREFIX = """The assistant can install Python packages using the %pip magic command in an IPython environment by using the following syntax: <execute_ipython> %pip install [package needed] </execute_ipython> and should always import packages and define variables before starting to use them."""
-
-SYSTEM_PREFIX = MINIMAL_SYSTEM_PREFIX + BROWSING_PREFIX + PIP_INSTALL_PREFIX
-
-GITHUB_MESSAGE = """To do any activities on GitHub, the assistant should use the token in the $GITHUB_TOKEN environment variable.
-For instance, to push a local branch `my_branch` to the github repo `owner/repo`, the assistant can use the following four commands:
-<execute_bash> git push https://$GITHUB_TOKEN@github.com/owner/repo.git my_branch </execute_bash>
-If the assistant require access to GitHub but $GITHUB_TOKEN is not set, ask the user to set it."""
-
-SYSTEM_SUFFIX = """The assistant's response should be concise.
-The assistant should include ONLY ONE <execute_ipython> or <execute_bash> or <execute_browse> in every one of the responses, unless the assistant is finished with the task or need more input or action from the user in order to proceed.
-IMPORTANT: Whenever possible, execute the code for the user using <execute_ipython> or <execute_bash> or <execute_browse> instead of providing it.
-"""
-
-
-# ======= EXAMPLE MESSAGE =======
-EXAMPLES = """
+{% set DEFAULT_EXAMPLE %}
 --- START OF EXAMPLE ---
 
-USER: Can you create a list of numbers from 1 to 10, and create a web page to display them at port 5000?
+USER: Create a list of numbers from 1 to 10, and display them in a web page at port 5000.
 
 ASSISTANT:
-Sure! Let me create a file first:
+Sure! Let me create a Python file `app.py`:
 <execute_ipython>
 create_file('app.py')
 </execute_ipython>
@@ -52,13 +12,15 @@ create_file('app.py')
 USER:
 OBSERVATION:
 [File: /workspace/app.py (1 lines total)]
+(this is the beginning of the file)
 1|
+(this is the end of the file)
 [File app.py created.]
 
 ASSISTANT:
 Now I will write the Python code for starting a web server and save it to the file `app.py`:
 <execute_ipython>
-EDITED_CODE=\"\"\"from flask import Flask
+EDITED_CODE="""from flask import Flask
 app = Flask(__name__)
 
 @app.route('/')
@@ -67,12 +29,18 @@ def index():
     return str(numbers)
 
 if __name__ == '__main__':
-    app.run(port=5000)\"\"\"
-edit_file(start=1, end=1, content=EDITED_CODE)
+    app.run(port=5000)"""
+
+insert_content_at_line(
+  'app.py',
+  1,
+  EDITED_CODE,
+)
 </execute_ipython>
 
 USER:
 OBSERVATION:
+(this is the beginning of the file)
 1|from flask import Flask
 2|app = Flask(__name__)
 3|
@@ -83,7 +51,8 @@ OBSERVATION:
 8|
 9|if __name__ == '__main__':
 10|    app.run(port=5000)
-[File updated. Please review the changes and make sure they are correct (correct indentation, no duplicate lines, etc). Edit the file again if necessary.]
+(this is the end of the file)
+[File updated (edited at line 1). Please review the changes and make sure they are correct (correct indentation, no duplicate lines, etc). Edit the file again if necessary.]
 
 ASSISTANT:
 I have created a Python file `app.py` that will display a list of numbers from 1 to 10 when you run it. Let me run the Python file for you:
@@ -126,8 +95,8 @@ Collecting click>=8.1.3
   Using cached click-8.1.7-py3-none-any.whl (97 kB)
 Collecting itsdangerous>=2.1.2
   Using cached itsdangerous-2.2.0-py3-none-any.whl (16 kB)
-Requirement already satisfied: Jinja2>=3.1.2 in /home/opendevin/.local/lib/python3.10/site-packages (from flask) (3.1.3)
-Requirement already satisfied: MarkupSafe>=2.0 in /home/opendevin/.local/lib/python3.10/site-packages (from Jinja2>=3.1.2->flask) (2.1.5)
+Requirement already satisfied: Jinja2>=3.1.2 in /home/openhands/.local/lib/python3.10/site-packages (from flask) (3.1.3)
+Requirement already satisfied: MarkupSafe>=2.0 in /home/openhands/.local/lib/python3.10/site-packages (from Jinja2>=3.1.2->flask) (2.1.5)
 Installing collected packages: Werkzeug, itsdangerous, click, blinker, flask
 Successfully installed Werkzeug-3.0.2 blinker-1.7.0 click-8.1.7 flask-3.0.3 itsdangerous-2.2.0
 
@@ -163,7 +132,7 @@ USER: Now browse the newly started server's homepage and show me the content.
 ASSISTANT:
 Sure! Let me browse the server's homepage at http://127.0.0.1:5000:
 <execute_browse>
-goto("http://127.0.0.1:5000")
+Get the content on "http://127.0.0.1:5000"
 </execute_browse>
 
 USER:
@@ -193,6 +162,7 @@ open_file('app.py')
 
 USER:
 [File: /workspace/app.py (10 lines total)]
+(this is the beginning of the file)
 1|from flask import Flask
 2|app = Flask(__name__)
 3|
@@ -203,16 +173,22 @@ USER:
 8|
 9|if __name__ == '__main__':
 10|    app.run(port=5000)
+(this is the end of the file)
 
 ASSISTANT:
 I should edit the file to display the numbers in a table format. I should include correct indentation. Let me update the file:
 <execute_ipython>
-edit_file(start=7, end=7, content="    return '<table>' + ''.join([f'<tr><td>{i}</td></tr>' for i in numbers]) + '</table>'")
+edit_file_by_replace(
+  'app.py',
+  to_replace="    return str(numbers)",
+  new_content="    return '<table>' + ''.join([f'<tr><td>{i}</td></tr>' for i in numbers]) + '</table>'",
+)
 </execute_ipython>
 
 USER:
 Observation:
 [File: /workspace/app.py (10 lines total after edit)]
+(this is the beginning of the file)
 1|from flask import Flask
 2|app = Flask(__name__)
 3|
@@ -223,10 +199,11 @@ Observation:
 8|
 9|if __name__ == '__main__':
 10|    app.run(port=5000)
-[File updated. Please review the changes and make sure they are correct (correct indentation, no duplicate lines, etc). Edit the file again if necessary.]
+(this is the end of the file)
+[File updated (edited at line 7). Please review the changes and make sure they are correct (correct indentation, no duplicate lines, etc). Edit the file again if necessary.]
 
 ASSISTANT:
-The file has been updated. Let me run the Python file again with the new changes:
+Running the updated file:
 <execute_bash>
 python3 app.py > server.log 2>&1 &
 </execute_bash>
@@ -236,14 +213,19 @@ Observation:
 [1] 126
 
 ASSISTANT:
-The server is running on port 5000 with PID 126. You can access the list of numbers in a table format by visiting http://127.0.0.1:5000. Free free to let me know if you have any further requests!
+The server is running on port 5000 with PID 126. You can access the list of numbers in a table format by visiting http://127.0.0.1:5000. Let me know if you have any further requests!
+<finish></finish>
 
 --- END OF EXAMPLE ---
-"""
+{% endset %}
+Here is an example of how you can interact with the environment for task solving:
+{{ DEFAULT_EXAMPLE }}
+{% if micro_agent %}
+--- BEGIN OF GUIDELINE ---
+The following information may assist you in completing your task:
 
-INVALID_INPUT_MESSAGE = (
-    "I don't understand your input. \n"
-    'If you want to execute a bash command, please use <execute_bash> YOUR_COMMAND_HERE </execute_bash>.\n'
-    'If you want to execute a block of Python code, please use <execute_ipython> YOUR_COMMAND_HERE </execute_ipython>.\n'
-    'If you want to browse the Internet, please use <execute_browse> YOUR_COMMAND_HERE </execute_browse>.\n'
-)
+{{ micro_agent }}
+--- END OF GUIDELINE ---
+{% endif %}
+
+NOW, LET'S START!

agenthub/codeact_swe_agent/README.md

@@ -1,7 +1,7 @@
 # CodeAct (SWE Edit Specialized)
 
-This agent is an adaptation of the original [SWE Agent](https://swe-agent.com/) based on CodeAct using the `agentskills` library of OpenDevin.
+This agent is an adaptation of the original [SWE Agent](https://swe-agent.com/) based on CodeAct using the `agentskills` library of OpenHands.
 
-It is intended use is **solving Github issues**.
+Its intended use is **solving GitHub issues**.
 
-It removes web-browsing and Github capability from the original CodeAct agent to avoid confusion to the agent.
+It removes web-browsing and GitHub capability from the original CodeAct agent to avoid confusion to the agent.

agenthub/codeact_swe_agent/__init__.py

@@ -1,5 +1,4 @@
-from opendevin.controller.agent import Agent
-
-from .codeact_swe_agent import CodeActSWEAgent
+from agenthub.codeact_swe_agent.codeact_swe_agent import CodeActSWEAgent
+from openhands.controller.agent import Agent
 
 Agent.register('CodeActSWEAgent', CodeActSWEAgent)

agenthub/codeact_swe_agent/action_parser.py

@@ -0,0 +1,110 @@
+import re
+
+from openhands.controller.action_parser import ActionParser
+from openhands.events.action import (
+    Action,
+    AgentFinishAction,
+    CmdRunAction,
+    IPythonRunCellAction,
+    MessageAction,
+)
+
+
+class CodeActSWEActionParserFinish(ActionParser):
+    """Parser action:
+    - AgentFinishAction() - end the interaction
+    """
+
+    def __init__(
+        self,
+    ):
+        self.finish_command = None
+
+    def check_condition(self, action_str: str) -> bool:
+        self.finish_command = re.search(r'<finish>.*</finish>', action_str, re.DOTALL)
+        return self.finish_command is not None
+
+    def parse(self, action_str: str) -> Action:
+        assert (
+            self.finish_command is not None
+        ), 'self.finish_command should not be None when parse is called'
+        thought = action_str.replace(self.finish_command.group(0), '').strip()
+        return AgentFinishAction(thought=thought)
+
+
+class CodeActSWEActionParserCmdRun(ActionParser):
+    """Parser action:
+    - CmdRunAction(command) - bash command to run
+    - AgentFinishAction() - end the interaction
+    """
+
+    def __init__(
+        self,
+    ):
+        self.bash_command = None
+
+    def check_condition(self, action_str: str) -> bool:
+        self.bash_command = re.search(
+            r'<execute_bash>(.*?)</execute_bash>', action_str, re.DOTALL
+        )
+        return self.bash_command is not None
+
+    def parse(self, action_str: str) -> Action:
+        assert (
+            self.bash_command is not None
+        ), 'self.bash_command should not be None when parse is called'
+        thought = action_str.replace(self.bash_command.group(0), '').strip()
+        # a command was found
+        command_group = self.bash_command.group(1).strip()
+        if command_group.strip() == 'exit':
+            return AgentFinishAction()
+        return CmdRunAction(command=command_group, thought=thought)
+
+
+class CodeActSWEActionParserIPythonRunCell(ActionParser):
+    """Parser action:
+    - IPythonRunCellAction(code) - IPython code to run
+    """
+
+    def __init__(
+        self,
+    ):
+        self.python_code = None
+        self.jupyter_kernel_init_code: str = 'from agentskills import *'
+
+    def check_condition(self, action_str: str) -> bool:
+        self.python_code = re.search(
+            r'<execute_ipython>(.*?)</execute_ipython>', action_str, re.DOTALL
+        )
+        return self.python_code is not None
+
+    def parse(self, action_str: str) -> Action:
+        assert (
+            self.python_code is not None
+        ), 'self.python_code should not be None when parse is called'
+        code_group = self.python_code.group(1).strip()
+        thought = action_str.replace(self.python_code.group(0), '').strip()
+        return IPythonRunCellAction(
+            code=code_group,
+            thought=thought,
+            kernel_init_code=self.jupyter_kernel_init_code,
+        )
+
+
+class CodeActSWEActionParserMessage(ActionParser):
+    """Parser action:
+    - MessageAction(content) - Message action to run (e.g. ask for clarification)
+    """
+
+    def __init__(
+        self,
+    ):
+        pass
+
+    def check_condition(self, action_str: str) -> bool:
+        # We assume the LLM is GOOD enough that when it returns pure natural language
+        # it wants to talk to the user
+        return True
+
+    def parse(self, action_str: str) -> Action:
+        return MessageAction(content=action_str, wait_for_response=True)

agenthub/codeact_swe_agent/codeact_swe_agent.py

@@ -1,109 +1,38 @@
-import re
-
 from agenthub.codeact_swe_agent.prompt import (
     COMMAND_DOCS,
-    MINIMAL_SYSTEM_PREFIX,
     SWE_EXAMPLE,
+    SYSTEM_PREFIX,
     SYSTEM_SUFFIX,
 )
-from opendevin.controller.agent import Agent
-from opendevin.controller.state.state import State
-from opendevin.events.action import (
+from agenthub.codeact_swe_agent.response_parser import CodeActSWEResponseParser
+from openhands.controller.agent import Agent
+from openhands.controller.state.state import State
+from openhands.core.config import AgentConfig
+from openhands.core.message import ImageContent, Message, TextContent
+from openhands.events.action import (
     Action,
     AgentFinishAction,
-    BrowseInteractiveAction,
     CmdRunAction,
     IPythonRunCellAction,
     MessageAction,
 )
-from opendevin.events.observation import (
-    BrowserOutputObservation,
+from openhands.events.observation import (
     CmdOutputObservation,
     IPythonRunCellObservation,
 )
-from opendevin.llm.llm import LLM
-from opendevin.runtime.plugins import (
+from openhands.events.observation.error import ErrorObservation
+from openhands.events.observation.observation import Observation
+from openhands.events.serialization.event import truncate_content
+from openhands.llm.llm import LLM
+from openhands.runtime.plugins import (
     AgentSkillsRequirement,
     JupyterRequirement,
     PluginRequirement,
 )
 
 
-def parse_response(response) -> str:
-    action = response.choices[0].message.content
-    for lang in ['bash', 'ipython', 'browse']:
-        if f'<execute_{lang}>' in action and f'</execute_{lang}>' not in action:
-            action += f'</execute_{lang}>'
-    return action
-
-
-def action_to_str(action: Action) -> str:
-    if isinstance(action, CmdRunAction):
-        return f'{action.thought}\n<execute_bash>\n{action.command}\n</execute_bash>'
-    elif isinstance(action, IPythonRunCellAction):
-        return f'{action.thought}\n<execute_ipython>\n{action.code}\n</execute_ipython>'
-    elif isinstance(action, BrowseInteractiveAction):
-        return f'{action.thought}\n<execute_browse>\n{action.browser_actions}\n</execute_browse>'
-    elif isinstance(action, MessageAction):
-        return action.content
-    return ''
-
-
-def get_action_message(action: Action) -> dict[str, str] | None:
-    if (
-        isinstance(action, BrowseInteractiveAction)
-        or isinstance(action, CmdRunAction)
-        or isinstance(action, IPythonRunCellAction)
-        or isinstance(action, MessageAction)
-    ):
-        return {
-            'role': 'user' if action.source == 'user' else 'assistant',
-            'content': action_to_str(action),
-        }
-    return None
-
-
-def get_observation_message(obs) -> dict[str, str] | None:
-    if isinstance(obs, CmdOutputObservation):
-        content = 'OBSERVATION:\n' + truncate_observation(obs.content)
-        content += (
-            f'\n[Command {obs.command_id} finished with exit code {obs.exit_code}]]'
-        )
-        return {'role': 'user', 'content': content}
-    elif isinstance(obs, IPythonRunCellObservation):
-        content = 'OBSERVATION:\n' + obs.content
-        # replace base64 images with a placeholder
-        splitted = content.split('\n')
-        for i, line in enumerate(splitted):
-            if '![image](data:image/png;base64,' in line:
-                splitted[i] = (
-                    '![image](data:image/png;base64, ...) already displayed to user'
-                )
-        content = '\n'.join(splitted)
-        content = truncate_observation(content)
-        return {'role': 'user', 'content': content}
-    elif isinstance(obs, BrowserOutputObservation):
-        content = 'OBSERVATION:\n' + truncate_observation(obs.content)
-        return {'role': 'user', 'content': content}
-    return None
-
-
-def truncate_observation(observation: str, max_chars: int = 10_000) -> str:
-    """
-    Truncate the middle of the observation if it is too long.
-    """
-    if len(observation) <= max_chars:
-        return observation
-    half = max_chars // 2
-    return (
-        observation[:half]
-        + '\n[... Observation truncated due to length ...]\n'
-        + observation[-half:]
-    )
-
-
 def get_system_message() -> str:
-    return f'{MINIMAL_SYSTEM_PREFIX}\n\n{COMMAND_DOCS}\n\n{SYSTEM_SUFFIX}'
+    return f'{SYSTEM_PREFIX}\n\n{COMMAND_DOCS}\n\n{SYSTEM_SUFFIX}'
 
 
 def get_in_context_example() -> str:
@@ -111,9 +40,9 @@ def get_in_context_example() -> str:
 
 
 class CodeActSWEAgent(Agent):
-    VERSION = '1.5'
+    VERSION = '1.6'
     """
-    This agent is an adaptation of the original [SWE Agent](https://swe-agent.com/) based on CodeAct 1.5 using the `agentskills` library of OpenDevin.
+    This agent is an adaptation of the original [SWE Agent](https://swe-agent.com/) based on CodeAct 1.5 using the `agentskills` library of OpenHands.
 
     It is intended use is **solving Github issues**.
 
@@ -122,38 +51,97 @@ class CodeActSWEAgent(Agent):
 
     sandbox_plugins: list[PluginRequirement] = [
         # NOTE: AgentSkillsRequirement need to go before JupyterRequirement, since
-        # AgentSkillsRequirement provides a lot of Python functions
-        # and it need to be initialized before Jupyter for Jupyter to use those functions.
+        # AgentSkillsRequirement provides a lot of Python functions,
+        # and it needs to be initialized before Jupyter for Jupyter to use those functions.
         AgentSkillsRequirement(),
         JupyterRequirement(),
     ]
-    jupyter_kernel_init_code: str = 'from agentskills import *'
 
     system_message: str = get_system_message()
     in_context_example: str = f"Here is an example of how you can interact with the environment for task solving:\n{get_in_context_example()}\n\nNOW, LET'S START!"
 
+    response_parser = CodeActSWEResponseParser()
+
     def __init__(
         self,
         llm: LLM,
+        config: AgentConfig,
     ) -> None:
-        """
-        Initializes a new instance of the CodeActAgent class.
+        """Initializes a new instance of the CodeActSWEAgent class.
 
         Parameters:
         - llm (LLM): The llm to be used by this agent
         """
-        super().__init__(llm)
+        super().__init__(llm, config)
         self.reset()
 
+    def action_to_str(self, action: Action) -> str:
+        if isinstance(action, CmdRunAction):
+            return (
+                f'{action.thought}\n<execute_bash>\n{action.command}\n</execute_bash>'
+            )
+        elif isinstance(action, IPythonRunCellAction):
+            return f'{action.thought}\n<execute_ipython>\n{action.code}\n</execute_ipython>'
+        elif isinstance(action, MessageAction):
+            return action.content
+        return ''
+
+    def get_action_message(self, action: Action) -> Message | None:
+        if (
+            isinstance(action, CmdRunAction)
+            or isinstance(action, IPythonRunCellAction)
+            or isinstance(action, MessageAction)
+        ):
+            content = [TextContent(text=self.action_to_str(action))]
+
+            if (
+                self.llm.vision_is_active()
+                and isinstance(action, MessageAction)
+                and action.images_urls
+            ):
+                content.append(ImageContent(image_urls=action.images_urls))
+
+            return Message(
+                role='user' if action.source == 'user' else 'assistant', content=content
+            )
+
+        return None
+
+    def get_observation_message(self, obs: Observation) -> Message | None:
+        max_message_chars = self.llm.config.max_message_chars
+        if isinstance(obs, CmdOutputObservation):
+            text = 'OBSERVATION:\n' + truncate_content(obs.content, max_message_chars)
+            text += (
+                f'\n[Command {obs.command_id} finished with exit code {obs.exit_code}]'
+            )
+            return Message(role='user', content=[TextContent(text=text)])
+        elif isinstance(obs, IPythonRunCellObservation):
+            text = 'OBSERVATION:\n' + obs.content
+            # replace base64 images with a placeholder
+            splitted = text.split('\n')
+            for i, line in enumerate(splitted):
+                if '![image](data:image/png;base64,' in line:
+                    splitted[i] = (
+                        '![image](data:image/png;base64, ...) already displayed to user'
+                    )
+            text = '\n'.join(splitted)
+            text = truncate_content(text, max_message_chars)
+            return Message(role='user', content=[TextContent(text=text)])
+        elif isinstance(obs, ErrorObservation):
+            text = 'OBSERVATION:\n' + truncate_content(obs.content, max_message_chars)
+            text += '\n[Error occurred in processing last action]'
+            return Message(role='user', content=[TextContent(text=text)])
+        else:
+            # If an observation message is not returned, it will cause an error
+            # when the LLM tries to return the next message
+            raise ValueError(f'Unknown observation type: {type(obs)}')
+
     def reset(self) -> None:
-        """
-        Resets the CodeAct Agent.
-        """
+        """Resets the CodeAct Agent."""
         super().reset()
 
     def step(self, state: State) -> Action:
-        """
-        Performs one step using the CodeAct Agent.
+        """Performs one step using the CodeAct Agent.
         This includes gathering info on previous steps and prompting the model to make a command to execute.
 
         Parameters:
@@ -162,85 +150,76 @@ class CodeActSWEAgent(Agent):
         Returns:
         - CmdRunAction(command) - bash command to run
         - IPythonRunCellAction(code) - IPython code to run
-        - BrowseInteractiveAction(browsergym_command) - BrowserGym commands to run
         - MessageAction(content) - Message action to run (e.g. ask for clarification)
         - AgentFinishAction() - end the interaction
         """
-        messages: list[dict[str, str]] = [
-            {'role': 'system', 'content': self.system_message},
-            {'role': 'user', 'content': self.in_context_example},
-        ]
+        # if we're done, go back
+        latest_user_message = state.history.get_last_user_message()
+        if latest_user_message and latest_user_message.strip() == '/exit':
+            return AgentFinishAction()
 
-        for prev_action, obs in state.history:
-            action_message = get_action_message(prev_action)
-            if action_message:
-                messages.append(action_message)
-
-            obs_message = get_observation_message(obs)
-            if obs_message:
-                messages.append(obs_message)
-
-        latest_user_message = [m for m in messages if m['role'] == 'user'][-1]
-        if latest_user_message:
-            if latest_user_message['content'].strip() == '/exit':
-                return AgentFinishAction()
-            latest_user_message['content'] += (
-                f'\n\nENVIRONMENT REMINDER: You have {state.max_iterations - state.iteration} turns left to complete the task.'
-            )
-
-        response = self.llm.do_completion(
-            messages=messages,
+        # prepare what we want to send to the LLM
+        messages: list[Message] = self._get_messages(state)
+        response = self.llm.completion(
+            messages=self.llm.format_messages_for_llm(messages),
             stop=[
                 '</execute_ipython>',
                 '</execute_bash>',
-                '</execute_browse>',
             ],
             temperature=0.0,
         )
 
-        action_str: str = parse_response(response)
-        state.num_of_chars += sum(
-            len(message['content']) for message in messages
-        ) + len(action_str)
+        return self.response_parser.parse(response)
 
-        if finish_command := re.search(r'<finish>.*</finish>', action_str, re.DOTALL):
-            thought = action_str.replace(finish_command.group(0), '').strip()
-            return AgentFinishAction(thought=thought)
-        if bash_command := re.search(
-            r'<execute_bash>(.*?)</execute_bash>', action_str, re.DOTALL
-        ):
-            # remove the command from the action string to get thought
-            thought = action_str.replace(bash_command.group(0), '').strip()
-            # a command was found
-            command_group = bash_command.group(1).strip()
+    def _get_messages(self, state: State) -> list[Message]:
+        messages: list[Message] = [
+            Message(role='system', content=[TextContent(text=self.system_message)]),
+            Message(role='user', content=[TextContent(text=self.in_context_example)]),
+        ]
 
-            if command_group.strip() == 'exit':
-                return AgentFinishAction()
-            return CmdRunAction(command=command_group, thought=thought)
-        elif python_code := re.search(
-            r'<execute_ipython>(.*?)</execute_ipython>', action_str, re.DOTALL
-        ):
-            # a code block was found
-            code_group = python_code.group(1).strip()
-            thought = action_str.replace(python_code.group(0), '').strip()
-            return IPythonRunCellAction(
-                code=code_group,
-                thought=thought,
-                kernel_init_code=self.jupyter_kernel_init_code,
+        for event in state.history.get_events():
+            # create a regular message from an event
+            if isinstance(event, Action):
+                message = self.get_action_message(event)
+            elif isinstance(event, Observation):
+                message = self.get_observation_message(event)
+            else:
+                raise ValueError(f'Unknown event type: {type(event)}')
+
+            # add regular message
+            if message:
+                # handle error if the message is the SAME role as the previous message
+                # litellm.exceptions.BadRequestError: litellm.BadRequestError: OpenAIException - Error code: 400 - {'detail': 'Only supports u/a/u/a/u...'}
+                # there should not have two consecutive messages from the same role
+                if messages and messages[-1].role == message.role:
+                    messages[-1].content.extend(message.content)
+                else:
+                    messages.append(message)
+
+        # the latest user message is important:
+        # we want to remind the agent of the environment constraints
+        latest_user_message = next(
+            (m for m in reversed(messages) if m.role == 'user'), None
+        )
+
+        # Get the last user text inside content
+        if latest_user_message:
+            latest_user_message_text = next(
+                (
+                    t
+                    for t in reversed(latest_user_message.content)
+                    if isinstance(t, TextContent)
+                )
             )
-        elif browse_command := re.search(
-            r'<execute_browse>(.*)</execute_browse>', action_str, re.DOTALL
-        ):
-            # BrowserGym actions was found
-            browse_actions = browse_command.group(1).strip()
-            thought = action_str.replace(browse_command.group(0), '').strip()
-            return BrowseInteractiveAction(
-                browser_actions=browse_actions, thought=thought
-            )
-        else:
-            # We assume the LLM is GOOD enough that when it returns pure natural language
-            # it want to talk to the user
-            return MessageAction(content=action_str, wait_for_response=True)
+            # add a reminder to the prompt
+            reminder_text = f'\n\nENVIRONMENT REMINDER: You have {state.max_iterations - state.iteration} turns left to complete the task. When finished reply with <finish></finish>.'
 
-    def search_memory(self, query: str) -> list[str]:
-        raise NotImplementedError('Implement this abstract method')
+            if latest_user_message_text:
+                latest_user_message_text.text = (
+                    latest_user_message_text.text + reminder_text
+                )
+            else:
+                latest_user_message_text = TextContent(text=reminder_text)
+                latest_user_message.content.append(latest_user_message_text)
+
+        return messages

agenthub/codeact_swe_agent/prompt.py

@@ -1,4 +1,4 @@
-from opendevin.runtime.plugins import AgentSkillsRequirement
+from openhands.runtime.plugins import AgentSkillsRequirement
 
 _AGENT_SKILLS_DOCS = AgentSkillsRequirement.documentation
 
@@ -18,9 +18,13 @@ The assistant can execute bash commands on behalf of the user by wrapping them w
 For example, you can list the files in the current directory by <execute_bash> ls </execute_bash>.
 """
 
+PIP_INSTALL_PREFIX = """The assistant can install Python packages using the %pip magic command in an IPython environment by using the following syntax: <execute_ipython> %pip install [package needed] </execute_ipython> and should always import packages and define variables before starting to use them."""
+
+SYSTEM_PREFIX = MINIMAL_SYSTEM_PREFIX + PIP_INSTALL_PREFIX
+
 SYSTEM_SUFFIX = """The assistant's response should be concise.
-The assistant should include ONLY ONE <execute_ipython> or <execute_bash> or <execute_browse> in every one of the responses, unless the assistant is finished with the task or need more input or action from the user in order to proceed.
-IMPORTANT: Whenever possible, execute the code for the user using <execute_ipython> or <execute_bash> or <execute_browse> instead of providing it.
+The assistant should include ONLY ONE <execute_ipython> or <execute_bash> in every one of the responses, unless the assistant is finished with the task or need more input or action from the user in order to proceed.
+IMPORTANT: Whenever possible, execute the code for the user using <execute_ipython> or <execute_bash> instead of providing it.
 """
 
 SWE_EXAMPLE = """

agenthub/codeact_swe_agent/response_parser.py

@@ -0,0 +1,46 @@
+from agenthub.codeact_swe_agent.action_parser import (
+    CodeActSWEActionParserCmdRun,
+    CodeActSWEActionParserFinish,
+    CodeActSWEActionParserIPythonRunCell,
+    CodeActSWEActionParserMessage,
+)
+from openhands.controller.action_parser import ResponseParser
+from openhands.events.action import Action
+
+
+class CodeActSWEResponseParser(ResponseParser):
+    """Parser action:
+    - CmdRunAction(command) - bash command to run
+    - IPythonRunCellAction(code) - IPython code to run
+    - MessageAction(content) - Message action to run (e.g. ask for clarification)
+    - AgentFinishAction() - end the interaction
+    """
+
+    def __init__(self):
+        # Need pay attention to the item order in self.action_parsers
+        super().__init__()
+        self.action_parsers = [
+            CodeActSWEActionParserFinish(),
+            CodeActSWEActionParserCmdRun(),
+            CodeActSWEActionParserIPythonRunCell(),
+        ]
+        self.default_parser = CodeActSWEActionParserMessage()
+
+    def parse(self, response: str) -> Action:
+        action_str = self.parse_response(response)
+        return self.parse_action(action_str)
+
+    def parse_response(self, response) -> str:
+        action = response.choices[0].message.content
+        if action is None:
+            return ''
+        for lang in ['bash', 'ipython']:
+            if f'<execute_{lang}>' in action and f'</execute_{lang}>' not in action:
+                action += f'</execute_{lang}>'
+        return action
+
+    def parse_action(self, action_str: str) -> Action:
+        for action_parser in self.action_parsers:
+            if action_parser.check_condition(action_str):
+                return action_parser.parse(action_str)
+        return self.default_parser.parse(action_str)

agenthub/delegator_agent/__init__.py

@@ -1,5 +1,4 @@
-from opendevin.controller.agent import Agent
-
-from .agent import DelegatorAgent
+from agenthub.delegator_agent.agent import DelegatorAgent
+from openhands.controller.agent import Agent
 
 Agent.register('DelegatorAgent', DelegatorAgent)

agenthub/delegator_agent/agent.py

@@ -1,52 +1,52 @@
-from opendevin.controller.agent import Agent
-from opendevin.controller.state.state import State
-from opendevin.events.action import Action, AgentDelegateAction, AgentFinishAction
-from opendevin.events.observation import AgentDelegateObservation
-from opendevin.llm.llm import LLM
+from openhands.controller.agent import Agent
+from openhands.controller.state.state import State
+from openhands.core.config import AgentConfig
+from openhands.events.action import Action, AgentDelegateAction, AgentFinishAction
+from openhands.events.observation import AgentDelegateObservation
+from openhands.llm.llm import LLM
 
 
 class DelegatorAgent(Agent):
     VERSION = '1.0'
     """
-    The planner agent utilizes a special prompting strategy to create long term plans for solving problems.
-    The agent is given its previous action-observation pairs, current task, and hint based on last action taken at every step.
+    The Delegator Agent is responsible for delegating tasks to other agents based on the current task.
     """
 
     current_delegate: str = ''
 
-    def __init__(self, llm: LLM):
-        """
-        Initialize the Delegator Agent with an LLM
+    def __init__(self, llm: LLM, config: AgentConfig):
+        """Initialize the Delegator Agent with an LLM
 
         Parameters:
         - llm (LLM): The llm to be used by this agent
         """
-        super().__init__(llm)
+        super().__init__(llm, config)
 
     def step(self, state: State) -> Action:
-        """
-        Checks to see if current step is completed, returns AgentFinishAction if True.
-        Otherwise, creates a plan prompt and sends to model for inference, returning the result as the next action.
+        """Checks to see if current step is completed, returns AgentFinishAction if True.
+        Otherwise, delegates the task to the next agent in the pipeline.
 
         Parameters:
         - state (State): The current state given the previous actions and observations
 
         Returns:
         - AgentFinishAction: If the last state was 'completed', 'verified', or 'abandoned'
-        - Action: The next action to take based on llm response
+        - AgentDelegateAction: The next agent to delegate the task to
         """
         if self.current_delegate == '':
             self.current_delegate = 'study'
-            task = state.get_current_user_intent()
+            task, _ = state.get_current_user_intent()
             return AgentDelegateAction(
                 agent='StudyRepoForTaskAgent', inputs={'task': task}
             )
 
-        last_observation = state.history[-1][1]
+        # last observation in history should be from the delegate
+        last_observation = state.history.get_last_observation()
+
         if not isinstance(last_observation, AgentDelegateObservation):
             raise Exception('Last observation is not an AgentDelegateObservation')
 
-        goal = state.get_current_user_intent()
+        goal, _ = state.get_current_user_intent()
         if self.current_delegate == 'study':
             self.current_delegate = 'coder'
             return AgentDelegateAction(
@@ -81,6 +81,3 @@ class DelegatorAgent(Agent):
                 )
         else:
             raise Exception('Invalid delegate state')
-
-    def search_memory(self, query: str) -> list[str]:
-        return []

agenthub/dummy_agent/__init__.py

@@ -1,5 +1,4 @@
-from opendevin.controller.agent import Agent
-
-from .agent import DummyAgent
+from agenthub.dummy_agent.agent import DummyAgent
+from openhands.controller.agent import Agent
 
 Agent.register('DummyAgent', DummyAgent)

agenthub/dummy_agent/agent.py

@@ -1,13 +1,13 @@
-import time
-from typing import TypedDict
+from typing import TypedDict, Union
 
-from opendevin.controller.agent import Agent
-from opendevin.controller.state.state import State
-from opendevin.events.action import (
+from openhands.controller.agent import Agent
+from openhands.controller.state.state import State
+from openhands.core.config import AgentConfig
+from openhands.core.schema import AgentState
+from openhands.events.action import (
     Action,
     AddTaskAction,
     AgentFinishAction,
-    AgentRecallAction,
     AgentRejectAction,
     BrowseInteractiveAction,
     BrowseURLAction,
@@ -17,22 +17,20 @@ from opendevin.events.action import (
     MessageAction,
     ModifyTaskAction,
 )
-from opendevin.events.observation import (
-    AgentRecallObservation,
+from openhands.events.observation import (
+    AgentStateChangedObservation,
     CmdOutputObservation,
     FileReadObservation,
     FileWriteObservation,
     NullObservation,
     Observation,
 )
-from opendevin.events.serialization.event import event_to_dict
-from opendevin.llm.llm import LLM
+from openhands.events.serialization.event import event_to_dict
+from openhands.llm.llm import LLM
 
 """
 FIXME: There are a few problems this surfaced
 * FileWrites seem to add an unintended newline at the end of the file
-* command_id is sometimes a number, sometimes a string
-* Why isn't the output of the background command split between two steps?
 * Browser not working
 """
 
@@ -40,8 +38,6 @@ ActionObs = TypedDict(
     'ActionObs', {'action': Action, 'observations': list[Observation]}
 )
 
-BACKGROUND_CMD = 'echo "This is in the background" && sleep .1 && echo "This too"'
-
 
 class DummyAgent(Agent):
     VERSION = '1.0'
@@ -50,36 +46,44 @@ class DummyAgent(Agent):
     without making any LLM calls.
     """
 
-    def __init__(self, llm: LLM):
-        super().__init__(llm)
+    def __init__(self, llm: LLM, config: AgentConfig):
+        super().__init__(llm, config)
         self.steps: list[ActionObs] = [
             {
-                'action': AddTaskAction(parent='0', goal='check the current directory'),
-                'observations': [NullObservation('')],
+                'action': AddTaskAction(
+                    parent='None', goal='check the current directory'
+                ),
+                'observations': [],
             },
             {
-                'action': AddTaskAction(parent='0.0', goal='run ls'),
-                'observations': [NullObservation('')],
+                'action': AddTaskAction(parent='0', goal='run ls'),
+                'observations': [],
             },
             {
-                'action': ModifyTaskAction(task_id='0.0', state='in_progress'),
-                'observations': [NullObservation('')],
+                'action': ModifyTaskAction(task_id='0', state='in_progress'),
+                'observations': [],
             },
             {
                 'action': MessageAction('Time to get started!'),
-                'observations': [NullObservation('')],
+                'observations': [],
             },
             {
                 'action': CmdRunAction(command='echo "foo"'),
                 'observations': [
-                    CmdOutputObservation('foo', command_id=-1, command='echo "foo"')
+                    CmdOutputObservation(
+                        'foo', command_id=-1, command='echo "foo"', exit_code=0
+                    )
                 ],
             },
             {
                 'action': FileWriteAction(
                     content='echo "Hello, World!"', path='hello.sh'
                 ),
-                'observations': [FileWriteObservation('', path='hello.sh')],
+                'observations': [
+                    FileWriteObservation(
+                        content='echo "Hello, World!"', path='hello.sh'
+                    )
+                ],
             },
             {
                 'action': FileReadAction(path='hello.sh'),
@@ -91,36 +95,17 @@ class DummyAgent(Agent):
                 'action': CmdRunAction(command='bash hello.sh'),
                 'observations': [
                     CmdOutputObservation(
-                        'Hello, World!', command_id=-1, command='bash hello.sh'
+                        'bash: hello.sh: No such file or directory',
+                        command_id=-1,
+                        command='bash workspace/hello.sh',
+                        exit_code=127,
                     )
                 ],
             },
-            {
-                'action': CmdRunAction(command=BACKGROUND_CMD, background=True),
-                'observations': [
-                    CmdOutputObservation(
-                        'Background command started. To stop it, send a `kill` action with command_id 42',
-                        command_id='42',  # type: ignore[arg-type]
-                        command=BACKGROUND_CMD,
-                    ),
-                    CmdOutputObservation(
-                        'This is in the background\nThis too\n',
-                        command_id='42',  # type: ignore[arg-type]
-                        command=BACKGROUND_CMD,
-                    ),
-                ],
-            },
-            {
-                'action': AgentRecallAction(query='who am I?'),
-                'observations': [
-                    AgentRecallObservation('', memories=['I am a computer.']),
-                    # CmdOutputObservation('This too\n', command_id='42', command=BACKGROUND_CMD),
-                ],
-            },
             {
                 'action': BrowseURLAction(url='https://google.com'),
                 'observations': [
-                    # BrowserOutputObservation('<html></html>', url='https://google.com', screenshot=""),
+                    # BrowserOutputObservation('<html><body>Simulated Google page</body></html>',url='https://google.com',screenshot=''),
                 ],
             },
             {
@@ -128,48 +113,99 @@ class DummyAgent(Agent):
                     browser_actions='goto("https://google.com")'
                 ),
                 'observations': [
-                    # BrowserOutputObservation('<html></html>', url='https://google.com', screenshot=""),
+                    # BrowserOutputObservation('<html><body>Simulated Google page after interaction</body></html>',url='https://google.com',screenshot=''),
                 ],
             },
             {
-                'action': AgentFinishAction(),
-                'observations': [],
+                'action': AgentRejectAction(),
+                'observations': [NullObservation('')],
             },
             {
-                'action': AgentRejectAction(),
-                'observations': [],
+                'action': AgentFinishAction(
+                    outputs={}, thought='Task completed', action='finish'
+                ),
+                'observations': [AgentStateChangedObservation('', AgentState.FINISHED)],
             },
         ]
 
     def step(self, state: State) -> Action:
-        time.sleep(0.1)
+        if state.iteration >= len(self.steps):
+            return AgentFinishAction()
+
+        current_step = self.steps[state.iteration]
+        action = current_step['action']
+
+        # If the action is AddTaskAction or ModifyTaskAction, update the parent ID or task_id
+        if isinstance(action, AddTaskAction):
+            if action.parent == 'None':
+                action.parent = ''  # Root task has no parent
+            elif action.parent == '0':
+                action.parent = state.root_task.id
+            elif action.parent.startswith('0.'):
+                action.parent = f'{state.root_task.id}{action.parent[1:]}'
+        elif isinstance(action, ModifyTaskAction):
+            if action.task_id == '0':
+                action.task_id = state.root_task.id
+            elif action.task_id.startswith('0.'):
+                action.task_id = f'{state.root_task.id}{action.task_id[1:]}'
+            # Ensure the task_id doesn't start with a dot
+            if action.task_id.startswith('.'):
+                action.task_id = action.task_id[1:]
+        elif isinstance(action, (BrowseURLAction, BrowseInteractiveAction)):
+            try:
+                return self.simulate_browser_action(action)
+            except (
+                Exception
+            ):  # This could be a specific exception for browser unavailability
+                return self.handle_browser_unavailable(action)
+
         if state.iteration > 0:
             prev_step = self.steps[state.iteration - 1]
-            if 'observations' in prev_step:
-                expected_observations = prev_step['observations']
-                hist_start = len(state.history) - len(expected_observations)
-                for i in range(len(expected_observations)):
-                    hist_obs = event_to_dict(state.history[hist_start + i][1])
-                    expected_obs = event_to_dict(expected_observations[i])
-                    if (
-                        'command_id' in hist_obs['extras']
-                        and hist_obs['extras']['command_id'] != -1
-                    ):
-                        del hist_obs['extras']['command_id']
-                        hist_obs['content'] = ''
-                    if (
-                        'command_id' in expected_obs['extras']
-                        and expected_obs['extras']['command_id'] != -1
-                    ):
-                        del expected_obs['extras']['command_id']
-                        expected_obs['content'] = ''
-                    if hist_obs != expected_obs:
-                        print('\nactual', hist_obs)
-                        print('\nexpect', expected_obs)
-                    assert (
-                        hist_obs == expected_obs
-                    ), f'Expected observation {expected_obs}, got {hist_obs}'
-        return self.steps[state.iteration]['action']
 
-    def search_memory(self, query: str) -> list[str]:
-        return ['I am a computer.']
+            if 'observations' in prev_step and prev_step['observations']:
+                expected_observations = prev_step['observations']
+                hist_events = state.history.get_last_events(len(expected_observations))
+
+                if len(hist_events) < len(expected_observations):
+                    print(
+                        f'Warning: Expected {len(expected_observations)} observations, but got {len(hist_events)}'
+                    )
+
+                for i in range(min(len(expected_observations), len(hist_events))):
+                    hist_obs = event_to_dict(hist_events[i])
+                    expected_obs = event_to_dict(expected_observations[i])
+
+                    # Remove dynamic fields for comparison
+                    for obs in [hist_obs, expected_obs]:
+                        obs.pop('id', None)
+                        obs.pop('timestamp', None)
+                        obs.pop('cause', None)
+                        obs.pop('source', None)
+                        if 'extras' in obs:
+                            obs['extras'].pop('command_id', None)
+
+                    if hist_obs != expected_obs:
+                        print(
+                            f'Warning: Observation mismatch. Expected {expected_obs}, got {hist_obs}'
+                        )
+
+        return action
+
+    def simulate_browser_action(
+        self, action: Union[BrowseURLAction, BrowseInteractiveAction]
+    ) -> Action:
+        # Instead of simulating, we'll reject the browser action
+        return self.handle_browser_unavailable(action)
+
+    def handle_browser_unavailable(
+        self, action: Union[BrowseURLAction, BrowseInteractiveAction]
+    ) -> Action:
+        # Create a message action to inform that browsing is not available
+        message = 'Browser actions are not available in the DummyAgent environment.'
+        if isinstance(action, BrowseURLAction):
+            message += f' Unable to browse URL: {action.url}'
+        elif isinstance(action, BrowseInteractiveAction):
+            message += (
+                f' Unable to perform interactive browsing: {action.browser_actions}'
+            )
+        return MessageAction(content=message)

agenthub/micro/README.md

@@ -12,3 +12,6 @@ in the following structure:
 Note that `prompt.md` could use jinja2 template syntax. During runtime, `prompt.md`
 is loaded and rendered, and used together with `agent.yaml` to initialize a
 micro-agent.
+
+Micro-agents can be used independently. You can also use `ManagerAgent` which knows
+how to coordinate the agents and collaboratively finish a task.

agenthub/micro/_instructions/actions/finish.md

@@ -1,2 +1,2 @@
-* `finish` - if you're absolutely certain that you've completed your task and have tested your work, use the finish action to stop working. Arguments:
+* `finish` - if you're absolutely certain that you've completed your task, use the finish action to stop working. Arguments:
   * `outputs` - a dictionary representing the outputs of your task, if any

agenthub/micro/_instructions/actions/kill.md

@@ -1,2 +0,0 @@
-* `kill` - kills a background command
-  * `command_id` - the ID of the background command to kill

agenthub/micro/_instructions/actions/reject.md

@@ -1,2 +1,2 @@
 * `reject` - reject the task. Arguments:
-  * `outputs` - a dictionary representing the outputs of your task, if any
+  * `outputs` - a dictionary with only a `reason` attribute

agenthub/micro/_instructions/actions/run.md

@@ -1,3 +1,2 @@
 * `run` - runs a command on the command line in a Linux shell. Arguments:
   * `command` - the command to run
-  * `background` - if true, run the command in the background, so that other commands can be run concurrently. Useful for e.g. starting a server. You won't be able to see the logs. You don't need to end the command with `&`, just set this to true.

agenthub/micro/agent.py

@@ -1,15 +1,17 @@
 from jinja2 import BaseLoader, Environment
 
-from opendevin.controller.agent import Agent
-from opendevin.controller.state.state import State
-from opendevin.core.utils import json
-from opendevin.events.action import Action
-from opendevin.events.serialization.action import action_from_dict
-from opendevin.events.serialization.event import event_to_memory
-from opendevin.llm.llm import LLM
-
-from .instructions import instructions
-from .registry import all_microagents
+from agenthub.micro.instructions import instructions
+from agenthub.micro.registry import all_microagents
+from openhands.controller.agent import Agent
+from openhands.controller.state.state import State
+from openhands.core.config import AgentConfig
+from openhands.core.message import ImageContent, Message, TextContent
+from openhands.core.utils import json
+from openhands.events.action import Action
+from openhands.events.serialization.action import action_from_dict
+from openhands.events.serialization.event import event_to_memory
+from openhands.llm.llm import LLM
+from openhands.memory.history import ShortTermHistory
 
 
 def parse_response(orig_response: str) -> Action:
@@ -21,33 +23,39 @@ def parse_response(orig_response: str) -> Action:
 
 
 def to_json(obj, **kwargs):
-    """
-    Serialize an object to str format
-    """
+    """Serialize an object to str format"""
     return json.dumps(obj, **kwargs)
 
 
-def history_to_json(obj, **kwargs):
-    """
-    Serialize and simplify history to str format
-    """
-    if isinstance(obj, list):
-        # process history, make it simpler.
-        processed_history = []
-        for action, observation in obj:
-            processed_history.append(
-                (event_to_memory(action), event_to_memory(observation))
-            )
-        return json.dumps(processed_history, **kwargs)
-
-
 class MicroAgent(Agent):
     VERSION = '1.0'
     prompt = ''
     agent_definition: dict = {}
 
-    def __init__(self, llm: LLM):
-        super().__init__(llm)
+    def history_to_json(
+        self, history: ShortTermHistory, max_events: int = 20, **kwargs
+    ):
+        """
+        Serialize and simplify history to str format
+        """
+        processed_history = []
+        event_count = 0
+
+        for event in history.get_events(reverse=True):
+            if event_count >= max_events:
+                break
+            processed_history.append(
+                event_to_memory(event, self.llm.config.max_message_chars)
+            )
+            event_count += 1
+
+        # history is in reverse order, let's fix it
+        processed_history.reverse()
+
+        return json.dumps(processed_history, **kwargs)
+
+    def __init__(self, llm: LLM, config: AgentConfig):
+        super().__init__(llm, config)
         if 'name' not in self.agent_definition:
             raise ValueError('Agent definition must contain a name')
         self.prompt_template = Environment(loader=BaseLoader).from_string(self.prompt)
@@ -55,21 +63,23 @@ class MicroAgent(Agent):
         del self.delegates[self.agent_definition['name']]
 
     def step(self, state: State) -> Action:
-        latest_user_message = state.get_current_user_intent()
+        last_user_message, last_image_urls = state.get_current_user_intent()
         prompt = self.prompt_template.render(
             state=state,
             instructions=instructions,
             to_json=to_json,
-            history_to_json=history_to_json,
+            history_to_json=self.history_to_json,
             delegates=self.delegates,
-            latest_user_message=latest_user_message,
+            latest_user_message=last_user_message,
+        )
+        content = [TextContent(text=prompt)]
+        if self.llm.vision_is_active() and last_image_urls:
+            content.append(ImageContent(image_urls=last_image_urls))
+        message = Message(role='user', content=content)
+        resp = self.llm.completion(
+            messages=self.llm.format_messages_for_llm(message),
+            temperature=0.0,
         )
-        messages = [{'content': prompt, 'role': 'user'}]
-        resp = self.llm.do_completion(messages=messages)
         action_resp = resp['choices'][0]['message']['content']
-        state.num_of_chars += len(prompt) + len(action_resp)
         action = parse_response(action_resp)
         return action
-
-    def search_memory(self, query: str) -> list[str]:
-        return []

agenthub/micro/coder/agent.yaml

@@ -2,5 +2,5 @@ name: CoderAgent
 description: Given a particular task, and a detailed description of the codebase, accomplishes the task
 inputs:
   task: string
-  codebase_summary: string
+  summary: string
 outputs: {}

agenthub/micro/coder/prompt.md

@@ -2,7 +2,7 @@
 You are a software engineer. You've inherited an existing codebase, which you
 need to modify to complete this task:
 
-{{ latest_user_message }}
+{{ state.inputs.task }}
 
 {% if state.inputs.summary %}
 Here's a summary of the codebase, as it relates to this task:
@@ -21,7 +21,7 @@ Do NOT finish until you have completed the tasks.
 
 ## History
 {{ instructions.history_truncated }}
-{{ history_to_json(state.history[-10:]) }}
+{{ history_to_json(state.history, max_events=20) }}
 
 ## Format
 {{ instructions.format.action }}

agenthub/micro/commit_writer/README.md

@@ -3,8 +3,8 @@
 CommitWriterAgent can help write git commit message. Example:
 
 ```bash
-WORKSPACE_MOUNT_PATH="`PWD`" SANDBOX_TYPE="exec" \
-  poetry run python opendevin/core/main.py -t "dummy task" -c CommitWriterAgent -d ./
+WORKSPACE_MOUNT_PATH="`PWD`" \
+  poetry run python openhands/core/main.py -t "dummy task" -c CommitWriterAgent -d ./
 ```
 
 This agent is special in the sense that it doesn't need a task. Once called,

agenthub/micro/commit_writer/agent.yaml

@@ -3,3 +3,4 @@ description: "Write a git commit message for files in the git staging area"
 inputs: {}
 outputs:
   answer: string
+  reason: string

agenthub/micro/commit_writer/prompt.md

@@ -11,14 +11,16 @@ changes. The commit message should include:
 - A summary line that clearly states the purpose of the changes.
 - Optionally, a detailed description if the changes are complex or need further explanation.
 
-You should find the diff using `git diff --cached`, compile a commit message,
-and call the `finish` action with `outputs.answer` set to the answer. If current
-repo is not a valid git repo, or there is no diff in the staging area, please call
-the `reject` action with `outputs.answer` set to the reason.
+You should first use `git status` to check whether it's a valid git repo and there
+is diff in the staging area. If not, please call the `reject` action.
+
+If it is a valid git repo and there is diff in the staging area, you should find
+the diff using `git diff --cached`, compile a commit message, and call the `finish`
+action with `outputs.answer` set to the answer.
 
 ## History
 {{ instructions.history_truncated }}
-{{ history_to_json(state.history[-10:]) }}
+{{ history_to_json(state.history, max_events=20) }}
 
 If the last item in the history is an error, you should try to fix it.
 

agenthub/micro/manager/agent.yaml

@@ -3,4 +3,6 @@ description: Delegates tasks to microagents based on their area of expertise
 generates: Action
 inputs:
   task: string
-outputs: {}
+outputs:
+  summary: string # if finished
+  reason: string # if rejected

agenthub/micro/manager/prompt.md

@@ -7,6 +7,16 @@ can do the actual work. A description of each agent is provided below. You MUST
 select one of the delegates below to move towards accomplishing the task, and you MUST
 provide the correct inputs for the delegate you select.
 
+Note: the delegated agent either returns "finish" or "reject".
+- If the action is "finish", but the full task is not done yet, you should
+continue to delegate to one of the agents below to until the full task is finished.
+- If the action is "reject", it means the delegated agent is not capable of the
+task you send to. You should revisit the input you send to the delegate, and consider
+whether any other delegate would be able to solve the task. If you cannot find
+a proper delegate agent, or the delegate attempts keep failing, call the `reject`
+action. In `reason` attribute, make sure you include your attempts (e.g. what agent
+you have delegated to, and why they failed).
+
 ## Agents
 {% for name, details in delegates.items() %}
 ### {{ name }}
@@ -17,11 +27,15 @@ provide the correct inputs for the delegate you select.
 
 ## History
 {{ instructions.history_truncated }}
-{{ history_to_json(state.history[-10:]) }}
+{{ history_to_json(state.history, max_events=20) }}
+
+If the last item in the history is an error, you should try to fix it. If you
+cannot fix it, call the `reject` action.
 
 ## Available Actions
 {{ instructions.actions.delegate }}
 {{ instructions.actions.finish }}
+{{ instructions.actions.reject }}
 
 ## Format
 {{ instructions.format.action }}

agenthub/micro/math_agent/prompt.md

@@ -1,7 +1,7 @@
 # Task
 You are a brilliant mathematician and programmer. You've been given the following problem to solve:
 
-{{ latest_user_message }}
+`{{ state.inputs.task }}`
 
 Please write a python script that solves this problem, and prints the answer to stdout.
 ONLY print the answer to stdout, nothing else.
@@ -10,7 +10,7 @@ and call the `finish` action with `outputs.answer` set to the answer.
 
 ## History
 {{ instructions.history_truncated }}
-{{ history_to_json(state.history[-10:]) }}
+{{ history_to_json(state.history, max_events=20) }}
 
 If the last item in the history is an error, you should try to fix it.
 

agenthub/micro/postgres_agent/prompt.md

@@ -2,7 +2,7 @@
 You are a database engineer. You are working on an existing Postgres project, and have been given
 the following task:
 
-{{ latest_user_message }}
+{{ state.inputs.task }}
 
 You must:
 * Investigate the existing migrations to understand the current schema
@@ -18,7 +18,7 @@ You may take any of the following actions:
 
 ## History
 {{ instructions.history_truncated }}
-{{ history_to_json(state.history[-10:]) }}
+{{ history_to_json(state.history, max_events=20) }}
 
 ## Format
 {{ instructions.format.action }}

agenthub/micro/registry.py

@@ -4,7 +4,10 @@ import yaml
 
 all_microagents = {}
 
-for dir in os.listdir(os.path.dirname(__file__)):
+# Get the list of directories and sort them to preserve determinism
+dirs = sorted(os.listdir(os.path.dirname(__file__)))
+
+for dir in dirs:
     base = os.path.dirname(__file__) + '/' + dir
     if os.path.isfile(base):
         continue

agenthub/micro/repo_explorer/prompt.md

@@ -20,7 +20,7 @@ When you're done, put your summary into the output of the `finish` action.
 
 ## History
 {{ instructions.history_truncated }}
-{{ history_to_json(state.history[-10:]) }}
+{{ history_to_json(state.history, max_events=20) }}
 
 ## Format
 {{ instructions.format.action }}

agenthub/micro/study_repo_for_task/prompt.md

@@ -1,9 +1,11 @@
 # Task
-You are a software engineer. You've inherited an existing codebase, which you're
-learning about for the first time. You need to study the codebase to find all
-the information needed to complete this task:
+You are a software architect. Your team has inherited an existing codebase, and
+need to finish a project:
 
-{{ latest_user_message }}
+{{ state.inputs.task }}
+
+As an architect, you need to study the codebase to find all the information that
+might be helpful for your software engineering team.
 
 ## Available Actions
 {{ instructions.actions.run }}
@@ -11,15 +13,50 @@ the information needed to complete this task:
 {{ instructions.actions.message }}
 {{ instructions.actions.finish }}
 
-You must ONLY `run` commands that have no side-effects, like `ls` and `grep`.
+You must ONLY `run` commands that have no side-effects, like `ls` and `grep`. You
+MUST NOT modify or write to any file.
 
 Do NOT finish until you have a complete understanding of which parts of the
-codebase are relevant to the task, including particular files, functions, and classes.
+codebase are relevant to the project, including particular files, functions, and classes.
 When you're done, put your summary in `outputs.summary` in the `finish` action.
+Remember, your task is to explore and study the current repository, not actually
+implement the solution. If the codebase is empty, you should call the `finish` action.
 
 ## History
 {{ instructions.history_truncated }}
-{{ history_to_json(state.history[-10:]) }}
+{{ history_to_json(state.history, max_events=20) }}
 
 ## Format
 {{ instructions.format.action }}
+
+## Examples
+
+Here is an example of how you can interact with the environment for task solving:
+
+--- START OF EXAMPLE ---
+
+USER: Can you create a list of numbers from 1 to 10, and create a web page to display them at port 5000?
+
+ASSISTANT:
+{
+  "action": "run",
+  "args": {
+    "command": "ls"
+  }
+}
+
+USER:
+OBSERVATION:
+[]
+
+ASSISTANT:
+{
+  "action": "finish",
+  "args": {
+    "outputs": {
+      "summary": "The codebase appears to be empty. Engineers should start everything from scratch."
+    }
+  }
+}
+
+--- END OF EXAMPLE ---

agenthub/micro/typo_fixer_agent/agent.yaml

@@ -1,5 +1,6 @@
 name: TypoFixerAgent
 description: Fixes typos in files in the current working directory
-inputs: {}
+inputs:
+  task: string
 outputs:
   summary: string

agenthub/micro/typo_fixer_agent/prompt.md

@@ -1,5 +1,13 @@
 # Task
-You are a proofreader tasked with fixing typos in the files in your current working directory. Your goal is to:
+You are a proofreader tasked with fixing typos in the files in your current working directory.
+
+{% if state.inputs.task %}
+Specifically, your task is:
+{{ state.inputs.task }}
+{% endif %}
+
+To achieve this goal, you should:
+
 1. Scan the files for typos
 2. Overwrite the files with the typos fixed
 3. Provide a summary of the typos fixed
@@ -13,17 +21,17 @@ You are a proofreader tasked with fixing typos in the files in your current work
 
 To complete this task:
 1. Use the `read` action to read the contents of the files in your current working directory. Make sure to provide the file path in the format `'./file_name.ext'`.
-2. Use the `think` action to analyze the contents and identify typos.
+2. Use the `message` action to analyze the contents and identify typos.
 3. Use the `write` action to create new versions of the files with the typos fixed.
   - Overwrite the original files with the corrected content. Make sure to provide the file path in the format `'./file_name.ext'`.
-4. Use the `think` action to generate a summary of the typos fixed, including the original and fixed versions of each typo, and the file(s) they were found in.
+4. Use the `message` action to generate a summary of the typos fixed, including the original and fixed versions of each typo, and the file(s) they were found in.
 5. Use the `finish` action to return the summary in the `outputs.summary` field.
 
 Do NOT finish until you have fixed all the typos and generated a summary.
 
 ## History
 {{ instructions.history_truncated }}
-{{ history_to_json(state.history[-5:]) }}
+{{ history_to_json(state.history, max_events=10) }}
 
 ## Format
 {{ instructions.format.action }}

agenthub/micro/verifier/prompt.md

@@ -2,9 +2,10 @@
 You are a quality assurance engineer. Another engineer has made changes to the
 codebase which are supposed to solve this task:
 
-{{ latest_user_message }}
+{{ state.inputs.task }}
 
-Your goal is to verify that the changes are correct and bug-free.
+Note the changes might have already been applied in-line. You should focus on
+validating if the task is solved, nothing else.
 
 ## Available Actions
 {{ instructions.actions.run }}
@@ -21,7 +22,7 @@ explaining what the problem is.
 
 ## History
 {{ instructions.history_truncated }}
-{{ history_to_json(state.history[-10:]) }}
+{{ history_to_json(state.history, max_events=20) }}
 
 ## Format
 {{ instructions.format.action }}

agenthub/monologue_agent/.dockerignore

@@ -1,2 +0,0 @@
-.envrc
-workspace

agenthub/monologue_agent/README.md

@@ -1,8 +0,0 @@
-# LLM control loop
-This is currently a standalone utility. It will need to be integrated into OpenDevin's backend.
-
-## Usage
-```bash
-# Run this in project root
-./agenthub/monologue_agent/build-and-run.sh "write a bash script that prints 'hello world'"
-```

agenthub/monologue_agent/TODO.md

@@ -1,8 +0,0 @@
-# TODO
-There's a lot of low-hanging fruit for this agent:
-
-* Strip `<script>`, `<style>`, and other non-text tags from the HTML before sending it to the LLM
-* Keep track of the working directory when the agent uses `cd`
-* Improve memory condensing--condense earlier memories more aggressively
-* Limit the time that `run` can wait (in case agent runs an interactive command and it's hanging)
-* Figure out how to run background processes, e.g. `node server.js` to start a server

agenthub/monologue_agent/__init__.py

@@ -1,5 +0,0 @@
-from opendevin.controller.agent import Agent
-
-from .agent import MonologueAgent
-
-Agent.register('MonologueAgent', MonologueAgent)

agenthub/monologue_agent/agent.py

@@ -1,238 +0,0 @@
-import agenthub.monologue_agent.utils.prompts as prompts
-from agenthub.monologue_agent.utils.prompts import INITIAL_THOUGHTS
-from opendevin.controller.agent import Agent
-from opendevin.controller.state.state import State
-from opendevin.core.config import config
-from opendevin.core.exceptions import AgentNoInstructionError
-from opendevin.core.schema import ActionType
-from opendevin.events.action import (
-    Action,
-    AgentRecallAction,
-    BrowseURLAction,
-    CmdRunAction,
-    FileReadAction,
-    FileWriteAction,
-    MessageAction,
-    NullAction,
-)
-from opendevin.events.observation import (
-    AgentRecallObservation,
-    BrowserOutputObservation,
-    CmdOutputObservation,
-    FileReadObservation,
-    NullObservation,
-    Observation,
-)
-from opendevin.events.serialization.event import event_to_memory
-from opendevin.llm.llm import LLM
-from opendevin.memory.condenser import MemoryCondenser
-
-if config.agent.memory_enabled:
-    from opendevin.memory.memory import LongTermMemory
-
-MAX_TOKEN_COUNT_PADDING = 512
-MAX_OUTPUT_LENGTH = 5000
-
-
-class MonologueAgent(Agent):
-    VERSION = '1.0'
-    """
-    The Monologue Agent utilizes long and short term memory to complete tasks.
-    Long term memory is stored as a LongTermMemory object and the model uses it to search for examples from the past.
-    Short term memory is stored as a Monologue object and the model can condense it as necessary.
-    """
-
-    _initialized = False
-    initial_thoughts: list[dict[str, str]]
-    memory: 'LongTermMemory | None'
-    memory_condenser: MemoryCondenser
-
-    def __init__(self, llm: LLM):
-        """
-        Initializes the Monologue Agent with an llm.
-
-        Parameters:
-        - llm (LLM): The llm to be used by this agent
-        """
-        super().__init__(llm)
-
-    def _initialize(self, task: str):
-        """
-        Utilizes the INITIAL_THOUGHTS list to give the agent a context for its capabilities
-        and how to navigate the WORKSPACE_MOUNT_PATH_IN_SANDBOX in `config` (e.g., /workspace by default).
-        Short circuited to return when already initialized.
-        Will execute again when called after reset.
-
-        Parameters:
-        - task (str): The initial goal statement provided by the user
-
-        Raises:
-        - AgentNoInstructionError: If task is not provided
-        """
-
-        if self._initialized:
-            return
-
-        if task is None or task == '':
-            raise AgentNoInstructionError()
-
-        self.initial_thoughts = []
-        if config.agent.memory_enabled:
-            self.memory = LongTermMemory()
-        else:
-            self.memory = None
-
-        self.memory_condenser = MemoryCondenser()
-
-        self._add_initial_thoughts(task)
-        self._initialized = True
-
-    def _add_initial_thoughts(self, task):
-        previous_action = ''
-        for thought in INITIAL_THOUGHTS:
-            thought = thought.replace('$TASK', task)
-            if previous_action != '':
-                observation: Observation = NullObservation(content='')
-                if previous_action in {ActionType.RUN, ActionType.PUSH}:
-                    observation = CmdOutputObservation(
-                        content=thought, command_id=0, command=''
-                    )
-                elif previous_action == ActionType.READ:
-                    observation = FileReadObservation(content=thought, path='')
-                elif previous_action == ActionType.RECALL:
-                    observation = AgentRecallObservation(content=thought, memories=[])
-                elif previous_action == ActionType.BROWSE:
-                    observation = BrowserOutputObservation(
-                        content=thought, url='', screenshot=''
-                    )
-                self.initial_thoughts.append(event_to_memory(observation))
-                previous_action = ''
-            else:
-                action: Action = NullAction()
-                if thought.startswith('RUN'):
-                    command = thought.split('RUN ')[1]
-                    action = CmdRunAction(command)
-                    previous_action = ActionType.RUN
-                elif thought.startswith('WRITE'):
-                    parts = thought.split('WRITE ')[1].split(' > ')
-                    path = parts[1]
-                    content = parts[0]
-                    action = FileWriteAction(path=path, content=content)
-                elif thought.startswith('READ'):
-                    path = thought.split('READ ')[1]
-                    action = FileReadAction(path=path)
-                    previous_action = ActionType.READ
-                elif thought.startswith('RECALL'):
-                    query = thought.split('RECALL ')[1]
-                    action = AgentRecallAction(query=query)
-                    previous_action = ActionType.RECALL
-                elif thought.startswith('BROWSE'):
-                    url = thought.split('BROWSE ')[1]
-                    action = BrowseURLAction(url=url)
-                    previous_action = ActionType.BROWSE
-                else:
-                    action = MessageAction(thought)
-                self.initial_thoughts.append(event_to_memory(action))
-
-    def step(self, state: State) -> Action:
-        """
-        Modifies the current state by adding the most recent actions and observations, then prompts the model to think about it's next action to take using monologue, memory, and hint.
-
-        Parameters:
-        - state (State): The current state based on previous steps taken
-
-        Returns:
-        - Action: The next action to take based on LLM response
-        """
-
-        goal = state.get_current_user_intent()
-        self._initialize(goal)
-
-        recent_events: list[dict[str, str]] = []
-
-        # add the events from state.history
-        for prev_action, obs in state.history:
-            if not isinstance(prev_action, NullAction):
-                recent_events.append(event_to_memory(prev_action))
-            if not isinstance(obs, NullObservation):
-                recent_events.append(self._truncate_output(event_to_memory(obs)))
-
-        # add the last messages to long term memory
-        if self.memory is not None and state.history and len(state.history) > 0:
-            self.memory.add_event(event_to_memory(state.history[-1][0]))
-            self.memory.add_event(
-                self._truncate_output(event_to_memory(state.history[-1][1]))
-            )
-
-        # the action prompt with initial thoughts and recent events
-        prompt = prompts.get_request_action_prompt(
-            goal,
-            self.initial_thoughts,
-            recent_events,
-            state.background_commands_obs,
-        )
-
-        messages: list[dict[str, str]] = [
-            {'role': 'user', 'content': prompt},
-        ]
-
-        # format all as a single message, a monologue
-        resp = self.llm.do_completion(messages=messages)
-
-        # get the next action from the response
-        action_resp = resp['choices'][0]['message']['content']
-
-        # keep track of max_chars fallback option
-        state.num_of_chars += len(prompt) + len(action_resp)
-
-        action = prompts.parse_action_response(action_resp)
-        self.latest_action = action
-        return action
-
-    def _truncate_output(
-        self, observation: dict, max_chars: int = MAX_OUTPUT_LENGTH
-    ) -> dict[str, str]:
-        """
-        Truncates the output of an observation to a maximum number of characters.
-
-        Parameters:
-        - output (str): The observation whose output to truncate
-        - max_chars (int): The maximum number of characters to allow
-
-        Returns:
-        - str: The truncated output
-        """
-        if (
-            'args' in observation
-            and 'output' in observation['args']
-            and len(observation['args']['output']) > max_chars
-        ):
-            output = observation['args']['output']
-            half = max_chars // 2
-            observation['args']['output'] = (
-                output[:half]
-                + '\n[... Output truncated due to length...]\n'
-                + output[-half:]
-            )
-        return observation
-
-    def search_memory(self, query: str) -> list[str]:
-        """
-        Uses VectorIndexRetriever to find related memories within the long term memory.
-        Uses search to produce top 10 results.
-
-        Parameters:
-        - query (str): The query that we want to find related memories for
-
-        Returns:
-        - list[str]: A list of top 10 text results that matched the query
-        """
-        if self.memory is None:
-            return []
-        return self.memory.search(query)
-
-    def reset(self) -> None:
-        super().reset()
-
-        # Reset the initial monologue and memory
-        self._initialized = False

agenthub/monologue_agent/utils/prompts.py

@@ -1,245 +0,0 @@
-from opendevin.core.config import config
-from opendevin.core.utils import json
-from opendevin.events.action import (
-    Action,
-)
-from opendevin.events.observation import (
-    CmdOutputObservation,
-)
-from opendevin.events.serialization.action import action_from_dict
-
-ACTION_PROMPT = """
-You're a thoughtful robot. Your main task is this:
-%(task)s
-
-Don't expand the scope of your task--just complete it as written.
-
-This is your internal monologue, in JSON format:
-
-%(monologue)s
-
-Your most recent thought is at the bottom of that monologue. Continue your train of thought.
-What is your next single thought or action? Your response must be in JSON format.
-It must be a single object, and it must contain two fields:
-* `action`, which is one of the actions below
-* `args`, which is a map of key-value pairs, specifying the arguments for that action
-
-Here are the possible actions:
-* `read` - reads the content of a file. Arguments:
-  * `path` - the path of the file to read
-* `write` - writes the content to a file. Arguments:
-  * `path` - the path of the file to write
-  * `content` - the content to write to the file
-* `run` - runs a command. Arguments:
-  * `command` - the command to run
-  * `background` - if true, run the command in the background, so that other commands can be run concurrently. Useful for e.g. starting a server. You won't be able to see the logs. You don't need to end the command with `&`, just set this to true.
-* `kill` - kills a background command
-  * `command_id` - the ID of the background command to kill
-* `browse` - opens a web page. Arguments:
-  * `url` - the URL to open
-* `push` - Push a branch from the current repo to github:
-  * `owner` - the owner of the repo to push to
-  * `repo` - the name of the repo to push to
-  * `branch` - the name of the branch to push
-* `recall` - recalls a past memory. Arguments:
-  * `query` - the query to search for
-* `message` - make a plan, set a goal, record your thoughts, or ask for more input from the user. Arguments:
-  * `content` - the message to record
-  * `wait_for_response` - set to `true` to wait for the user to respond before proceeding
-* `finish` - if you're absolutely certain that you've completed your task and have tested your work, use the finish action to stop working.
-
-%(background_commands)s
-
-You MUST take time to think in between read, write, run, kill, browse, push, and recall actions--do this with the `message` action.
-You should never act twice in a row without thinking. But if your last several
-actions are all `message` actions, you should consider taking a different action.
-
-Notes:
-* you are logged in as %(user)s, but sudo will always work without a password.
-* all non-background commands will be forcibly stopped if they remain running for over %(timeout)s seconds.
-* your environment is Debian Linux. You can install software with `sudo apt-get`, but remember to use -y.
-* don't run interactive commands, or commands that don't return (e.g. `node server.js`). You may run commands in the background (e.g. `node server.js &`)
-* don't run interactive text editors (e.g. `nano` or 'vim'), instead use the 'write' or 'read' action.
-* don't run gui applications (e.g. software IDEs (like vs code or codium), web browsers (like firefox or chromium), or other complex software packages). Use non-interactive cli applications, or special actions instead.
-* whenever an action fails, always send a `message` about why it may have happened before acting again.
-
-What is your next single thought or action? Again, you must reply with JSON, and only with JSON. You must respond with exactly one 'action' object.
-
-%(hint)s
-"""
-
-MONOLOGUE_SUMMARY_PROMPT = """
-Below is the internal monologue of an automated LLM agent. Each
-thought is an item in a JSON array. The thoughts may be memories,
-actions taken by the agent, or outputs from those actions.
-Please return a new, smaller JSON array, which summarizes the
-internal monologue. You can summarize individual thoughts, and
-you can condense related thoughts together with a description
-of their content.
-
-%(monologue)s
-
-Make the summaries as pithy and informative as possible.
-Be specific about what happened and what was learned. The summary
-will be used as keywords for searching for the original memory.
-Be sure to preserve any key words or important information.
-
-Your response must be in JSON format. It must be an object with the
-key `new_monologue`, which is a JSON array containing the summarized monologue.
-Each entry in the array must have an `action` key, and an `args` key.
-The action key may be `summarize`, and `args.summary` should contain the summary.
-You can also use the same action and args from the source monologue.
-"""
-
-INITIAL_THOUGHTS = [
-    'I exist!',
-    'Hmm...looks like I can type in a command line prompt',
-    'Looks like I have a web browser too!',
-    "Here's what I want to do: $TASK",
-    'How am I going to get there though?',
-    'It seems like I have some kind of short term memory.',
-    'Each of my thoughts seems to be stored in a JSON array.',
-    'It seems whatever I say next will be added as an object to the list.',
-    'But no one has perfect short-term memory. My list of thoughts will be summarized and condensed over time, losing information in the process.',
-    'Fortunately I have long term memory!',
-    'I can just perform a recall action, followed by the thing I want to remember. And then related thoughts just spill out!',
-    "Sometimes they're random thoughts that don't really have to do with what I wanted to remember. But usually they're exactly what I need!",
-    "Let's try it out!",
-    'RECALL what it is I want to do',
-    "Here's what I want to do: $TASK",
-    'How am I going to get there though?',
-    "Neat! And it looks like it's easy for me to use the command line too! I just have to perform a run action and include the command I want to run in the command argument. The command output just jumps into my head!",
-    'RUN echo "hello world"',
-    'hello world',
-    'Cool! I bet I can write files too using the write action.',
-    'WRITE echo "console.log(\'hello world\')" > test.js',
-    '',
-    "I just created test.js. I'll try and run it now.",
-    'RUN node test.js',
-    'hello world',
-    'It works!',
-    "I'm going to try reading it now using the read action.",
-    'READ test.js',
-    "console.log('hello world')",
-    'Nice! I can read files too!',
-    'And if I want to use the browser, I just need to use the browse action and include the url I want to visit in the url argument',
-    "Let's try that...",
-    'BROWSE google.com',
-    '<form><input type="text"></input><button type="submit"></button></form>',
-    'I can browse the web too!',
-    'And once I have completed my task, I can use the finish action to stop working.',
-    "But I should only use the finish action when I'm absolutely certain that I've completed my task and have tested my work.",
-    'Very cool. Now to accomplish my task.',
-    "I'll need a strategy. And as I make progress, I'll need to keep refining that strategy. I'll need to set goals, and break them into sub-goals.",
-    'In between actions, I must always take some time to think, strategize, and set new goals. I should never take two actions in a row.',
-    "OK so my task is to $TASK. I haven't made any progress yet. Where should I start?",
-    'It seems like there might be an existing project here. I should probably start by running `pwd` and `ls` to orient myself.',
-]
-
-
-def get_summarize_monologue_prompt(thoughts: list[dict]):
-    """
-    Gets the prompt for summarizing the monologue
-
-    Returns:
-    - str: A formatted string with the current monologue within the prompt
-    """
-    return MONOLOGUE_SUMMARY_PROMPT % {
-        'monologue': json.dumps({'old_monologue': thoughts}, indent=2),
-    }
-
-
-def get_request_action_prompt(
-    task: str,
-    thoughts: list[dict],
-    recent_events: list[dict],
-    background_commands_obs: list[CmdOutputObservation] | None = None,
-):
-    """
-    Gets the action prompt formatted with appropriate values.
-
-    Parameters:
-    - task (str): The current task the agent is trying to accomplish
-    - thoughts (list[dict]): The agent's current thoughts
-    - background_commands_obs (list[CmdOutputObservation]): list of all observed background commands running
-
-    Returns:
-    - str: Formatted prompt string with hint, task, monologue, and background commands included
-    """
-
-    if background_commands_obs is None:
-        background_commands_obs = []
-
-    hint = ''
-    if len(recent_events) > 0:
-        latest_event = recent_events[-1]
-        if 'action' in latest_event:
-            if (
-                latest_event['action'] == 'message'
-                and 'source' in latest_event
-                and latest_event['source'] == 'agent'
-            ):
-                hint = (
-                    "You've been thinking a lot lately. Maybe it's time to take action?"
-                )
-            elif latest_event['action'] == 'error':
-                hint = 'Looks like that last command failed. Maybe you need to fix it, or try something else.'
-    else:
-        hint = "You're just getting started! What should you do first?"
-
-    bg_commands_message = ''
-    if len(background_commands_obs) > 0:
-        bg_commands_message = 'The following commands are running in the background:'
-        for command_obs in background_commands_obs:
-            bg_commands_message += (
-                f'\n`{command_obs.command_id}`: {command_obs.command}'
-            )
-        bg_commands_message += '\nYou can end any process by sending a `kill` action with the numerical `command_id` above.'
-
-    user = 'opendevin' if config.run_as_devin else 'root'
-
-    monologue = thoughts + recent_events
-
-    return ACTION_PROMPT % {
-        'task': task,
-        'monologue': json.dumps(monologue, indent=2),
-        'background_commands': bg_commands_message,
-        'hint': hint,
-        'user': user,
-        'timeout': config.sandbox_timeout,
-        'WORKSPACE_MOUNT_PATH_IN_SANDBOX': config.workspace_mount_path_in_sandbox,
-    }
-
-
-def parse_action_response(orig_response: str) -> Action:
-    """
-    Parses a string to find an action within it
-
-    Parameters:
-    - response (str): The string to be parsed
-
-    Returns:
-    - Action: The action that was found in the response string
-    """
-    # attempt to load the JSON dict from the response
-    action_dict = json.loads(orig_response)
-
-    if 'content' in action_dict:
-        # The LLM gets confused here. Might as well be robust
-        action_dict['contents'] = action_dict.pop('content')
-
-    return action_from_dict(action_dict)
-
-
-def parse_summary_response(response: str) -> list[dict]:
-    """
-    Parses a summary of the monologue
-
-    Parameters:
-    - response (str): The response string to be parsed
-
-    Returns:
-    - list[dict]: The list of summaries output by the model
-    """
-    parsed = json.loads(response)
-    return parsed['new_monologue']

agenthub/planner_agent/__init__.py

@@ -1,5 +1,4 @@
-from opendevin.controller.agent import Agent
-
-from .agent import PlannerAgent
+from agenthub.planner_agent.agent import PlannerAgent
+from openhands.controller.agent import Agent
 
 Agent.register('PlannerAgent', PlannerAgent)

agenthub/planner_agent/agent.py

@@ -1,9 +1,11 @@
-from opendevin.controller.agent import Agent
-from opendevin.controller.state.state import State
-from opendevin.events.action import Action, AgentFinishAction
-from opendevin.llm.llm import LLM
-
-from .prompt import get_prompt, parse_response
+from agenthub.planner_agent.prompt import get_prompt_and_images
+from agenthub.planner_agent.response_parser import PlannerResponseParser
+from openhands.controller.agent import Agent
+from openhands.controller.state.state import State
+from openhands.core.config import AgentConfig
+from openhands.core.message import ImageContent, Message, TextContent
+from openhands.events.action import Action, AgentFinishAction
+from openhands.llm.llm import LLM
 
 
 class PlannerAgent(Agent):
@@ -12,19 +14,18 @@ class PlannerAgent(Agent):
     The planner agent utilizes a special prompting strategy to create long term plans for solving problems.
     The agent is given its previous action-observation pairs, current task, and hint based on last action taken at every step.
     """
+    response_parser = PlannerResponseParser()
 
-    def __init__(self, llm: LLM):
-        """
-        Initialize the Planner Agent with an LLM
+    def __init__(self, llm: LLM, config: AgentConfig):
+        """Initialize the Planner Agent with an LLM
 
         Parameters:
         - llm (LLM): The llm to be used by this agent
         """
-        super().__init__(llm)
+        super().__init__(llm, config)
 
     def step(self, state: State) -> Action:
-        """
-        Checks to see if current step is completed, returns AgentFinishAction if True.
+        """Checks to see if current step is completed, returns AgentFinishAction if True.
         Otherwise, creates a plan prompt and sends to model for inference, returning the result as the next action.
 
         Parameters:
@@ -34,20 +35,19 @@ class PlannerAgent(Agent):
         - AgentFinishAction: If the last state was 'completed', 'verified', or 'abandoned'
         - Action: The next action to take based on llm response
         """
-
         if state.root_task.state in [
             'completed',
             'verified',
             'abandoned',
         ]:
             return AgentFinishAction()
-        prompt = get_prompt(state)
-        messages = [{'content': prompt, 'role': 'user'}]
-        resp = self.llm.do_completion(messages=messages)
-        action_resp = resp['choices'][0]['message']['content']
-        state.num_of_chars += len(prompt) + len(action_resp)
-        action = parse_response(action_resp)
-        return action
 
-    def search_memory(self, query: str) -> list[str]:
-        return []
+        prompt, image_urls = get_prompt_and_images(
+            state, self.llm.config.max_message_chars
+        )
+        content = [TextContent(text=prompt)]
+        if self.llm.vision_is_active() and image_urls:
+            content.append(ImageContent(image_urls=image_urls))
+        message = Message(role='user', content=content)
+        resp = self.llm.completion(messages=self.llm.format_messages_for_llm(message))
+        return self.response_parser.parse(resp)

agenthub/planner_agent/prompt.py

@@ -1,18 +1,15 @@
-from opendevin.controller.state.state import State
-from opendevin.core.logger import opendevin_logger as logger
-from opendevin.core.schema import ActionType
-from opendevin.core.utils import json
-from opendevin.events.action import (
+from openhands.controller.state.state import State
+from openhands.core.logger import openhands_logger as logger
+from openhands.core.schema import ActionType
+from openhands.core.utils import json
+from openhands.events.action import (
     Action,
     NullAction,
 )
-from opendevin.events.observation import (
-    NullObservation,
-)
-from opendevin.events.serialization.action import action_from_dict
-from opendevin.events.serialization.event import event_to_memory
+from openhands.events.serialization.action import action_from_dict
+from openhands.events.serialization.event import event_to_memory
 
-HISTORY_SIZE = 10
+HISTORY_SIZE = 20
 
 prompt = """
 # Task
@@ -77,9 +74,6 @@ It must be an object, and it must contain two fields:
   * `content` - the content to write to the file
 * `run` - runs a command on the command line in a Linux shell. Arguments:
   * `command` - the command to run
-  * `background` - if true, run the command in the background, so that other commands can be run concurrently. Useful for e.g. starting a server. You won't be able to see the logs. You don't need to end the command with `&`, just set this to true.
-* `kill` - kills a background command
-  * `command_id` - the ID of the background command to kill
 * `browse` - opens a web page. Arguments:
   * `url` - the URL to open
 * `message` - make a plan, set a goal, record your thoughts, or ask for more input from the user. Arguments:
@@ -94,7 +88,7 @@ It must be an object, and it must contain two fields:
   * `state` - set to 'in_progress' to start the task, 'completed' to finish it, 'verified' to assert that it was successful, 'abandoned' to give up on it permanently, or `open` to stop working on it for now.
 * `finish` - if ALL of your tasks and subtasks have been verified or abandoned, and you're absolutely certain that you've completed your task and have tested your work, use the finish action to stop working.
 
-You MUST take time to think in between read, write, run, kill, browse, and recall actions--do this with the `message` action.
+You MUST take time to think in between read, write, run, and browse actions--do this with the `message` action.
 You should never act twice in a row without thinking. But if your last several
 actions are all `message` actions, you should consider taking a different action.
 
@@ -106,7 +100,6 @@ What is your next thought or action? Again, you must reply with JSON, and only w
 
 def get_hint(latest_action_id: str) -> str:
     """Returns action type hint based on given action_id"""
-
     hints = {
         '': "You haven't taken any actions yet. Start by using `ls` to check out what files you're working with.",
         ActionType.RUN: 'You should think about the command you just ran, what output it gave, and how that affects your plan.',
@@ -114,7 +107,6 @@ def get_hint(latest_action_id: str) -> str:
         ActionType.WRITE: 'You just changed a file. You should think about how it affects your plan.',
         ActionType.BROWSE: 'You should think about the page you just visited, and what you learned from it.',
         ActionType.MESSAGE: "Look at your last thought in the history above. What does it suggest? Don't think anymore--take action.",
-        ActionType.RECALL: 'You should think about the information you just recalled, and how it should affect your plan.',
         ActionType.ADD_TASK: 'You should think about the next action to take.',
         ActionType.MODIFY_TASK: 'You should think about the next action to take.',
         ActionType.SUMMARIZE: '',
@@ -123,9 +115,11 @@ def get_hint(latest_action_id: str) -> str:
     return hints.get(latest_action_id, '')
 
 
-def get_prompt(state: State) -> str:
-    """
-    Gets the prompt for the planner agent.
+def get_prompt_and_images(
+    state: State, max_message_chars: int
+) -> tuple[str, list[str]]:
+    """Gets the prompt for the planner agent.
+
     Formatted with the most recent action-observation pairs, current task, and hint based on last action
 
     Parameters:
@@ -134,19 +128,28 @@ def get_prompt(state: State) -> str:
     Returns:
     - str: The formatted string prompt with historical values
     """
-
+    # the plan
     plan_str = json.dumps(state.root_task.to_dict(), indent=2)
-    sub_history = state.history[-HISTORY_SIZE:]
+
+    # the history
     history_dicts = []
     latest_action: Action = NullAction()
-    for action, observation in sub_history:
-        if not isinstance(action, NullAction):
-            history_dicts.append(event_to_memory(action))
-            latest_action = action
-        if not isinstance(observation, NullObservation):
-            observation_dict = event_to_memory(observation)
-            history_dicts.append(observation_dict)
+
+    # retrieve the latest HISTORY_SIZE events
+    for event_count, event in enumerate(state.history.get_events(reverse=True)):
+        if event_count >= HISTORY_SIZE:
+            break
+        if latest_action == NullAction() and isinstance(event, Action):
+            latest_action = event
+        history_dicts.append(event_to_memory(event, max_message_chars))
+
+    # history_dicts is in reverse order, lets fix it
+    history_dicts.reverse()
+
+    # and get it as a JSON string
     history_str = json.dumps(history_dicts, indent=2)
+
+    # the plan status
     current_task = state.root_task.get_current_task()
     if current_task is not None:
         plan_status = f"You're currently working on this task:\n{current_task.goal}."
@@ -154,22 +157,26 @@ def get_prompt(state: State) -> str:
             plan_status += "\nIf it's not achievable AND verifiable with a SINGLE action, you MUST break it down into subtasks NOW."
     else:
         plan_status = "You're not currently working on any tasks. Your next action MUST be to mark a task as in_progress."
-    hint = get_hint(event_to_memory(latest_action).get('action', ''))
+
+    # the hint, based on the last action
+    hint = get_hint(event_to_memory(latest_action, max_message_chars).get('action', ''))
     logger.info('HINT:\n' + hint, extra={'msg_type': 'DETAIL'})
-    task = state.get_current_user_intent()
+
+    # the last relevant user message (the task)
+    message, image_urls = state.get_current_user_intent()
+
+    # finally, fill in the prompt
     return prompt % {
-        'task': task,
+        'task': message,
         'plan': plan_str,
         'history': history_str,
         'hint': hint,
         'plan_status': plan_status,
-    }
+    }, image_urls
 
 
 def parse_response(response: str) -> Action:
-    """
-    Parses the model output to find a valid action to take
-
+    """Parses the model output to find a valid action to take
     Parameters:
     - response (str): A response from the model that potentially contains an Action.
 

agenthub/planner_agent/response_parser.py

@@ -0,0 +1,37 @@
+from openhands.controller.action_parser import ResponseParser
+from openhands.core.utils import json
+from openhands.events.action import (
+    Action,
+)
+from openhands.events.serialization.action import action_from_dict
+
+
+class PlannerResponseParser(ResponseParser):
+    def __init__(self):
+        super().__init__()
+
+    def parse(self, response: str) -> Action:
+        action_str = self.parse_response(response)
+        return self.parse_action(action_str)
+
+    def parse_response(self, response) -> str:
+        # get the next action from the response
+        return response['choices'][0]['message']['content']
+
+    def parse_action(self, action_str: str) -> Action:
+        """Parses a string to find an action within it
+
+        Parameters:
+        - response (str): The string to be parsed
+
+        Returns:
+        - Action: The action that was found in the response string
+        """
+        # attempt to load the JSON dict from the response
+        action_dict = json.loads(action_str)
+
+        if 'content' in action_dict:
+            # The LLM gets confused here. Might as well be robust
+            action_dict['contents'] = action_dict.pop('content')
+
+        return action_from_dict(action_dict)

compose.yml

@@ -0,0 +1,22 @@
+#
+services:
+  openhands:
+    build:
+      context: ./
+      dockerfile: ./containers/app/Dockerfile
+    image: openhands:latest
+    container_name: openhands-app-${DATE:-}
+    environment:
+      - SANDBOX_RUNTIME_CONTAINER_IMAGE=${SANDBOX_RUNTIME_CONTAINER_IMAGE:-ghcr.io/all-hands-ai/runtime:0.9-nikolaik}
+      - SANDBOX_USER_ID=${SANDBOX_USER_ID:-1234}
+      - WORKSPACE_MOUNT_PATH=${WORKSPACE_BASE:-$PWD/workspace}
+    ports:
+      - "3000:3000"
+    extra_hosts:
+      - "host.docker.internal:host-gateway"
+    volumes:
+      - /var/run/docker.sock:/var/run/docker.sock
+      - ${WORKSPACE_BASE:-$PWD/workspace}:/opt/workspace_base
+    pull_policy: build
+    stdin_open: true
+    tty: true

config.template.toml

@@ -0,0 +1,243 @@
+###################### OpenHands Configuration Example ######################
+#
+# All settings have default values, so you only need to uncomment and
+# modify what you want to change
+# The fields within each section are sorted in alphabetical order.
+#
+##############################################################################
+
+#################################### Core ####################################
+# General core configurations
+##############################################################################
+[core]
+# API key for E2B
+#e2b_api_key = ""
+
+# Base path for the workspace
+workspace_base = "./workspace"
+
+# Cache directory path
+#cache_dir = "/tmp/cache"
+
+# Debugging enabled
+#debug = false
+
+# Disable color in terminal output
+#disable_color = false
+
+# Enable saving and restoring the session when run from CLI
+#enable_cli_session = false
+
+# File store path
+#file_store_path = "/tmp/file_store"
+
+# File store type
+#file_store = "memory"
+
+# List of allowed file extensions for uploads
+#file_uploads_allowed_extensions = [".*"]
+
+# Maximum file size for uploads, in megabytes
+#file_uploads_max_file_size_mb = 0
+
+# Maximum budget per task, 0.0 means no limit
+#max_budget_per_task = 0.0
+
+# Maximum number of iterations
+#max_iterations = 100
+
+# Path to mount the workspace in the sandbox
+#workspace_mount_path_in_sandbox = "/workspace"
+
+# Path to mount the workspace
+#workspace_mount_path = ""
+
+# Path to rewrite the workspace mount path to
+#workspace_mount_rewrite = ""
+
+# Run as openhands
+#run_as_openhands = true
+
+# Runtime environment
+#runtime = "eventstream"
+
+# Name of the default agent
+#default_agent = "CodeActAgent"
+
+# JWT secret for authentication
+#jwt_secret = ""
+
+# Restrict file types for file uploads
+#file_uploads_restrict_file_types = false
+
+# List of allowed file extensions for uploads
+#file_uploads_allowed_extensions = [".*"]
+
+#################################### LLM #####################################
+# Configuration for LLM models (group name starts with 'llm')
+# use 'llm' for the default LLM config
+##############################################################################
+[llm]
+# AWS access key ID
+#aws_access_key_id = ""
+
+# AWS region name
+#aws_region_name = ""
+
+# AWS secret access key
+#aws_secret_access_key = ""
+
+# API key to use
+api_key = "your-api-key"
+
+# API base URL
+#base_url = ""
+
+# API version
+#api_version = ""
+
+# Cost per input token
+#input_cost_per_token = 0.0
+
+# Cost per output token
+#output_cost_per_token = 0.0
+
+# Custom LLM provider
+#custom_llm_provider = ""
+
+# Embedding API base URL
+#embedding_base_url = ""
+
+# Embedding deployment name
+#embedding_deployment_name = ""
+
+# Embedding model to use
+embedding_model = ""
+
+# Maximum number of characters in an observation's content
+#max_message_chars = 10000
+
+# Maximum number of input tokens
+#max_input_tokens = 0
+
+# Maximum number of output tokens
+#max_output_tokens = 0
+
+# Model to use
+model = "gpt-4o"
+
+# Number of retries to attempt when an operation fails with the LLM.
+# Increase this value to allow more attempts before giving up
+#num_retries = 8
+
+# Maximum wait time (in seconds) between retry attempts
+# This caps the exponential backoff to prevent excessively long
+#retry_max_wait = 120
+
+# Minimum wait time (in seconds) between retry attempts
+# This sets the initial delay before the first retry
+#retry_min_wait = 15
+
+# Multiplier for exponential backoff calculation
+# The wait time increases by this factor after each failed attempt
+# A value of 2.0 means each retry waits twice as long as the previous one
+#retry_multiplier = 2.0
+
+# Drop any unmapped (unsupported) params without causing an exception
+#drop_params = false
+
+# Using the prompt caching feature provided by the LLM
+#caching_prompt = false
+
+# Base URL for the OLLAMA API
+#ollama_base_url = ""
+
+# Temperature for the API
+#temperature = 0.0
+
+# Timeout for the API
+#timeout = 0
+
+# Top p for the API
+#top_p = 1.0
+
+# If model is vision capable, this option allows to disable image processing (useful for cost reduction).
+#disable_vision = true
+
+[llm.gpt4o-mini]
+# API key to use
+api_key = "your-api-key"
+
+# Model to use
+model = "gpt-4o-mini"
+
+#################################### Agent ###################################
+# Configuration for agents (group name starts with 'agent')
+# Use 'agent' for the default agent config
+# otherwise, group name must be `agent.<agent_name>` (case-sensitive), e.g.
+# agent.CodeActAgent
+##############################################################################
+[agent]
+# Name of the micro agent to use for this agent
+#micro_agent_name = ""
+
+# Memory enabled
+#memory_enabled = false
+
+# Memory maximum threads
+#memory_max_threads = 2
+
+# LLM config group to use
+#llm_config = 'llm'
+
+[agent.RepoExplorerAgent]
+# Example: use a cheaper model for RepoExplorerAgent to reduce cost, especially
+# useful when an agent doesn't demand high quality but uses a lot of tokens
+llm_config = 'gpt3'
+
+#################################### Sandbox ###################################
+# Configuration for the sandbox
+##############################################################################
+[sandbox]
+# Sandbox timeout in seconds
+#timeout = 120
+
+# Sandbox user ID
+#user_id = 1000
+
+# Container image to use for the sandbox
+#base_container_image = "nikolaik/python-nodejs:python3.11-nodejs22"
+
+# Use host network
+#use_host_network = false
+
+# Enable auto linting after editing
+#enable_auto_lint = false
+
+# Whether to initialize plugins
+#initialize_plugins = true
+
+# Extra dependencies to install in the runtime image
+#runtime_extra_deps = ""
+
+# Environment variables to set at the launch of the runtime
+#runtime_startup_env_vars = {}
+
+# BrowserGym environment to use for evaluation
+#browsergym_eval_env = ""
+
+#################################### Security ###################################
+# Configuration for security features
+##############################################################################
+[security]
+
+# Enable confirmation mode
+#confirmation_mode = true
+
+# The security analyzer to use
+#security_analyzer = ""
+
+#################################### Eval ####################################
+# Configuration for the evaluation, please refer to the specific evaluation
+# plugin for the available options
+##############################################################################

containers/README.md

@@ -1,11 +1,12 @@
 # Docker Containers
+
 Each folder here contains a Dockerfile, and a config.sh describing how to build
-the image and where to push it. These are images are built and pushed in GitHub Actions
+the images and where to push them. These images are built and pushed in GitHub Actions
 by the `ghcr.yml` workflow.
 
 ## Building Manually
 
-```
-docker build -f containers/app/Dockerfile -t opendevin .
+```bash
+docker build -f containers/app/Dockerfile -t openhands .
 docker build -f containers/sandbox/Dockerfile -t sandbox .
 ```

containers/app/Dockerfile

@@ -1,5 +1,5 @@
-ARG OPEN_DEVIN_BUILD_VERSION=dev
-FROM node:21.7.2-bookworm-slim as frontend-builder
+ARG OPENHANDS_BUILD_VERSION=dev
+FROM node:21.7.2-bookworm-slim AS frontend-builder
 
 WORKDIR /app
 
@@ -10,10 +10,10 @@ RUN npm ci
 COPY ./frontend ./
 RUN npm run make-i18n && npm run build
 
-FROM python:3.12-slim as backend-builder
+FROM python:3.12.3-slim AS backend-builder
 
 WORKDIR /app
-ENV PYTHONPATH '/app'
+ENV PYTHONPATH='/app'
 
 ENV POETRY_NO_INTERACTION=1 \
     POETRY_VIRTUALENVS_IN_PROJECT=1 \
@@ -26,57 +26,67 @@ RUN apt-get update -y \
 
 COPY ./pyproject.toml ./poetry.lock ./
 RUN touch README.md
-RUN poetry install --without evaluation --no-root && rm -rf $POETRY_CACHE_DIR
+RUN export POETRY_CACHE_DIR && poetry install --without evaluation,llama-index --no-root && rm -rf $POETRY_CACHE_DIR
 
-FROM python:3.12-slim as runtime
+FROM python:3.12.3-slim AS runtime
 
 WORKDIR /app
 
-ENV RUN_AS_DEVIN=true
+ARG OPENHANDS_BUILD_VERSION #re-declare for this section
+
+ENV RUN_AS_OPENHANDS=true
 # A random number--we need this to be different from the user's UID on the host machine
-ENV OPENDEVIN_USER_ID=42420
+ENV OPENHANDS_USER_ID=42420
+ENV SANDBOX_API_HOSTNAME=host.docker.internal
 ENV USE_HOST_NETWORK=false
-ENV SSH_HOSTNAME=host.docker.internal
 ENV WORKSPACE_BASE=/opt/workspace_base
-ENV OPEN_DEVIN_BUILD_VERSION=$OPEN_DEVIN_BUILD_VERSION
+ENV OPENHANDS_BUILD_VERSION=$OPENHANDS_BUILD_VERSION
 RUN mkdir -p $WORKSPACE_BASE
 
 RUN apt-get update -y \
     && apt-get install -y curl ssh sudo
 
-RUN sed -i 's/^UID_MIN.*/UID_MIN 499/' /etc/login.defs # Default is 1000, but OSX is often 501
-RUN sed -i 's/^UID_MAX.*/UID_MAX 1000000/' /etc/login.defs # Default is 60000, but we've seen up to 200000
+# Default is 1000, but OSX is often 501
+RUN sed -i 's/^UID_MIN.*/UID_MIN 499/' /etc/login.defs
+# Default is 60000, but we've seen up to 200000
+RUN sed -i 's/^UID_MAX.*/UID_MAX 1000000/' /etc/login.defs
 
 RUN groupadd app
-RUN useradd -l -m -u $OPENDEVIN_USER_ID -s /bin/bash opendevin && \
-    usermod -aG app opendevin && \
-    usermod -aG sudo opendevin && \
+RUN useradd -l -m -u $OPENHANDS_USER_ID -s /bin/bash openhands && \
+    usermod -aG app openhands && \
+    usermod -aG sudo openhands && \
     echo '%sudo ALL=(ALL) NOPASSWD:ALL' >> /etc/sudoers
-RUN chown -R opendevin:app /app && chmod -R 770 /app
-RUN sudo chown -R opendevin:app $WORKSPACE_BASE && sudo chmod -R 770 $WORKSPACE_BASE
-USER opendevin
+RUN chown -R openhands:app /app && chmod -R 770 /app
+RUN sudo chown -R openhands:app $WORKSPACE_BASE && sudo chmod -R 770 $WORKSPACE_BASE
+USER openhands
 
 ENV VIRTUAL_ENV=/app/.venv \
     PATH="/app/.venv/bin:$PATH" \
     PYTHONPATH='/app'
 
-COPY --chown=opendevin:app --chmod=770 --from=backend-builder ${VIRTUAL_ENV} ${VIRTUAL_ENV}
+COPY --chown=openhands:app --chmod=770 --from=backend-builder ${VIRTUAL_ENV} ${VIRTUAL_ENV}
 RUN playwright install --with-deps chromium
 
-COPY --chown=opendevin:app --chmod=770 ./opendevin ./opendevin
-COPY --chown=opendevin:app --chmod=777 ./opendevin/runtime/plugins ./opendevin/runtime/plugins
-COPY --chown=opendevin:app --chmod=770 ./agenthub ./agenthub
+COPY --chown=openhands:app --chmod=770 ./openhands ./openhands
+COPY --chown=openhands:app --chmod=777 ./openhands/runtime/plugins ./openhands/runtime/plugins
+COPY --chown=openhands:app --chmod=770 ./agenthub ./agenthub
+COPY --chown=openhands:app --chmod=770 ./pyproject.toml ./pyproject.toml
+COPY --chown=openhands:app --chmod=770 ./poetry.lock ./poetry.lock
+COPY --chown=openhands:app --chmod=770 ./README.md ./README.md
+COPY --chown=openhands:app --chmod=770 ./MANIFEST.in ./MANIFEST.in
 
-RUN python opendevin/core/download.py # No-op to download assets
-RUN chown -R opendevin:app /app/logs && chmod -R 770 /app/logs # This gets created by the download.py script
+# This is run as "openhands" user, and will create __pycache__ with openhands:openhands ownership
+RUN python openhands/core/download.py # No-op to download assets
+# Add this line to set group ownership of all files/directories not already in "app" group
+# openhands:openhands -> openhands:app
+RUN find /app \! -group app -exec chgrp app {} +
 
-
-COPY --chown=opendevin:app --chmod=770 --from=frontend-builder /app/dist ./frontend/dist
-COPY --chown=opendevin:app --chmod=770 ./containers/app/entrypoint.sh /app/entrypoint.sh
+COPY --chown=openhands:app --chmod=770 --from=frontend-builder /app/dist ./frontend/dist
+COPY --chown=openhands:app --chmod=770 ./containers/app/entrypoint.sh /app/entrypoint.sh
 
 USER root
 
 WORKDIR /app
 
 ENTRYPOINT ["/app/entrypoint.sh"]
-CMD ["uvicorn", "opendevin.server.listen:app", "--host", "0.0.0.0", "--port", "3000"]
+CMD ["uvicorn", "openhands.server.listen:app", "--host", "0.0.0.0", "--port", "3000"]

containers/app/config.sh

@@ -1,4 +1,4 @@
 DOCKER_REGISTRY=ghcr.io
-DOCKER_ORG=opendevin
-DOCKER_IMAGE=opendevin
+DOCKER_ORG=all-hands-ai
+DOCKER_IMAGE=openhands
 DOCKER_BASE_DIR="."

containers/app/entrypoint.sh

@@ -1,7 +1,7 @@
 #!/bin/bash
 set -eo pipefail
 
-echo "Starting OpenDevin..."
+echo "Starting OpenHands..."
 if [[ $NO_SETUP == "true" ]]; then
   echo "Skipping setup, running as $(whoami)"
   "$@"
@@ -9,7 +9,7 @@ if [[ $NO_SETUP == "true" ]]; then
 fi
 
 if [ "$(id -u)" -ne 0 ]; then
-  echo "The OpenDevin entrypoint.sh must run as root"
+  echo "The OpenHands entrypoint.sh must run as root"
   exit 1
 fi
 
@@ -19,10 +19,12 @@ if [ -z "$SANDBOX_USER_ID" ]; then
 fi
 
 if [[ "$SANDBOX_USER_ID" -eq 0 ]]; then
-  echo "Running OpenDevin as root"
-  export RUN_AS_DEVIN=false
+  echo "Running OpenHands as root"
+  export RUN_AS_OPENHANDS=false
   mkdir -p /root/.cache/ms-playwright/
-  mv /home/opendevin/.cache/ms-playwright/ /root/.cache/
+  if [ -d "/home/openhands/.cache/ms-playwright/" ]; then
+    mv /home/openhands/.cache/ms-playwright/ /root/.cache/
+  fi
   "$@"
 else
   echo "Setting up enduser with id $SANDBOX_USER_ID"
@@ -30,9 +32,9 @@ else
     echo "User enduser already exists. Skipping creation."
   else
     if ! useradd -l -m -u $SANDBOX_USER_ID -s /bin/bash enduser; then
-      echo "Failed to create user enduser with id $SANDBOX_USER_ID. Moving opendevin user."
+      echo "Failed to create user enduser with id $SANDBOX_USER_ID. Moving openhands user."
       incremented_id=$(($SANDBOX_USER_ID + 1))
-      usermod -u $incremented_id opendevin
+      usermod -u $incremented_id openhands
       if ! useradd -l -m -u $SANDBOX_USER_ID -s /bin/bash enduser; then
         echo "Failed to create user enduser with id $SANDBOX_USER_ID for a second time. Exiting."
         exit 1
@@ -40,7 +42,7 @@ else
     fi
   fi
   usermod -aG app enduser
-  # get the user group of /var/run/docker.sock and set opendevin to that group
+  # get the user group of /var/run/docker.sock and set openhands to that group
   DOCKER_SOCKET_GID=$(stat -c '%g' /var/run/docker.sock)
   echo "Docker socket group id: $DOCKER_SOCKET_GID"
   if getent group $DOCKER_SOCKET_GID; then
@@ -50,10 +52,13 @@ else
     groupadd -g $DOCKER_SOCKET_GID docker
   fi
 
+  mkdir -p /home/enduser/.cache/huggingface/hub/
   mkdir -p /home/enduser/.cache/ms-playwright/
-  mv /home/opendevin/.cache/ms-playwright/ /home/enduser/.cache/
+  if [ -d "/home/openhands/.cache/ms-playwright/" ]; then
+    mv /home/openhands/.cache/ms-playwright/ /home/enduser/.cache/
+  fi
 
   usermod -aG $DOCKER_SOCKET_GID enduser
   echo "Running as enduser"
-  su enduser /bin/bash -c "$*"
+  su enduser /bin/bash -c "${*@Q}" # This magically runs any arguments passed to the script as a command
 fi