Add RuntimeManager to centralize runtime management

.devcontainer/README.MD

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

.devcontainer/devcontainer.json

@@ -1,15 +0,0 @@
-{
-	"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

@@ -1,6 +0,0 @@
-#!/usr/bin/env bash
-sudo apt update
-sudo apt install -y netcat
-sudo add-apt-repository -y ppa:deadsnakes/ppa
-sudo apt install -y python3.12
-curl -sSL https://install.python-poetry.org | python3.12 -

.dockerignore

@@ -1,5 +0,0 @@
-frontend/node_modules
-config.toml
-.envrc
-.env
-.git

.gitattributes

@@ -1 +0,0 @@
-*.ipynb linguist-vendored

.github/.codecov.yml

@@ -1,19 +0,0 @@
-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:
-    patch:
-      default:
-        threshold: 100% # allow patch coverage to be lower than project coverage by any amount
-    project:
-      default:
-        threshold: 5% # allow project coverage to drop at most 5%
-
-comment: false
-github_checks:
-    annotations: false

.github/ISSUE_TEMPLATE/bug_template.yml

@@ -1,59 +0,0 @@
-name: Bug
-description: Report a problem with OpenHands
-title: '[Bug]: '
-labels: ['bug']
-body:
-  - type: markdown
-    attributes:
-      value: Thank you for taking the time to fill out this bug report. Please provide as much information as possible to help us understand and address the issue effectively.
-
-  - type: checkboxes
-    attributes:
-      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 existing issues.
-        required: true
-
-  - type: textarea
-    id: bug-description
-    attributes:
-      label: Describe the bug and reproduction steps
-      description: Provide a description of the issue along with any reproduction steps.
-    validations:
-      required: true
-
-  - type: dropdown
-    id: installation
-    attributes:
-      label: OpenHands Installation
-      description: How are you running OpenHands?
-      options:
-        - Docker command in README
-        - Development workflow
-      default: 0
-
-  - type: input
-    id: openhands-version
-    attributes:
-      label: OpenHands Version
-      description: What version of OpenHands are you using?
-      placeholder: ex. 0.9.8, main, etc.
-
-  - type: dropdown
-    id: os
-    attributes:
-      label: Operating System
-      options:
-        - MacOS
-        - Linux
-        - WSL on Windows
-
-  - type: textarea
-    id: additional-context
-    attributes:
-      label: Logs, Errors, Screenshots, and Additional Context
-      description: Please provide any additional information you think might help. 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

@@ -1,18 +0,0 @@
----
-name: Feature Request
-about: Suggest an idea for OpenHands features
-title: ''
-labels: 'enhancement'
-assignees: ''
-
----
-
-**What problem or use case are you trying to solve?**
-
-**Describe the UX of the solution you'd like**
-
-**Do you have thoughts on the technical implementation?**
-
-**Describe alternatives you've considered**
-
-**Additional context**

.github/ISSUE_TEMPLATE/technical_proposal.md

@@ -1,18 +0,0 @@
----
-name: Technical Proposal
-about: Propose a new architecture or technology
-title: ''
-labels: 'proposal'
-assignees: ''
-
----
-
-**Summary**
-
-**Motivation**
-
-**Technical Design**
-
-**Alternatives to Consider**
-
-**Additional context**

.github/dependabot.yml

@@ -1,38 +0,0 @@
-# To get started with Dependabot version updates, you'll need to specify which
-# package ecosystems to update and where the package manifests are located.
-# Please see the documentation for all configuration options:
-# https://docs.github.com/code-security/dependabot/dependabot-version-updates/configuration-options-for-the-dependabot.yml-file
-
-version: 2
-updates:
-  - package-ecosystem: "pip"
-    directory: "/"
-    schedule:
-      interval: "daily"
-    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

@@ -1,11 +0,0 @@
-**End-user friendly description of the problem this fixes or functionality that this introduces**
-
-- [ ] Include this change in the Release Notes. If checked, you must provide an **end-user friendly** description for your change below
-
----
-**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

@@ -1,69 +0,0 @@
-# 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

@@ -1,74 +0,0 @@
-# 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
-
-# If triggered by a PR, it will be in the same group. However, each commit on main will be in its own unique group
-concurrency:
-  group: ${{ github.workflow }}-${{ (github.head_ref && github.ref) || github.run_id }}
-  cancel-in-progress: true
-
-jobs:
-  # Build the documentation website
-  build:
-    if: github.repository == 'All-Hands-AI/OpenHands'
-    name: Build Docusaurus
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-      - uses: actions/setup-node@v4
-        with:
-          node-version: 18
-          cache: npm
-          cache-dependency-path: docs/package-lock.json
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: '3.12'
-      - 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
-    # Grant GITHUB_TOKEN the permissions required to make a Pages deployment
-    permissions:
-      pages: write # to deploy to Pages
-      id-token: write # to verify the deployment originates from an appropriate source
-    # Deploy to the github-pages environment
-    environment:
-      name: github-pages
-      url: ${{ steps.deployment.outputs.page_url }}
-    steps:
-      - name: Deploy to GitHub Pages
-        id: deployment
-        uses: actions/deploy-pages@v4

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

@@ -1,61 +0,0 @@
-# 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:
-    - main
-  pull_request:
-
-# If triggered by a PR, it will be in the same group. However, each commit on main will be in its own unique group
-concurrency:
-  group: ${{ github.workflow }}-${{ (github.head_ref && github.ref) || github.run_id }}
-  cancel-in-progress: true
-
-jobs:
-  test:
-    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: '3.12'
-          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: |
-          set -e
-          SANDBOX_FORCE_REBUILD_RUNTIME=True 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

@@ -1,44 +0,0 @@
-# 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'
-
-# If triggered by a PR, it will be in the same group. However, each commit on main will be in its own unique group
-concurrency:
-  group: ${{ github.workflow }}-${{ (github.head_ref && github.ref) || github.run_id }}
-  cancel-in-progress: true
-
-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-build.yml

@@ -1,409 +0,0 @@
-# Workflow that builds, tests and then pushes the OpenHands and runtime docker images to the ghcr.io repository
-name: Build, Test and Publish RT Image
-
-# 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: ''
-
-# If triggered by a PR, it will be in the same group. However, each commit on main will be in its own unique group
-concurrency:
-  group: ${{ github.workflow }}-${{ (github.head_ref && github.ref) || github.run_id }}
-  cancel-in-progress: true
-
-env:
-  BASE_IMAGE_FOR_HASH_EQUIVALENCE_TEST: nikolaik/python-nodejs:python3.12-nodejs22
-  RELEVANT_SHA: ${{ github.event.pull_request.head.sha || github.sha }}
-
-jobs:
-  # Builds the OpenHands Docker images
-  ghcr_build_app:
-    name: Build App Image
-    runs-on: ubuntu-latest
-    permissions:
-      contents: read
-      packages: write
-    outputs:
-      hash_from_app_image: ${{ steps.get_hash_in_app_image.outputs.hash_from_app_image }}
-    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@v3.0.0
-        with:
-          image: tonistiigi/binfmt:latest
-      - 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: Build and push app image
-        if: "!github.event.pull_request.head.repo.fork"
-        run: |
-          ./containers/build.sh -i openhands -o ${{ github.repository_owner }} --push
-      - name: Build app image
-        if: "github.event.pull_request.head.repo.fork"
-        run: |
-          ./containers/build.sh -i openhands -o ${{ github.repository_owner }} --load
-      - name: Get hash in App Image
-        id: get_hash_in_app_image
-        run: |
-          # Lowercase the repository owner
-          export REPO_OWNER=${{ github.repository_owner }}
-          REPO_OWNER=$(echo $REPO_OWNER | tr '[:upper:]' '[:lower:]')
-          # Run the build script in the app image
-          docker run -e SANDBOX_USER_ID=0 -v /var/run/docker.sock:/var/run/docker.sock ghcr.io/${REPO_OWNER}/openhands:${{ env.RELEVANT_SHA }} /bin/bash -c "mkdir -p containers/runtime; python3 openhands/runtime/utils/runtime_build.py --base_image ${{ env.BASE_IMAGE_FOR_HASH_EQUIVALENCE_TEST }} --build_folder containers/runtime --force_rebuild" 2>&1 | tee docker-outputs.txt
-          # Get the hash from the build script
-          hash_from_app_image=$(cat docker-outputs.txt | grep "Hash for docker build directory" | awk -F "): " '{print $2}' | uniq | head -n1)
-          echo "hash_from_app_image=$hash_from_app_image" >> $GITHUB_OUTPUT
-          echo "Hash from app image: $hash_from_app_image"
-      # This test should move when we have a test suite for the app image
-      - name: Test docker in App Image
-        run: |
-          # Lowercase the repository owner
-          export REPO_OWNER=${{ github.repository_owner }}
-          REPO_OWNER=$(echo $REPO_OWNER | tr '[:upper:]' '[:lower:]')
-
-          docker run -e SANDBOX_USER_ID=0 -v /var/run/docker.sock:/var/run/docker.sock ghcr.io/${REPO_OWNER}/openhands:${{ env.RELEVANT_SHA }} /bin/bash -c "docker run hello-world"
-
-  # 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:python3.12-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@v3.0.0
-        with:
-          image: tonistiigi/binfmt:latest
-      - 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: '3.12'
-      - 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 -i runtime -o ${{ github.repository_owner }} --push -t ${{ 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 ${{ matrix.base_image.image }} for fork
-        if: github.event.pull_request.head.repo.fork
-        uses: docker/build-push-action@v6
-        with:
-          tags: ghcr.io/all-hands-ai/runtime:${{ env.RELEVANT_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
-
-  verify_hash_equivalence_in_runtime_and_app:
-    name: Verify Hash Equivalence in Runtime and Docker images
-    runs-on: ubuntu-latest
-    needs: [ghcr_build_runtime, ghcr_build_app]
-    strategy:
-      fail-fast: false
-      matrix:
-        base_image: ['nikolaik']
-    steps:
-      - uses: actions/checkout@v4
-      - 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: '3.12'
-      - name: Install poetry via pipx
-        run: pipx install poetry
-      - name: Install Python dependencies using Poetry
-        run: make install-python-dependencies
-      - name: Get hash in App Image
-        run: |
-          echo "Hash from app image: ${{ needs.ghcr_build_app.outputs.hash_from_app_image }}"
-          echo "hash_from_app_image=${{ needs.ghcr_build_app.outputs.hash_from_app_image }}" >> $GITHUB_ENV
-
-      - name: Get hash using code (development mode)
-        run: |
-          mkdir -p containers/runtime
-          poetry run python3 openhands/runtime/utils/runtime_build.py --base_image ${{ env.BASE_IMAGE_FOR_HASH_EQUIVALENCE_TEST }} --build_folder containers/runtime --force_rebuild > output.txt 2>&1
-          hash_from_code=$(cat output.txt | grep "Hash for docker build directory" | awk -F "): " '{print $2}' | uniq | head -n1)
-          echo "hash_from_code=$hash_from_code" >> $GITHUB_ENV
-
-      - name: Compare hashes
-        run: |
-          echo "Hash from App Image: ${{ env.hash_from_app_image }}"
-          echo "Hash from Code: ${{ env.hash_from_code }}"
-          if [ "${{ env.hash_from_app_image }}" = "${{ env.hash_from_code }}" ]; then
-            echo "Hashes match!"
-          else
-            echo "Hashes do not match!"
-            exit 1
-          fi
-
-  # Run unit tests with the EventStream runtime Docker images as root
-  test_runtime_root:
-    name: RT Unit Tests (Root)
-    needs: [ghcr_build_runtime]
-    runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
-      matrix:
-        base_image: ['nikolaik']
-    steps:
-      - uses: actions/checkout@v4
-      - name: Free Disk Space (Ubuntu)
-        uses: jlumbroso/free-disk-space@main
-        with:
-          # this might remove tools that are actually needed,
-          # if set to "true" but frees about 6 GB
-          tool-cache: true
-          # all of these default to true, but feel free to set to
-          # "false" if necessary for your workflow
-          android: true
-          dotnet: true
-          haskell: true
-          large-packages: true
-          docker-images: false
-          swap-storage: true
-      - name: Set up Docker Buildx
-        id: buildx
-        uses: docker/setup-buildx-action@v3
-      # Forked repos can't push to GHCR, so we need to download the image as an artifact
-      - name: Download runtime image for fork
-        if: github.event.pull_request.head.repo.fork
-        uses: actions/download-artifact@v4
-        with:
-          name: runtime-${{ 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: |
-            ${{ runner.os }}-poetry-
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: '3.12'
-      - name: Install poetry via pipx
-        run: pipx install poetry
-      - name: Install Python dependencies using Poetry
-        run: make install-python-dependencies
-      - name: Run runtime tests
-        run: |
-          # We install pytest-xdist in order to run tests across CPUs
-          poetry run pip install pytest-xdist
-
-          # Install to be able to retry on failures for flaky tests
-          poetry run pip install pytest-rerunfailures
-
-          image_name=ghcr.io/${{ github.repository_owner }}/runtime:${{ env.RELEVANT_SHA }}-${{ 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 -raRs --reruns 2 --reruns-delay 5 --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: |
-            ${{ runner.os }}-poetry-
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: '3.12'
-      - name: Install poetry via pipx
-        run: pipx install poetry
-      - name: Install Python dependencies using Poetry
-        run: make install-python-dependencies
-      - name: Run runtime tests
-        run: |
-          # We install pytest-xdist in order to run tests across CPUs
-          poetry run pip install pytest-xdist
-
-          # Install to be able to retry on failures for flaky tests
-          poetry run pip install pytest-rerunfailures
-
-          image_name=ghcr.io/${{ github.repository_owner }}/runtime:${{ env.RELEVANT_SHA }}-${{ 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 -raRs --reruns 2 --reruns-delay 5 --cov=openhands --cov-report=xml -s ./tests/runtime
-      - 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, verify_hash_equivalence_in_runtime_and_app]
-    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, verify_hash_equivalence_in_runtime_and_app]
-    steps:
-      - name: Some tests failed
-        run: |
-          echo "Some runtime tests failed or were cancelled"
-          exit 1

.github/workflows/lint.yml

@@ -1,54 +0,0 @@
-# Workflow that runs lint on the frontend and python code
-name: Lint
-
-# 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:
-
-# If triggered by a PR, it will be in the same group. However, each commit on main will be in its own unique group
-concurrency:
-  group: ${{ github.workflow }}-${{ (github.head_ref && github.ref) || github.run_id }}
-  cancel-in-progress: true
-
-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
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-      - name: Set up python
-        uses: actions/setup-python@v5
-        with:
-          python-version: 3.12
-          cache: 'pip'
-      - name: Install pre-commit
-        run: pip install pre-commit==3.7.0
-      - name: Run pre-commit hooks
-        run: pre-commit run --files openhands/**/* evaluation/**/* tests/**/* --show-diff-on-failure --config ./dev_config/python/.pre-commit-config.yaml

.github/workflows/openhands-resolver.yml

@@ -1,13 +0,0 @@
-name: Resolve Issues with OpenHands
-
-on:
-  issues:
-    types: [labeled]
-
-jobs:
-  call-openhands-resolver:
-    uses: All-Hands-AI/openhands-resolver/.github/workflows/openhands-resolver.yml@main
-    if: github.event.label.name == 'fix-me'
-    with:
-      issue_number: ${{ github.event.issue.number }}
-    secrets: inherit

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

@@ -1,137 +0,0 @@
-# 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:
-
-# If triggered by a PR, it will be in the same group. However, each commit on main will be in its own unique group
-concurrency:
-  group: ${{ github.workflow }}-${{ (github.head_ref && github.ref) || github.run_id }}
-  cancel-in-progress: true
-
-jobs:
-  # Run python unit tests on macOS
-  test-on-macos:
-    name: Python Unit Tests on macOS
-    runs-on: macos-14
-    env:
-      INSTALL_DOCKER: '1' # Set to '0' to skip Docker installation
-    strategy:
-      matrix:
-        python-version: ['3.12']
-    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=openhands --cov-report=xml ./tests/unit --ignore=tests/unit/test_memory.py
-      - 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: ['3.12']
-    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=openhands --cov-report=xml -svv ./tests/unit --ignore=tests/unit/test_memory.py
-      - name: Upload coverage to Codecov
-        uses: codecov/codecov-action@v4
-        env:
-          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}

.github/workflows/pypi-release.yml

@@ -1,31 +0,0 @@
-# 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: 3.12
-      - 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: ./build.sh
-      - name: publish
-        run: poetry publish -u __token__ -p ${{ secrets.PYPI_TOKEN }}

.github/workflows/review-pr.yml

@@ -1,81 +0,0 @@
-# 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:
-    types: [synchronize, labeled]
-
-permissions:
-  contents: write
-  pull-requests: write
-
-jobs:
-  dogfood:
-    if: contains(github.event.pull_request.labels.*.name, 'review-this')
-    runs-on: ubuntu-latest
-    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: '3.12'
-    - name: install git, github cli
-      run: |
-        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." > 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 "${{ github.event.pull_request.title }}" >> task.txt
-        echo "" >> task.txt
-        echo "Description" >> task.txt
-        echo "${{ github.event.pull_request.body }}" >> task.txt
-        echo "" >> task.txt
-        echo "Diff file is: ${{ github.event.pull_request.number }}.diff" >> task.txt
-    - name: Set up environment
-      run: |
-        curl -sSL https://install.python-poetry.org | python3 -
-        export PATH="/github/home/.local/bin:$PATH"
-        poetry install --without evaluation,llama-index
-        poetry run playwright install --with-deps chromium
-    - name: Run OpenHands
-      env:
-        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: |
-        ls -la
-        if [[ -s review-${{ github.event.pull_request.number }}.txt ]]; then
-          echo "non_empty=true" >> $GITHUB_OUTPUT
-        fi
-      shell: bash
-    - name: Create PR review if file is non-empty
-      env:
-        GH_TOKEN: ${{ github.token }}
-      if: steps.check_file.outputs.non_empty == 'true'
-      run: |
-        gh pr review ${{ github.event.pull_request.number }} --comment --body-file "review-${{ github.event.pull_request.number }}.txt"

.github/workflows/stale.yml

@@ -1,21 +0,0 @@
-# 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 * * *'
-
-jobs:
-  stale:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/stale@v9
-        with:
-          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.'
-          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.'
-          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

@@ -1,230 +0,0 @@
-# Byte-compiled / optimized / DLL files
-__pycache__/
-*.py[cod]
-*$py.class
-
-# C extensions
-*.so
-
-# Distribution / packaging
-.Python
-build/
-develop-eggs/
-dist/
-downloads/
-eggs/
-.eggs/
-./lib/
-lib64/
-parts/
-sdist/
-var/
-wheels/
-share/python-wheels/
-*.egg-info/
-.installed.cfg
-*.egg
-MANIFEST
-requirements.txt
-
-# PyInstaller
-#  Usually these files are written by a python script from a template
-#  before PyInstaller builds the exe, so as to inject date/other infos into it.
-*.manifest
-*.spec
-
-# Installer logs
-pip-log.txt
-pip-delete-this-directory.txt
-
-# Unit test / coverage reports
-htmlcov/
-.tox/
-.nox/
-.coverage
-.coverage.*
-.cache
-nosetests.xml
-coverage.xml
-*.cover
-*.py,cover
-.hypothesis/
-.pytest_cache/
-cover/
-
-# Translations
-*.mo
-*.pot
-
-# Django stuff:
-local_settings.py
-db.sqlite3
-db.sqlite3-journal
-
-# Flask stuff:
-instance/
-.webassets-cache
-
-# Scrapy stuff:
-.scrapy
-
-# Sphinx documentation
-docs/_build/
-
-# PyBuilder
-.pybuilder/
-target/
-
-# Jupyter Notebook
-.ipynb_checkpoints
-
-# IPython
-profile_default/
-ipython_config.py
-
-# pyenv
-#   For a library or package, you might want to ignore these files since the code is
-#   intended to run in multiple environments; otherwise, check them in:
-.python-version
-
-# pipenv
-#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
-#   However, in case of collaboration, if having platform-specific dependencies or dependencies
-#   having no cross-platform support, pipenv may install dependencies that don't work, or not
-#   install all needed dependencies.
-#Pipfile.lock
-
-# poetry
-#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
-#   This is especially recommended for binary packages to ensure reproducibility, and is more
-#   commonly ignored for libraries.
-#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
-# poetry.lock
-
-# pdm
-#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
-#pdm.lock
-#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
-#   in version control.
-#   https://pdm.fming.dev/#use-with-ide
-.pdm.toml
-
-# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
-__pypackages__/
-
-# Celery stuff
-celerybeat-schedule
-celerybeat.pid
-
-# SageMath parsed files
-*.sage.py
-
-# Environments
-.env
-frontend/.env
-.venv
-env/
-venv/
-ENV/
-env.bak/
-.env.bak
-venv.bak/
-*venv/
-
-# Spyder project settings
-.spyderproject
-.spyproject
-
-# Rope project settings
-.ropeproject
-
-# mkdocs documentation
-/site
-
-# mypy
-.mypy_cache/
-.dmypy.json
-dmypy.json
-
-# Pyre type checker
-.pyre/
-
-# pytype static type analyzer
-.pytype/
-
-# Cython debug symbols
-cython_debug/
-
-# PyCharm
-#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
-#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
-#  and can be added to the global gitignore or merged into this file.  For a more nuclear
-#  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
-
-# dependencies
-frontend/node_modules
-frontend/.pnp
-frontend/bun.lockb
-frontend/yarn.lock
-.pnp.js
-
-# testing
-frontend/coverage
-test_results*
-/_test_files_tmp/
-
-# production
-frontend/build
-frontend/dist
-
-# misc
-.DS_Store
-.env.local
-.env.development.local
-.env.test.local
-.env.production.local
-
-npm-debug.log*
-yarn-debug.log*
-yarn-error.log*
-
-logs
-
-# agent
-.envrc
-/workspace
-/_test_workspace
-/debug
-cache
-
-# configuration
-config.toml
-config.toml_
-config.toml.bak
-
-# swe-bench-eval
-image_build_logs
-run_instance_logs
-
-runtime_*.tar
-
-# docker build
-containers/runtime/Dockerfile
-containers/runtime/project.tar.gz
-containers/runtime/code

.openhands_instructions

@@ -1,28 +0,0 @@
-OpenHands is an automated AI software engineer. It is a repo with a Python backend
-(in the `openhands` directory) and TypeScript frontend (in the `frontend` directory).
-
-General Setup:
-- To set up the entire repo, including frontend and backend, run `make build`
-- To run linting and type-checking before finishing the job, run `poetry run pre-commit run --all-files --config ./dev_config/python/.pre-commit-config.yaml`
-
-Backend:
-- Located in the `openhands` directory
-- Testing:
-  - All tests are in `tests/unit/test_*.py`
-  - To test new code, run `poetry run pytest tests/unit/test_xxx.py` where `xxx` is the appropriate file for the current functionality
-  - Write all tests with pytest
-
-Frontend:
-- Located in the `frontend` directory
-- Prerequisites: A recent version of NodeJS / NPM
-- Setup: Run `npm install` in the frontend directory
-- Testing:
-  - Run tests: `npm run test`
-  - To run specific tests: `npm run test -- -t "TestName"`
-- Building:
-  - Build for production: `npm run build`
-- Environment Variables:
-  - Set in `frontend/.env` or as environment variables
-  - Available variables: VITE_BACKEND_HOST, VITE_USE_TLS, VITE_INSECURE_SKIP_VERIFY, VITE_FRONTEND_PORT
-- Internationalization:
-  - Generate i18n declaration file: `npm run make-i18n`

CODE_OF_CONDUCT.md

@@ -1,133 +0,0 @@
-
-# Contributor Covenant Code of Conduct
-
-## Our Pledge
-
-We as members, contributors, and leaders pledge to make participation in our
-community a harassment-free experience for everyone, regardless of age, body
-size, visible or invisible disability, ethnicity, sex characteristics, gender
-identity and expression, level of experience, education, socio-economic status,
-nationality, personal appearance, race, caste, color, religion, or sexual
-identity and orientation.
-
-We pledge to act and interact in ways that contribute to an open, welcoming,
-diverse, inclusive, and healthy community.
-
-## Our Standards
-
-Examples of behavior that contributes to a positive environment for our
-community include:
-
-* Demonstrating empathy and kindness toward other people
-* Being respectful of differing opinions, viewpoints, and experiences
-* Giving and gracefully accepting constructive feedback
-* Accepting responsibility and apologizing to those affected by our mistakes,
-  and learning from the experience
-* Focusing on what is best not just for us as individuals, but for the overall
-  community
-
-Examples of unacceptable behavior include:
-
-* The use of sexualized language or imagery, and sexual attention or advances of
-  any kind
-* Trolling, insulting or derogatory comments, and personal or political attacks
-* Public or private harassment
-* Publishing others' private information, such as a physical or email address,
-  without their explicit permission
-* Other conduct which could reasonably be considered inappropriate in a
-  professional setting
-
-## Enforcement Responsibilities
-
-Community leaders are responsible for clarifying and enforcing our standards of
-acceptable behavior and will take appropriate and fair corrective action in
-response to any behavior that they deem inappropriate, threatening, offensive,
-or harmful.
-
-Community leaders have the right and responsibility to remove, edit, or reject
-comments, commits, code, wiki edits, issues, and other contributions that are
-not aligned to this Code of Conduct, and will communicate reasons for moderation
-decisions when appropriate.
-
-## Scope
-
-This Code of Conduct applies within all community spaces, and also applies when
-an individual is officially representing the community in public spaces.
-Examples of representing our community include using an official email address,
-posting via an official social media account, or acting as an appointed
-representative at an online or offline event.
-
-## Enforcement
-
-Instances of abusive, harassing, or otherwise unacceptable behavior may be
-reported to the community leaders responsible for enforcement at
-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
-reporter of any incident.
-
-## Enforcement Guidelines
-
-Community leaders will follow these Community Impact Guidelines in determining
-the consequences for any action they deem in violation of this Code of Conduct:
-
-### 1. Correction
-
-**Community Impact**: Use of inappropriate language or other behavior deemed
-unprofessional or unwelcome in the community.
-
-**Consequence**: A private, written warning from community leaders, providing
-clarity around the nature of the violation and an explanation of why the
-behavior was inappropriate. A public apology may be requested.
-
-### 2. Warning
-
-**Community Impact**: A violation through a single incident or series of
-actions.
-
-**Consequence**: A warning with consequences for continued behavior. No
-interaction with the people involved, including unsolicited interaction with
-those enforcing the Code of Conduct, for a specified period of time. This
-includes avoiding interactions in community spaces as well as external channels
-like social media. Violating these terms may lead to a temporary or permanent
-ban.
-
-### 3. Temporary Ban
-
-**Community Impact**: A serious violation of community standards, including
-sustained inappropriate behavior.
-
-**Consequence**: A temporary ban from any sort of interaction or public
-communication with the community for a specified period of time. No public or
-private interaction with the people involved, including unsolicited interaction
-with those enforcing the Code of Conduct, is allowed during this period.
-Violating these terms may lead to a permanent ban.
-
-### 4. Permanent Ban
-
-**Community Impact**: Demonstrating a pattern of violation of community
-standards, including sustained inappropriate behavior, harassment of an
-individual, or aggression toward or disparagement of classes of individuals.
-
-**Consequence**: A permanent ban from any sort of public interaction within the
-community.
-
-## Attribution
-
-This Code of Conduct is adapted from the [Contributor Covenant][homepage],
-version 2.1, available at
-[https://www.contributor-covenant.org/version/2/1/code_of_conduct.html][v2.1].
-
-Community Impact Guidelines were inspired by
-[Mozilla's code of conduct enforcement ladder][Mozilla CoC].
-
-For answers to common questions about this code of conduct, see the FAQ at
-[https://www.contributor-covenant.org/faq][FAQ]. Translations are available at
-[https://www.contributor-covenant.org/translations][translations].
-
-[homepage]: https://www.contributor-covenant.org
-[v2.1]: https://www.contributor-covenant.org/version/2/1/code_of_conduct.html
-[Mozilla CoC]: https://github.com/mozilla/diversity
-[FAQ]: https://www.contributor-covenant.org/faq
-[translations]: https://www.contributor-covenant.org/translations

CONTRIBUTING.md

@@ -1,117 +0,0 @@
-# Contributing
-
-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 issues](https://github.com/All-Hands-AI/OpenHands/labels/good%20first%20issue) 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)
-- [evaluation](./evaluation/README.md)
-- [openhands](./openhands/README.md)
-   - [agenthub](./openhands/agenthub/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
-
-### 1. Fork the Official Repository
-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>/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:
-
-```shell
-cd OpenHands
-git remote add upstream git@github.com:All-Hands-AI/OpenHands.git
-```
-
-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:
-
-```shell
-git fetch upstream
-git checkout main
-git merge upstream/main
-git push origin main
-```
-
-### 4. Set up the Development Environment
-
-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 [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
-- `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
-- `perf`: A code change that improves performance
-- `test`: Adding missing tests or correcting existing tests
-- `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
-
-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/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 contains a lot of changes, it's better to write more details.

CREDITS.md

@@ -1,312 +0,0 @@
-# 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. We greatly appreciate your dedication and hard work.
-
-## 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 agent hub
-
-#### [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

@@ -1,126 +0,0 @@
-# Development Guide
-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)  [ 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/) = 3.12
-* [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`.
-
-#### Develop without sudo access
-If you want to develop without system admin/sudo access to upgrade/install `Python` and/or `NodeJs`, you can use `conda` or `mamba` to manage the packages for you:
-
-```bash
-# Download and install Mamba (a faster version of conda)
-curl -L -O "https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-$(uname)-$(uname -m).sh"
-bash Miniforge3-$(uname)-$(uname -m).sh
-
-# Install Python 3.12, nodejs, and poetry
-mamba install python=3.12
-mamba install conda-forge::nodejs
-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:
-
-```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.
-
-To configure the LM of your choice, run:
-
-   ```bash
-   make setup-config
-   ```
-
-   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 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).
-
-### 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
-    ```
-
-- **Start the Frontend Server:** Similarly, you can start the frontend server on its own to work on frontend-related components or interface enhancements.
-    ```bash
-    make start-frontend
-    ```
-
-### 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.
-OpenHands will then log the prompts and responses in the logs/llm/CURRENT_DATE directory, allowing you to identify the causes.
-
-### 7. 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_*.py
-```
-
-### 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`
-
-### 9. Use existing Docker image
-To reduce build time (e.g., if no changes were made to the client-runtime component), you can use an existing Docker container image. Follow these steps:
-1. Set the SANDBOX_RUNTIME_CONTAINER_IMAGE environment variable to the desired Docker image.
-2. Example: export SANDBOX_RUNTIME_CONTAINER_IMAGE=ghcr.io/all-hands-ai/runtime:0.9-nikolaik
-
-## 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

@@ -1,25 +0,0 @@
-# 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

LICENSE

@@ -1,25 +0,0 @@
-The MIT License (MIT)
-=====================
-
-Copyright © 2023
-
-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.

MANIFEST.in

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

Makefile

@@ -1,332 +0,0 @@
-SHELL=/bin/bash
-# Makefile for OpenHands project
-
-# Variables
-BACKEND_HOST ?= "127.0.0.1"
-BACKEND_PORT = 3000
-BACKEND_HOST_PORT = "$(BACKEND_HOST):$(BACKEND_PORT)"
-FRONTEND_PORT = 3001
-DEFAULT_WORKSPACE_DIR = "./workspace"
-DEFAULT_MODEL = "gpt-4o"
-CONFIG_FILE = config.toml
-PRE_COMMIT_CONFIG_PATH = "./dev_config/python/.pre-commit-config.yaml"
-PYTHON_VERSION = 3.12
-
-# ANSI color codes
-GREEN=$(shell tput -Txterm setaf 2)
-YELLOW=$(shell tput -Txterm setaf 3)
-RED=$(shell tput -Txterm setaf 1)
-BLUE=$(shell tput -Txterm setaf 6)
-RESET=$(shell tput -Txterm sgr0)
-
-# Build
-build:
-	@echo "$(GREEN)Building project...$(RESET)"
-	@$(MAKE) -s check-dependencies
-	@$(MAKE) -s install-python-dependencies
-	@$(MAKE) -s install-frontend-dependencies
-	@$(MAKE) -s install-pre-commit-hooks
-	@$(MAKE) -s build-frontend
-	@echo "$(GREEN)Build completed successfully.$(RESET)"
-
-check-dependencies:
-	@echo "$(YELLOW)Checking dependencies...$(RESET)"
-	@$(MAKE) -s check-system
-	@$(MAKE) -s check-python
-	@$(MAKE) -s check-npm
-	@$(MAKE) -s check-nodejs
-ifeq ($(INSTALL_DOCKER),)
-	@$(MAKE) -s check-docker
-endif
-	@$(MAKE) -s check-poetry
-	@echo "$(GREEN)Dependencies checked successfully.$(RESET)"
-
-check-system:
-	@echo "$(YELLOW)Checking system...$(RESET)"
-	@if [ "$(shell uname)" = "Darwin" ]; then \
-		echo "$(BLUE)macOS detected.$(RESET)"; \
-	elif [ "$(shell uname)" = "Linux" ]; then \
-		if [ -f "/etc/manjaro-release" ]; then \
-			echo "$(BLUE)Manjaro Linux detected.$(RESET)"; \
-		else \
-			echo "$(BLUE)Linux detected.$(RESET)"; \
-		fi; \
-	elif [ "$$(uname -r | grep -i microsoft)" ]; then \
-		echo "$(BLUE)Windows Subsystem for Linux detected.$(RESET)"; \
-	else \
-		echo "$(RED)Unsupported system detected. Please use macOS, Linux, or Windows Subsystem for Linux (WSL).$(RESET)"; \
-		exit 1; \
-	fi
-
-check-python:
-	@echo "$(YELLOW)Checking Python installation...$(RESET)"
-	@if command -v python$(PYTHON_VERSION) > /dev/null; then \
-		echo "$(BLUE)$(shell python$(PYTHON_VERSION) --version) is already installed.$(RESET)"; \
-	else \
-		echo "$(RED)Python $(PYTHON_VERSION) is not installed. Please install Python $(PYTHON_VERSION) to continue.$(RESET)"; \
-		exit 1; \
-	fi
-
-check-npm:
-	@echo "$(YELLOW)Checking npm installation...$(RESET)"
-	@if command -v npm > /dev/null; then \
-		echo "$(BLUE)npm $(shell npm --version) is already installed.$(RESET)"; \
-	else \
-		echo "$(RED)npm is not installed. Please install Node.js to continue.$(RESET)"; \
-		exit 1; \
-	fi
-
-check-nodejs:
-	@echo "$(YELLOW)Checking Node.js installation...$(RESET)"
-	@if command -v node > /dev/null; then \
-		NODE_VERSION=$(shell node --version | sed -E 's/v//g'); \
-		IFS='.' read -r -a NODE_VERSION_ARRAY <<< "$$NODE_VERSION"; \
-		if [ "$${NODE_VERSION_ARRAY[0]}" -gt 18 ] || ([ "$${NODE_VERSION_ARRAY[0]}" -eq 18 ] && [ "$${NODE_VERSION_ARRAY[1]}" -gt 17 ]) || ([ "$${NODE_VERSION_ARRAY[0]}" -eq 18 ] && [ "$${NODE_VERSION_ARRAY[1]}" -eq 17 ] && [ "$${NODE_VERSION_ARRAY[2]}" -ge 1 ]); then \
-			echo "$(BLUE)Node.js $$NODE_VERSION is already installed.$(RESET)"; \
-		else \
-			echo "$(RED)Node.js 18.17.1 or later is required. Please install Node.js 18.17.1 or later to continue.$(RESET)"; \
-			exit 1; \
-		fi; \
-	else \
-		echo "$(RED)Node.js is not installed. Please install Node.js to continue.$(RESET)"; \
-		exit 1; \
-	fi
-
-check-docker:
-	@echo "$(YELLOW)Checking Docker installation...$(RESET)"
-	@if command -v docker > /dev/null; then \
-		echo "$(BLUE)$(shell docker --version) is already installed.$(RESET)"; \
-	else \
-		echo "$(RED)Docker is not installed. Please install Docker to continue.$(RESET)"; \
-		exit 1; \
-	fi
-
-check-poetry:
-	@echo "$(YELLOW)Checking Poetry installation...$(RESET)"
-	@if command -v poetry > /dev/null; then \
-		POETRY_VERSION=$(shell poetry --version 2>&1 | sed -E 's/Poetry \(version ([0-9]+\.[0-9]+\.[0-9]+)\)/\1/'); \
-		IFS='.' read -r -a POETRY_VERSION_ARRAY <<< "$$POETRY_VERSION"; \
-		if [ $${POETRY_VERSION_ARRAY[0]} -ge 1 ] && [ $${POETRY_VERSION_ARRAY[1]} -ge 8 ]; then \
-			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 | 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 | python$(PYTHON_VERSION) -$(RESET)"; \
-		echo "$(RED)More detail here: https://python-poetry.org/docs/#installing-with-the-official-installer$(RESET)"; \
-		exit 1; \
-	fi
-
-install-python-dependencies:
-	@echo "$(GREEN)Installing Python dependencies...$(RESET)"
-	@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 --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 \
-		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)"
-
-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
-	echo "$(BLUE)Installing frontend dependencies with npm...$(RESET)"
-	@cd frontend && npm install
-	@echo "$(GREEN)Frontend dependencies installed successfully.$(RESET)"
-
-install-pre-commit-hooks:
-	@echo "$(YELLOW)Installing pre-commit hooks...$(RESET)"
-	@git config --unset-all core.hooksPath || true
-	@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 openhands/**/* agenthub/**/* evaluation/**/* --show-diff-on-failure --config $(PRE_COMMIT_CONFIG_PATH)
-
-lint-frontend:
-	@echo "$(YELLOW)Running linters for frontend...$(RESET)"
-	@cd frontend && npm run lint
-
-lint:
-	@$(MAKE) -s lint-frontend
-	@$(MAKE) -s lint-backend
-
-test-frontend:
-	@echo "$(YELLOW)Running tests for frontend...$(RESET)"
-	@cd frontend && npm run test
-
-test:
-	@$(MAKE) -s test-frontend
-
-build-frontend:
-	@echo "$(YELLOW)Building frontend...$(RESET)"
-	@cd frontend && npm run build
-
-# Start backend
-start-backend:
-	@echo "$(YELLOW)Starting backend...$(RESET)"
-	@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_PORT) VITE_FRONTEND_PORT=$(FRONTEND_PORT) npm run dev -- --port $(FRONTEND_PORT)
-
-# Common setup for running the app (non-callable)
-_run_setup:
-	@if [ "$(OS)" = "Windows_NT" ]; then \
-		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 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 dev -- --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)"
-	@$(MAKE) setup-config-prompts
-	@mv $(CONFIG_FILE).tmp $(CONFIG_FILE)
-	@echo "$(GREEN)Config.toml setup completed.$(RESET)"
-
-setup-config-prompts:
-	@echo "[core]" > $(CONFIG_FILE).tmp
-
-	@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=\"$$workspace_dir\"" >> $(CONFIG_FILE).tmp
-
-	@echo "" >> $(CONFIG_FILE).tmp
-
-	@echo "[llm]" >> $(CONFIG_FILE).tmp
-	@read -p "Enter your LLM model name, used for running without UI. Set the model in the UI after you start the app. (see https://docs.litellm.ai/docs/providers for full list) [default: $(DEFAULT_MODEL)]: " llm_model; \
-	 llm_model=$${llm_model:-$(DEFAULT_MODEL)}; \
-	 echo "model=\"$$llm_model\"" >> $(CONFIG_FILE).tmp
-
-	@read -p "Enter your LLM api key: " llm_api_key; \
-	 echo "api_key=\"$$llm_api_key\"" >> $(CONFIG_FILE).tmp
-
-	@read -p "Enter your LLM base URL [mostly used for local LLMs, leave blank if not needed - example: http://localhost:5001/v1/]: " llm_base_url; \
-	 if [[ ! -z "$$llm_base_url" ]]; then echo "base_url=\"$$llm_base_url\"" >> $(CONFIG_FILE).tmp; fi
-
-	@echo "Enter your LLM Embedding Model"; \
-		echo "Choices are:"; \
-		echo "  - openai"; \
-		echo "  - azureopenai"; \
-		echo "  - Embeddings available only with OllamaEmbedding:"; \
-		echo "    - llama2"; \
-		echo "    - mxbai-embed-large"; \
-		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=\"$$llm_embedding_model\"" >> $(CONFIG_FILE).tmp; \
-		if [ "$$llm_embedding_model" = "llama2" ] || [ "$$llm_embedding_model" = "mxbai-embed-large" ] || [ "$$llm_embedding_model" = "nomic-embed-text" ] || [ "$$llm_embedding_model" = "all-minilm" ] || [ "$$llm_embedding_model" = "stable-code" ]; then \
-			read -p "Enter the local model URL for the embedding model (will set llm.embedding_base_url): " llm_embedding_base_url; \
-				echo "embedding_base_url=\"$$llm_embedding_base_url\"" >> $(CONFIG_FILE).tmp; \
-		elif [ "$$llm_embedding_model" = "azureopenai" ]; then \
-			read -p "Enter the Azure endpoint URL (will overwrite llm.base_url): " llm_base_url; \
-				echo "base_url=\"$$llm_base_url\"" >> $(CONFIG_FILE).tmp; \
-			read -p "Enter the Azure LLM Embedding Deployment Name: " llm_embedding_deployment_name; \
-				echo "embedding_deployment_name=\"$$llm_embedding_deployment_name\"" >> $(CONFIG_FILE).tmp; \
-			read -p "Enter the Azure API Version: " llm_api_version; \
-				echo "api_version=\"$$llm_api_version\"" >> $(CONFIG_FILE).tmp; \
-		fi
-
-
-# Develop in container
-docker-dev:
-	@if [ -f /.dockerenv ]; then \
-		echo "Running inside a Docker container. Exiting..."; \
-		exit 0; \
-	else \
-		echo "$(YELLOW)Build and run in Docker $(OPTIONS)...$(RESET)"; \
-		./containers/dev/dev.sh $(OPTIONS); \
-	fi
-
-# Clean up all caches
-clean:
-	@echo "$(YELLOW)Cleaning up caches...$(RESET)"
-	@rm -rf openhands/.cache
-	@echo "$(GREEN)Caches cleaned up successfully.$(RESET)"
-
-# Help
-help:
-	@echo "$(BLUE)Usage: make [target]$(RESET)"
-	@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 OpenHands by providing LLM API key,"
-	@echo "                        LLM Model name, and workspace directory."
-	@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 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

@@ -1,132 +0,0 @@
-<a name="readme-top"></a>
-
-<div align="center">
-  <img src="./docs/static/img/logo.png" alt="Logo" width="200">
-  <h1 align="center">OpenHands: Code Less, Make More</h1>
-</div>
-
-
-<div align="center">
-  <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-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://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>
-
-Welcome to OpenHands (formerly OpenDevin), a platform for software development agents powered by AI.
-
-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.
-
-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)
-
-## ⚡ Quick Start
-
-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 [Installation](https://docs.all-hands.dev/modules/usage/installation) guide for
-system requirements and more information.
-
-```bash
-export WORKSPACE_BASE=$(pwd)/workspace
-
-docker pull ghcr.io/all-hands-ai/runtime:0.9-nikolaik
-
-docker run -it --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=ghcr.io/all-hands-ai/runtime:0.9-nikolaik \
-    -e SANDBOX_USER_ID=$(id -u) \
-    -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
-    -v $WORKSPACE_BASE:/opt/workspace_base \
-    -v /var/run/docker.sock:/var/run/docker.sock \
-    -p 3000:3000 \
-    --add-host host.docker.internal:host-gateway \
-    --name openhands-app-$(date +%Y%m%d%H%M%S) \
-    ghcr.io/all-hands-ai/openhands:0.9
-```
-
-You'll find OpenHands running at [http://localhost:3000](http://localhost:3000)!
-
-You'll need a model provider and API key. One option that works well: [Claude 3.5 Sonnet](https://www.anthropic.com/api), but you have [many options](https://docs.all-hands.dev/modules/usage/llms).
-
----
-
-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).
-
-Visit [Installation](https://docs.all-hands.dev/modules/usage/installation) for more information and setup instructions.
-
-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
-
-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 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 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-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://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>
-
-## 📜 License
-
-Distributed under the MIT License. See [`LICENSE`](./LICENSE) for more information.
-
-## 🙏 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{openhands,
-      title={{OpenHands: 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},
-}
-```

__init__.py

@@ -4,26 +4,6 @@ __package_name__ = 'openhands_ai'
 
 
 def get_version():
-    try:
-        from importlib.metadata import PackageNotFoundError, version
-
-        try:
-            return version(__package_name__)
-        except PackageNotFoundError:
-            pass
-    except ImportError:
-        pass
-
-    try:
-        from pkg_resources import DistributionNotFound, get_distribution
-
-        try:
-            return get_distribution(__package_name__).version
-        except DistributionNotFound:
-            pass
-    except ImportError:
-        pass
-
     # Try getting the version from pyproject.toml
     try:
         root_dir = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
@@ -34,6 +14,20 @@ def get_version():
     except FileNotFoundError:
         pass
 
+    try:
+        from importlib.metadata import PackageNotFoundError, version
+
+        return version(__package_name__)
+    except (ImportError, PackageNotFoundError):
+        pass
+
+    try:
+        from pkg_resources import DistributionNotFound, get_distribution
+
+        return get_distribution(__package_name__).version
+    except (ImportError, DistributionNotFound):
+        pass
+
     return 'unknown'
 
 

agenthub/__init__.py

@@ -10,7 +10,6 @@ load_dotenv()
 from openhands.agenthub import (  # noqa: E402
     browsing_agent,
     codeact_agent,
-    codeact_swe_agent,
     delegator_agent,
     dummy_agent,
     planner_agent,
@@ -18,7 +17,6 @@ from openhands.agenthub import (  # noqa: E402
 
 __all__ = [
     'codeact_agent',
-    'codeact_swe_agent',
     'planner_agent',
     'delegator_agent',
     'dummy_agent',

agenthub/browsing_agent/browsing_agent.py

@@ -150,13 +150,13 @@ class BrowsingAgent(Agent):
         last_obs = None
         last_action = None
 
-        if EVAL_MODE and len(state.history.get_events_as_list()) == 1:
+        if EVAL_MODE and len(state.history) == 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():
+        for event in state.history:
             if isinstance(event, BrowseInteractiveAction):
                 prev_actions.append(event.browser_actions)
                 last_action = event

agenthub/codeact_agent/codeact_agent.py

@@ -0,0 +1,504 @@
+import json
+import os
+from collections import deque
+
+from litellm import ModelResponse
+
+import openhands.agenthub.codeact_agent.function_calling as codeact_function_calling
+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 ImageContent, Message, TextContent
+from openhands.events.action import (
+    Action,
+    AgentDelegateAction,
+    AgentFinishAction,
+    BrowseInteractiveAction,
+    BrowseURLAction,
+    CmdRunAction,
+    FileEditAction,
+    IPythonRunCellAction,
+    MessageAction,
+)
+from openhands.events.observation import (
+    AgentDelegateObservation,
+    BrowserOutputObservation,
+    CmdOutputObservation,
+    FileEditObservation,
+    IPythonRunCellObservation,
+    UserRejectObservation,
+)
+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,
+)
+from openhands.utils.prompt import PromptManager
+
+
+class CodeActAgent(Agent):
+    VERSION = '2.2'
+    """
+    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/2402.01030), [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:
+
+    1. **Converse**: Communicate with humans in natural language to ask for clarification, confirmation, etc.
+    2. **CodeAct**: Choose to perform the task by executing code
+    - 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/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 needs to be initialized before Jupyter for Jupyter to use those functions.
+        AgentSkillsRequirement(),
+        JupyterRequirement(),
+    ]
+
+    def __init__(
+        self,
+        llm: LLM,
+        config: AgentConfig,
+    ) -> None:
+        """Initializes a new instance of the CodeActAgent class.
+
+        Parameters:
+        - llm (LLM): The llm to be used by this agent
+        """
+        super().__init__(llm, config)
+        self.reset()
+
+        self.mock_function_calling = False
+        if not self.llm.is_function_calling_active():
+            logger.info(
+                f'Function calling not enabled for model {self.llm.config.model}. '
+                'Mocking function calling via prompting.'
+            )
+            self.mock_function_calling = True
+
+        # Function calling mode
+        self.tools = codeact_function_calling.get_tools(
+            codeact_enable_browsing=self.config.codeact_enable_browsing,
+            codeact_enable_jupyter=self.config.codeact_enable_jupyter,
+            codeact_enable_llm_editor=self.config.codeact_enable_llm_editor,
+        )
+        logger.debug(
+            f'TOOLS loaded for CodeActAgent: {json.dumps(self.tools, indent=2)}'
+        )
+        self.prompt_manager = PromptManager(
+            microagent_dir=os.path.join(os.path.dirname(__file__), 'micro')
+            if self.config.use_microagents
+            else None,
+            prompt_dir=os.path.join(os.path.dirname(__file__), 'prompts'),
+            disabled_microagents=self.config.disabled_microagents,
+        )
+
+        self.pending_actions: deque[Action] = deque()
+
+    def get_action_message(
+        self,
+        action: Action,
+        pending_tool_call_action_messages: dict[str, Message],
+    ) -> list[Message]:
+        """Converts an action into a message format that can be sent to the LLM.
+
+        This method handles different types of actions and formats them appropriately:
+        1. For tool-based actions (AgentDelegate, CmdRun, IPythonRunCell, FileEdit) and agent-sourced AgentFinish:
+            - In function calling mode: Stores the LLM's response in pending_tool_call_action_messages
+            - In non-function calling mode: Creates a message with the action string
+        2. For MessageActions: Creates a message with the text content and optional image content
+
+        Args:
+            action (Action): The action to convert. Can be one of:
+                - CmdRunAction: For executing bash commands
+                - IPythonRunCellAction: For running IPython code
+                - FileEditAction: For editing files
+                - BrowseInteractiveAction: For browsing the web
+                - AgentFinishAction: For ending the interaction
+                - MessageAction: For sending messages
+            pending_tool_call_action_messages (dict[str, Message]): Dictionary mapping response IDs
+                to their corresponding messages. Used in function calling mode to track tool calls
+                that are waiting for their results.
+
+        Returns:
+            list[Message]: A list containing the formatted message(s) for the action.
+                May be empty if the action is handled as a tool call in function calling mode.
+
+        Note:
+            In function calling mode, tool-based actions are stored in pending_tool_call_action_messages
+            rather than being returned immediately. They will be processed later when all corresponding
+            tool call results are available.
+        """
+        # create a regular message from an event
+        if isinstance(
+            action,
+            (
+                AgentDelegateAction,
+                IPythonRunCellAction,
+                FileEditAction,
+                BrowseInteractiveAction,
+                BrowseURLAction,
+            ),
+        ) or (isinstance(action, CmdRunAction) and action.source == 'agent'):
+            tool_metadata = action.tool_call_metadata
+            assert tool_metadata is not None, (
+                'Tool call metadata should NOT be None when function calling is enabled. Action: '
+                + str(action)
+            )
+
+            llm_response: ModelResponse = tool_metadata.model_response
+            assistant_msg = llm_response.choices[0].message
+
+            # Add the LLM message (assistant) that initiated the tool calls
+            # (overwrites any previous message with the same response_id)
+            logger.debug(
+                f'Tool calls type: {type(assistant_msg.tool_calls)}, value: {assistant_msg.tool_calls}'
+            )
+            pending_tool_call_action_messages[llm_response.id] = Message(
+                role=assistant_msg.role,
+                # tool call content SHOULD BE a string
+                content=[TextContent(text=assistant_msg.content or '')]
+                if assistant_msg.content is not None
+                else [],
+                tool_calls=assistant_msg.tool_calls,
+            )
+            return []
+        elif isinstance(action, AgentFinishAction):
+            role = 'user' if action.source == 'user' else 'assistant'
+
+            # when agent finishes, it has tool_metadata
+            # which has already been executed, and it doesn't have a response
+            # when the user finishes (/exit), we don't have tool_metadata
+            tool_metadata = action.tool_call_metadata
+            if tool_metadata is not None:
+                # take the response message from the tool call
+                assistant_msg = tool_metadata.model_response.choices[0].message
+                content = assistant_msg.content or ''
+
+                # save content if any, to thought
+                if action.thought:
+                    if action.thought != content:
+                        action.thought += '\n' + content
+                else:
+                    action.thought = content
+
+                # remove the tool call metadata
+                action.tool_call_metadata = None
+            return [
+                Message(
+                    role=role,
+                    content=[TextContent(text=action.thought)],
+                )
+            ]
+        elif isinstance(action, MessageAction):
+            role = 'user' if action.source == 'user' else 'assistant'
+            content = [TextContent(text=action.content or '')]
+            if self.llm.vision_is_active() and action.image_urls:
+                content.append(ImageContent(image_urls=action.image_urls))
+            return [
+                Message(
+                    role=role,
+                    content=content,
+                )
+            ]
+        elif isinstance(action, CmdRunAction) and action.source == 'user':
+            content = [
+                TextContent(text=f'User executed the command:\n{action.command}')
+            ]
+            return [
+                Message(
+                    role='user',
+                    content=content,
+                )
+            ]
+        return []
+
+    def get_observation_message(
+        self,
+        obs: Observation,
+        tool_call_id_to_message: dict[str, Message],
+    ) -> list[Message]:
+        """Converts an observation into a message format that can be sent to the LLM.
+
+        This method handles different types of observations and formats them appropriately:
+        - CmdOutputObservation: Formats command execution results with exit codes
+        - IPythonRunCellObservation: Formats IPython cell execution results, replacing base64 images
+        - FileEditObservation: Formats file editing results
+        - AgentDelegateObservation: Formats results from delegated agent tasks
+        - ErrorObservation: Formats error messages from failed actions
+        - UserRejectObservation: Formats user rejection messages
+
+        In function calling mode, observations with tool_call_metadata are stored in
+        tool_call_id_to_message for later processing instead of being returned immediately.
+
+        Args:
+            obs (Observation): The observation to convert
+            tool_call_id_to_message (dict[str, Message]): Dictionary mapping tool call IDs
+                to their corresponding messages (used in function calling mode)
+
+        Returns:
+            list[Message]: A list containing the formatted message(s) for the observation.
+                May be empty if the observation is handled as a tool response in function calling mode.
+
+        Raises:
+            ValueError: If the observation type is unknown
+        """
+        message: Message
+        max_message_chars = self.llm.config.max_message_chars
+        if isinstance(obs, CmdOutputObservation):
+            # if it doesn't have tool call metadata, it was triggered by a user action
+            if obs.tool_call_metadata is None:
+                text = truncate_content(
+                    f'\nObserved result of command executed by user:\n{obs.content}',
+                    max_message_chars,
+                )
+            else:
+                text = truncate_content(
+                    obs.content + obs.interpreter_details, max_message_chars
+                )
+            text += f'\n[Command finished with exit code {obs.exit_code}]'
+            message = Message(role='user', content=[TextContent(text=text)])
+        elif isinstance(obs, IPythonRunCellObservation):
+            text = 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)
+            message = Message(role='user', content=[TextContent(text=text)])
+        elif isinstance(obs, FileEditObservation):
+            text = truncate_content(str(obs), max_message_chars)
+            message = Message(role='user', content=[TextContent(text=text)])
+        elif isinstance(obs, BrowserOutputObservation):
+            text = obs.get_agent_obs_text()
+            message = Message(
+                role='user',
+                content=[TextContent(text=text)],
+            )
+        elif isinstance(obs, AgentDelegateObservation):
+            text = truncate_content(
+                obs.outputs['content'] if 'content' in obs.outputs else '',
+                max_message_chars,
+            )
+            message = Message(role='user', content=[TextContent(text=text)])
+        elif isinstance(obs, ErrorObservation):
+            text = truncate_content(obs.content, max_message_chars)
+            text += '\n[Error occurred in processing last action]'
+            message = 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]'
+            message = 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)}')
+
+        # Update the message as tool response properly
+        if (tool_call_metadata := obs.tool_call_metadata) is not None:
+            tool_call_id_to_message[tool_call_metadata.tool_call_id] = Message(
+                role='tool',
+                content=message.content,
+                tool_call_id=tool_call_metadata.tool_call_id,
+                name=tool_call_metadata.function_name,
+            )
+            # No need to return the observation message
+            # because it will be added by get_action_message when all the corresponding
+            # tool calls in the SAME request are processed
+            return []
+
+        return [message]
+
+    def reset(self) -> None:
+        """Resets the CodeAct Agent."""
+        super().reset()
+
+    def step(self, state: State) -> Action:
+        """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
+
+        Returns:
+        - 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
+        """
+        # Continue with pending actions if any
+        if self.pending_actions:
+            return self.pending_actions.popleft()
+
+        # if we're done, go back
+        latest_user_message = state.get_last_user_message()
+        if latest_user_message and latest_user_message.content.strip() == '/exit':
+            return AgentFinishAction()
+
+        # prepare what we want to send to the LLM
+        messages = self._get_messages(state)
+        params: dict = {
+            'messages': self.llm.format_messages_for_llm(messages),
+        }
+        params['tools'] = self.tools
+        if self.mock_function_calling:
+            params['mock_function_calling'] = True
+        response = self.llm.completion(**params)
+        actions = codeact_function_calling.response_to_actions(response)
+        for action in actions:
+            self.pending_actions.append(action)
+        return self.pending_actions.popleft()
+
+    def _get_messages(self, state: State) -> list[Message]:
+        """Constructs the message history for the LLM conversation.
+
+        This method builds a structured conversation history by processing events from the state
+        and formatting them into messages that the LLM can understand. It handles both regular
+        message flow and function-calling scenarios.
+
+        The method performs the following steps:
+        1. Initializes with system prompt and optional initial user message
+        2. Processes events (Actions and Observations) into messages
+        3. Handles tool calls and their responses in function-calling mode
+        4. Manages message role alternation (user/assistant/tool)
+        5. Applies caching for specific LLM providers (e.g., Anthropic)
+        6. Adds environment reminders for non-function-calling mode
+
+        Args:
+            state (State): The current state object containing conversation history and other metadata
+
+        Returns:
+            list[Message]: A list of formatted messages ready for LLM consumption, including:
+                - System message with prompt
+                - Initial user message (if configured)
+                - Action messages (from both user and assistant)
+                - Observation messages (including tool responses)
+                - Environment reminders (in non-function-calling mode)
+
+        Note:
+            - In function-calling mode, tool calls and their responses are carefully tracked
+              to maintain proper conversation flow
+            - Messages from the same role are combined to prevent consecutive same-role messages
+            - For Anthropic models, specific messages are cached according to their documentation
+        """
+        if not self.prompt_manager:
+            raise Exception('Prompt Manager not instantiated.')
+
+        messages: list[Message] = [
+            Message(
+                role='system',
+                content=[
+                    TextContent(
+                        text=self.prompt_manager.get_system_message(),
+                        cache_prompt=self.llm.is_caching_prompt_active(),
+                    )
+                ],
+            )
+        ]
+        example_message = self.prompt_manager.get_example_user_message()
+        if example_message:
+            messages.append(
+                Message(
+                    role='user',
+                    content=[TextContent(text=example_message)],
+                    cache_prompt=self.llm.is_caching_prompt_active(),
+                )
+            )
+
+        pending_tool_call_action_messages: dict[str, Message] = {}
+        tool_call_id_to_message: dict[str, Message] = {}
+        events = list(state.history)
+        for event in events:
+            # create a regular message from an event
+            if isinstance(event, Action):
+                messages_to_add = self.get_action_message(
+                    action=event,
+                    pending_tool_call_action_messages=pending_tool_call_action_messages,
+                )
+            elif isinstance(event, Observation):
+                messages_to_add = self.get_observation_message(
+                    obs=event,
+                    tool_call_id_to_message=tool_call_id_to_message,
+                )
+            else:
+                raise ValueError(f'Unknown event type: {type(event)}')
+
+            # Check pending tool call action messages and see if they are complete
+            _response_ids_to_remove = []
+            for (
+                response_id,
+                pending_message,
+            ) in pending_tool_call_action_messages.items():
+                assert pending_message.tool_calls is not None, (
+                    'Tool calls should NOT be None when function calling is enabled & the message is considered pending tool call. '
+                    f'Pending message: {pending_message}'
+                )
+                if all(
+                    tool_call.id in tool_call_id_to_message
+                    for tool_call in pending_message.tool_calls
+                ):
+                    # If complete:
+                    # -- 1. Add the message that **initiated** the tool calls
+                    messages_to_add.append(pending_message)
+                    # -- 2. Add the tool calls **results***
+                    for tool_call in pending_message.tool_calls:
+                        messages_to_add.append(tool_call_id_to_message[tool_call.id])
+                        tool_call_id_to_message.pop(tool_call.id)
+                    _response_ids_to_remove.append(response_id)
+            # Cleanup the processed pending tool messages
+            for response_id in _response_ids_to_remove:
+                pending_tool_call_action_messages.pop(response_id)
+
+            for message in messages_to_add:
+                if message:
+                    if message.role == 'user':
+                        self.prompt_manager.enhance_message(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
+                    # NOTE: we shouldn't combine tool messages because each of them has a different tool_call_id
+                    if (
+                        messages
+                        and messages[-1].role == message.role
+                        and message.role != 'tool'
+                    ):
+                        messages[-1].content.extend(message.content)
+                    else:
+                        messages.append(message)
+
+        if self.llm.is_caching_prompt_active():
+            # NOTE: this is only needed for anthropic
+            # following logic here:
+            # https://github.com/anthropics/anthropic-quickstarts/blob/8f734fd08c425c6ec91ddd613af04ff87d70c5a0/computer-use-demo/computer_use_demo/loop.py#L241-L262
+            breakpoints_remaining = 3  # remaining 1 for system/tool
+            for message in reversed(messages):
+                if message.role == 'user' or message.role == 'tool':
+                    if breakpoints_remaining > 0:
+                        message.content[
+                            -1
+                        ].cache_prompt = True  # Last item inside the message content
+                        breakpoints_remaining -= 1
+                    else:
+                        break
+
+        return messages

agenthub/codeact_agent/function_calling.py

@@ -0,0 +1,554 @@
+"""This file contains the function calling implementation for different actions.
+
+This is similar to the functionality of `CodeActResponseParser`.
+"""
+
+import json
+
+from browsergym.core.action.highlevel import HighLevelActionSet
+from litellm import (
+    ChatCompletionToolParam,
+    ChatCompletionToolParamFunctionChunk,
+    ModelResponse,
+)
+
+from openhands.core.exceptions import FunctionCallNotExistsError
+from openhands.core.logger import openhands_logger as logger
+from openhands.events.action import (
+    Action,
+    AgentDelegateAction,
+    AgentFinishAction,
+    BrowseInteractiveAction,
+    BrowseURLAction,
+    CmdRunAction,
+    FileEditAction,
+    IPythonRunCellAction,
+    MessageAction,
+)
+from openhands.events.tool import ToolCallMetadata
+
+_BASH_DESCRIPTION = """Execute a bash command in the terminal.
+* Long running commands: For commands that may run indefinitely, it should be run in the background and the output should be redirected to a file, e.g. command = `python3 app.py > server.log 2>&1 &`.
+* Interactive: If a bash command returns exit code `-1`, this means the process is not yet finished. The assistant must then send a second call to terminal with an empty `command` (which will retrieve any additional logs), or it can send additional text (set `command` to the text) to STDIN of the running process, or it can send command=`ctrl+c` to interrupt the process.
+* Timeout: If a command execution result says "Command timed out. Sending SIGINT to the process", the assistant should retry running the command in the background.
+"""
+
+CmdRunTool = ChatCompletionToolParam(
+    type='function',
+    function=ChatCompletionToolParamFunctionChunk(
+        name='execute_bash',
+        description=_BASH_DESCRIPTION,
+        parameters={
+            'type': 'object',
+            'properties': {
+                'command': {
+                    'type': 'string',
+                    'description': 'The bash command to execute. Can be empty to view additional logs when previous exit code is `-1`. Can be `ctrl+c` to interrupt the currently running process.',
+                },
+            },
+            'required': ['command'],
+        },
+    ),
+)
+
+_IPYTHON_DESCRIPTION = """Run a cell of Python code in an IPython environment.
+* The assistant should define variables and import packages before using them.
+* The variable defined in the IPython environment will not be available outside the IPython environment (e.g., in terminal).
+"""
+
+IPythonTool = ChatCompletionToolParam(
+    type='function',
+    function=ChatCompletionToolParamFunctionChunk(
+        name='execute_ipython_cell',
+        description=_IPYTHON_DESCRIPTION,
+        parameters={
+            'type': 'object',
+            'properties': {
+                'code': {
+                    'type': 'string',
+                    'description': 'The Python code to execute. Supports magic commands like %pip.',
+                },
+            },
+            'required': ['code'],
+        },
+    ),
+)
+
+_FILE_EDIT_DESCRIPTION = """Edit a file.
+* The assistant can edit files by specifying the file path and providing a draft of the new file content.
+* The draft content doesn't need to be exactly the same as the existing file; the assistant may skip unchanged lines using comments like `# unchanged` to indicate unchanged sections.
+* IMPORTANT: For large files (e.g., > 300 lines), specify the range of lines to edit using `start` and `end` (1-indexed, inclusive). The range should be smaller than 300 lines.
+* To append to a file, set both `start` and `end` to `-1`.
+* If the file doesn't exist, a new file will be created with the provided content.
+
+**Example 1: general edit for short files**
+For example, given an existing file `/path/to/file.py` that looks like this:
+(this is the end of the file)
+1|class MyClass:
+2|    def __init__(self):
+3|        self.x = 1
+4|        self.y = 2
+5|        self.z = 3
+6|
+7|print(MyClass().z)
+8|print(MyClass().x)
+(this is the end of the file)
+
+The assistant wants to edit the file to look like this:
+(this is the end of the file)
+1|class MyClass:
+2|    def __init__(self):
+3|        self.x = 1
+4|        self.y = 2
+5|
+6|print(MyClass().y)
+(this is the end of the file)
+
+The assistant may produce an edit action like this:
+path="/path/to/file.txt" start=1 end=-1
+content=```
+class MyClass:
+    def __init__(self):
+        # no changes before
+        self.y = 2
+        # self.z is removed
+
+# MyClass().z is removed
+print(MyClass().y)
+```
+
+**Example 2: append to file for short files**
+For example, given an existing file `/path/to/file.py` that looks like this:
+(this is the end of the file)
+1|class MyClass:
+2|    def __init__(self):
+3|        self.x = 1
+4|        self.y = 2
+5|        self.z = 3
+6|
+7|print(MyClass().z)
+8|print(MyClass().x)
+(this is the end of the file)
+
+To append the following lines to the file:
+```python
+print(MyClass().y)
+```
+
+The assistant may produce an edit action like this:
+path="/path/to/file.txt" start=-1 end=-1
+content=```
+print(MyClass().y)
+```
+
+**Example 3: edit for long files**
+
+Given an existing file `/path/to/file.py` that looks like this:
+(1000 more lines above)
+1001|class MyClass:
+1002|    def __init__(self):
+1003|        self.x = 1
+1004|        self.y = 2
+1005|        self.z = 3
+1006|
+1007|print(MyClass().z)
+1008|print(MyClass().x)
+(2000 more lines below)
+
+The assistant wants to edit the file to look like this:
+
+(1000 more lines above)
+1001|class MyClass:
+1002|    def __init__(self):
+1003|        self.x = 1
+1004|        self.y = 2
+1005|
+1006|print(MyClass().y)
+(2000 more lines below)
+
+The assistant may produce an edit action like this:
+path="/path/to/file.txt" start=1001 end=1008
+content=```
+class MyClass:
+    def __init__(self):
+        # no changes before
+        self.y = 2
+        # self.z is removed
+
+# MyClass().z is removed
+print(MyClass().y)
+```
+"""
+
+LLMBasedFileEditTool = ChatCompletionToolParam(
+    type='function',
+    function=ChatCompletionToolParamFunctionChunk(
+        name='edit_file',
+        description=_FILE_EDIT_DESCRIPTION,
+        parameters={
+            'type': 'object',
+            'properties': {
+                'path': {
+                    'type': 'string',
+                    'description': 'The absolute path to the file to be edited.',
+                },
+                'new_content_draft': {
+                    'type': 'string',
+                    'description': 'A draft of the new content for the file being edited. Note that the assistant may skip unchanged lines.',
+                },
+                'start': {
+                    'type': 'integer',
+                    'description': 'The starting line number for the edit (1-indexed, inclusive). Default is 1.',
+                },
+                'end': {
+                    'type': 'integer',
+                    'description': 'The ending line number for the edit (1-indexed, inclusive). Default is -1 (end of file).',
+                },
+            },
+            'required': ['path', 'content'],
+        },
+    ),
+)
+
+_STR_REPLACE_EDITOR_DESCRIPTION = """Custom editing tool for viewing, creating and editing files
+* State is persistent across command calls and discussions with the user
+* If `path` is a file, `view` displays the result of applying `cat -n`. If `path` is a directory, `view` lists non-hidden files and directories up to 2 levels deep
+* The `create` command cannot be used if the specified `path` already exists as a file
+* If a `command` generates a long output, it will be truncated and marked with `<response clipped>`
+* The `undo_edit` command will revert the last edit made to the file at `path`
+
+Notes for using the `str_replace` command:
+* The `old_str` parameter should match EXACTLY one or more consecutive lines from the original file. Be mindful of whitespaces!
+* If the `old_str` parameter is not unique in the file, the replacement will not be performed. Make sure to include enough context in `old_str` to make it unique
+* The `new_str` parameter should contain the edited lines that should replace the `old_str`
+"""
+
+StrReplaceEditorTool = ChatCompletionToolParam(
+    type='function',
+    function=ChatCompletionToolParamFunctionChunk(
+        name='str_replace_editor',
+        description=_STR_REPLACE_EDITOR_DESCRIPTION,
+        parameters={
+            'type': 'object',
+            'properties': {
+                'command': {
+                    'description': 'The commands to run. Allowed options are: `view`, `create`, `str_replace`, `insert`, `undo_edit`.',
+                    'enum': ['view', 'create', 'str_replace', 'insert', 'undo_edit'],
+                    'type': 'string',
+                },
+                'path': {
+                    'description': 'Absolute path to file or directory, e.g. `/workspace/file.py` or `/workspace`.',
+                    'type': 'string',
+                },
+                'file_text': {
+                    'description': 'Required parameter of `create` command, with the content of the file to be created.',
+                    'type': 'string',
+                },
+                'old_str': {
+                    'description': 'Required parameter of `str_replace` command containing the string in `path` to replace.',
+                    'type': 'string',
+                },
+                'new_str': {
+                    'description': 'Optional parameter of `str_replace` command containing the new string (if not given, no string will be added). Required parameter of `insert` command containing the string to insert.',
+                    'type': 'string',
+                },
+                'insert_line': {
+                    'description': 'Required parameter of `insert` command. The `new_str` will be inserted AFTER the line `insert_line` of `path`.',
+                    'type': 'integer',
+                },
+                'view_range': {
+                    'description': 'Optional parameter of `view` command when `path` points to a file. If none is given, the full file is shown. If provided, the file will be shown in the indicated line number range, e.g. [11, 12] will show lines 11 and 12. Indexing at 1 to start. Setting `[start_line, -1]` shows all lines from `start_line` to the end of the file.',
+                    'items': {'type': 'integer'},
+                    'type': 'array',
+                },
+            },
+            'required': ['command', 'path'],
+        },
+    ),
+)
+
+
+_WEB_DESCRIPTION = """Read (convert to markdown) content from a webpage. You should prefer using the `webpage_read` tool over the `browser` tool, but do use the `browser` tool if you need to interact with a webpage (e.g., click a button, fill out a form, etc.).
+
+You may use the `webpage_read` tool to read content from a webpage, and even search the webpage content using a Google search query (e.g., url=`https://www.google.com/search?q=YOUR_QUERY`).
+"""
+
+WebReadTool = ChatCompletionToolParam(
+    type='function',
+    function=ChatCompletionToolParamFunctionChunk(
+        name='web_read',
+        description=_WEB_DESCRIPTION,
+        parameters={
+            'type': 'object',
+            'properties': {
+                'url': {
+                    'type': 'string',
+                    'description': 'The URL of the webpage to read. You can also use a Google search query here (e.g., `https://www.google.com/search?q=YOUR_QUERY`).',
+                }
+            },
+            'required': ['url'],
+        },
+    ),
+)
+
+# from browsergym/core/action/highlevel.py
+_browser_action_space = HighLevelActionSet(
+    subsets=['bid', 'nav'],
+    strict=False,  # less strict on the parsing of the actions
+    multiaction=True,  # enable to agent to take multiple actions at once
+)
+
+
+_BROWSER_DESCRIPTION = """Interact with the browser using Python code. Use it ONLY when you need to interact with a webpage.
+
+See the description of "code" parameter for more details.
+
+Multiple actions can be provided at once, but will be executed sequentially without any feedback from the page.
+More than 2-3 actions usually leads to failure or unexpected behavior. Example:
+fill('a12', 'example with "quotes"')
+click('a51')
+click('48', button='middle', modifiers=['Shift'])
+"""
+
+_BROWSER_TOOL_DESCRIPTION = """
+The following 15 functions are available. Nothing else is supported.
+
+goto(url: str)
+    Description: Navigate to a url.
+    Examples:
+        goto('http://www.example.com')
+
+go_back()
+    Description: Navigate to the previous page in history.
+    Examples:
+        go_back()
+
+go_forward()
+    Description: Navigate to the next page in history.
+    Examples:
+        go_forward()
+
+noop(wait_ms: float = 1000)
+    Description: Do nothing, and optionally wait for the given time (in milliseconds).
+    You can use this to get the current page content and/or wait for the page to load.
+    Examples:
+        noop()
+
+        noop(500)
+
+scroll(delta_x: float, delta_y: float)
+    Description: Scroll horizontally and vertically. Amounts in pixels, positive for right or down scrolling, negative for left or up scrolling. Dispatches a wheel event.
+    Examples:
+        scroll(0, 200)
+
+        scroll(-50.2, -100.5)
+
+fill(bid: str, value: str)
+    Description: Fill out a form field. It focuses the element and triggers an input event with the entered text. It works for <input>, <textarea> and [contenteditable] elements.
+    Examples:
+        fill('237', 'example value')
+
+        fill('45', 'multi-line\nexample')
+
+        fill('a12', 'example with "quotes"')
+
+select_option(bid: str, options: str | list[str])
+    Description: Select one or multiple options in a <select> element. You can specify option value or label to select. Multiple options can be selected.
+    Examples:
+        select_option('a48', 'blue')
+
+        select_option('c48', ['red', 'green', 'blue'])
+
+click(bid: str, button: Literal['left', 'middle', 'right'] = 'left', modifiers: list[typing.Literal['Alt', 'Control', 'ControlOrMeta', 'Meta', 'Shift']] = [])
+    Description: Click an element.
+    Examples:
+        click('a51')
+
+        click('b22', button='right')
+
+        click('48', button='middle', modifiers=['Shift'])
+
+dblclick(bid: str, button: Literal['left', 'middle', 'right'] = 'left', modifiers: list[typing.Literal['Alt', 'Control', 'ControlOrMeta', 'Meta', 'Shift']] = [])
+    Description: Double click an element.
+    Examples:
+        dblclick('12')
+
+        dblclick('ca42', button='right')
+
+        dblclick('178', button='middle', modifiers=['Shift'])
+
+hover(bid: str)
+    Description: Hover over an element.
+    Examples:
+        hover('b8')
+
+press(bid: str, key_comb: str)
+    Description: Focus the matching element and press a combination of keys. It accepts the logical key names that are emitted in the keyboardEvent.key property of the keyboard events: Backquote, Minus, Equal, Backslash, Backspace, Tab, Delete, Escape, ArrowDown, End, Enter, Home, Insert, PageDown, PageUp, ArrowRight, ArrowUp, F1 - F12, Digit0 - Digit9, KeyA - KeyZ, etc. You can alternatively specify a single character you'd like to produce such as "a" or "#". Following modification shortcuts are also supported: Shift, Control, Alt, Meta, ShiftLeft, ControlOrMeta. ControlOrMeta resolves to Control on Windows and Linux and to Meta on macOS.
+    Examples:
+        press('88', 'Backspace')
+
+        press('a26', 'ControlOrMeta+a')
+
+        press('a61', 'Meta+Shift+t')
+
+focus(bid: str)
+    Description: Focus the matching element.
+    Examples:
+        focus('b455')
+
+clear(bid: str)
+    Description: Clear the input field.
+    Examples:
+        clear('996')
+
+drag_and_drop(from_bid: str, to_bid: str)
+    Description: Perform a drag & drop. Hover the element that will be dragged. Press left mouse button. Move mouse to the element that will receive the drop. Release left mouse button.
+    Examples:
+        drag_and_drop('56', '498')
+
+upload_file(bid: str, file: str | list[str])
+    Description: Click an element and wait for a "filechooser" event, then select one or multiple input files for upload. Relative file paths are resolved relative to the current working directory. An empty list clears the selected files.
+    Examples:
+        upload_file('572', '/home/user/my_receipt.pdf')
+
+        upload_file('63', ['/home/bob/Documents/image.jpg', '/home/bob/Documents/file.zip'])
+"""
+
+
+for _, action in _browser_action_space.action_set.items():
+    assert (
+        action.signature in _BROWSER_TOOL_DESCRIPTION
+    ), f'Browser description mismatch. Please double check if the BrowserGym updated their action space.\n\nAction: {action.signature}'
+    assert (
+        action.description in _BROWSER_TOOL_DESCRIPTION
+    ), f'Browser description mismatch. Please double check if the BrowserGym updated their action space.\n\nAction: {action.description}'
+
+BrowserTool = ChatCompletionToolParam(
+    type='function',
+    function=ChatCompletionToolParamFunctionChunk(
+        name='browser',
+        description=_BROWSER_DESCRIPTION,
+        parameters={
+            'type': 'object',
+            'properties': {
+                'code': {
+                    'type': 'string',
+                    'description': (
+                        'The Python code that interacts with the browser.\n'
+                        + _BROWSER_TOOL_DESCRIPTION
+                    ),
+                }
+            },
+            'required': ['code'],
+        },
+    ),
+)
+
+_FINISH_DESCRIPTION = """Finish the interaction when the task is complete OR if the assistant cannot proceed further with the task."""
+
+FinishTool = ChatCompletionToolParam(
+    type='function',
+    function=ChatCompletionToolParamFunctionChunk(
+        name='finish',
+        description=_FINISH_DESCRIPTION,
+    ),
+)
+
+
+def combine_thought(action: Action, thought: str) -> Action:
+    if not hasattr(action, 'thought'):
+        return action
+    if thought:
+        action.thought = thought
+    return action
+
+
+def response_to_actions(response: ModelResponse) -> list[Action]:
+    actions: list[Action] = []
+    assert len(response.choices) == 1, 'Only one choice is supported for now'
+    assistant_msg = response.choices[0].message
+    if assistant_msg.tool_calls:
+        # Check if there's assistant_msg.content. If so, add it to the thought
+        thought = ''
+        if isinstance(assistant_msg.content, str):
+            thought = assistant_msg.content
+        elif isinstance(assistant_msg.content, list):
+            for msg in assistant_msg.content:
+                if msg['type'] == 'text':
+                    thought += msg['text']
+
+        # Process each tool call to OpenHands action
+        for i, tool_call in enumerate(assistant_msg.tool_calls):
+            action: Action
+            try:
+                arguments = json.loads(tool_call.function.arguments)
+            except json.decoder.JSONDecodeError as e:
+                raise RuntimeError(
+                    f'Failed to parse tool call arguments: {tool_call.function.arguments}'
+                ) from e
+            if tool_call.function.name == 'execute_bash':
+                action = CmdRunAction(**arguments)
+            elif tool_call.function.name == 'execute_ipython_cell':
+                action = IPythonRunCellAction(**arguments)
+            elif tool_call.function.name == 'delegate_to_browsing_agent':
+                action = AgentDelegateAction(
+                    agent='BrowsingAgent',
+                    inputs=arguments,
+                )
+            elif tool_call.function.name == 'finish':
+                action = AgentFinishAction()
+            elif tool_call.function.name == 'edit_file':
+                action = FileEditAction(**arguments)
+            elif tool_call.function.name == 'str_replace_editor':
+                # We implement this in agent_skills, which can be used via Jupyter
+                # convert tool_call.function.arguments to kwargs that can be passed to file_editor
+                code = f'print(file_editor(**{arguments}))'
+                logger.debug(
+                    f'TOOL CALL: str_replace_editor -> file_editor with code: {code}'
+                )
+                action = IPythonRunCellAction(code=code, include_extra=False)
+            elif tool_call.function.name == 'browser':
+                action = BrowseInteractiveAction(browser_actions=arguments['code'])
+            elif tool_call.function.name == 'web_read':
+                action = BrowseURLAction(url=arguments['url'])
+            else:
+                raise FunctionCallNotExistsError(
+                    f'Tool {tool_call.function.name} is not registered. (arguments: {arguments}). Please check the tool name and retry with an existing tool.'
+                )
+
+            # We only add thought to the first action
+            if i == 0:
+                action = combine_thought(action, thought)
+            # Add metadata for tool calling
+            action.tool_call_metadata = ToolCallMetadata(
+                tool_call_id=tool_call.id,
+                function_name=tool_call.function.name,
+                model_response=response,
+                total_calls_in_response=len(assistant_msg.tool_calls),
+            )
+            actions.append(action)
+    else:
+        actions.append(
+            MessageAction(content=assistant_msg.content, wait_for_response=True)
+        )
+
+    assert len(actions) >= 1
+    return actions
+
+
+def get_tools(
+    codeact_enable_browsing: bool = False,
+    codeact_enable_llm_editor: bool = False,
+    codeact_enable_jupyter: bool = False,
+) -> list[ChatCompletionToolParam]:
+    tools = [CmdRunTool, FinishTool]
+    if codeact_enable_browsing:
+        tools.append(WebReadTool)
+        tools.append(BrowserTool)
+    if codeact_enable_jupyter:
+        tools.append(IPythonTool)
+    if codeact_enable_llm_editor:
+        tools.append(LLMBasedFileEditTool)
+    else:
+        tools.append(StrReplaceEditorTool)
+    return tools

agenthub/codeact_agent/prompts/system_prompt.j2

@@ -0,0 +1,6 @@
+You are OpenHands agent, a helpful AI assistant that can interact with a computer to solve tasks.
+<IMPORTANT>
+* If user provides a path, you should NOT assume it's relative to the current working directory. Instead, you should explore the file system to find the file before working on it.
+* When configuring git credentials, use "openhands" as the user.name and "openhands@all-hands.dev" as the user.email by default, unless explicitly instructed otherwise.
+* The assistant MUST NOT include comments in the code unless they are necessary to describe non-obvious behavior.
+</IMPORTANT>

agenthub/delegator_agent/agent.py

@@ -2,7 +2,7 @@ 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.events.observation import AgentDelegateObservation, Observation
 from openhands.llm.llm import LLM
 
 
@@ -41,7 +41,11 @@ class DelegatorAgent(Agent):
             )
 
         # last observation in history should be from the delegate
-        last_observation = state.history.get_last_observation()
+        last_observation = None
+        for event in reversed(state.history):
+            if isinstance(event, Observation):
+                last_observation = event
+                break
 
         if not isinstance(last_observation, AgentDelegateObservation):
             raise Exception('Last observation is not an AgentDelegateObservation')

agenthub/dummy_agent/agent.py

@@ -164,7 +164,7 @@ class DummyAgent(Agent):
 
             if 'observations' in prev_step and prev_step['observations']:
                 expected_observations = prev_step['observations']
-                hist_events = state.history.get_last_events(len(expected_observations))
+                hist_events = state.history[-len(expected_observations) :]
 
                 if len(hist_events) < len(expected_observations):
                     print(

agenthub/micro/agent.py

@@ -8,10 +8,10 @@ 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.event import Event
 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:
@@ -32,16 +32,14 @@ class MicroAgent(Agent):
     prompt = ''
     agent_definition: dict = {}
 
-    def history_to_json(
-        self, history: ShortTermHistory, max_events: int = 20, **kwargs
-    ):
+    def history_to_json(self, history: list[Event], 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):
+        for event in reversed(history):
             if event_count >= max_events:
                 break
             processed_history.append(

agenthub/planner_agent/prompt.py

@@ -117,7 +117,7 @@ def get_hint(latest_action_id: str) -> str:
 
 def get_prompt_and_images(
     state: State, max_message_chars: int
-) -> tuple[str, list[str]]:
+) -> tuple[str, list[str] | None]:
     """Gets the prompt for the planner agent.
 
     Formatted with the most recent action-observation pairs, current task, and hint based on last action
@@ -136,7 +136,7 @@ def get_prompt_and_images(
     latest_action: Action = NullAction()
 
     # retrieve the latest HISTORY_SIZE events
-    for event_count, event in enumerate(state.history.get_events(reverse=True)):
+    for event_count, event in enumerate(reversed(state.history)):
         if event_count >= HISTORY_SIZE:
             break
         if latest_action == NullAction() and isinstance(event, Action):
@@ -160,7 +160,7 @@ def get_prompt_and_images(
 
     # 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'})
+    logger.debug('HINT:\n' + hint, extra={'msg_type': 'DETAIL'})
 
     # the last relevant user message (the task)
     message, image_urls = state.get_current_user_intent()

build.sh

@@ -1,5 +0,0 @@
-#!/bin/bash
-set -e
-
-cp pyproject.toml poetry.lock openhands
-poetry build -v

compose.yml

@@ -1,22 +0,0 @@
-#
-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

@@ -1,250 +0,0 @@
-###################### 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 = ""
-
-# API key for Modal
-#modal_api_token_id = ""
-#modal_api_token_secret = ""
-
-# 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
-
-# Path to store trajectories
-#trajectories_path="./trajectories"
-
-# 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 = "local"
-
-# 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 if provided by the LLM and supported
-#caching_prompt = true
-
-# 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 = 3
-
-# LLM config group to use
-#llm_config = 'your-llm-config-group'
-
-[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:python3.12-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 = false
-
-# 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

@@ -1,12 +0,0 @@
-# Docker Containers
-
-Each folder here contains a Dockerfile, and a config.sh describing how to build
-the images and where to push them. These images are built and pushed in GitHub Actions
-by the `ghcr.yml` workflow.
-
-## Building Manually
-
-```bash
-docker build -f containers/app/Dockerfile -t openhands .
-docker build -f containers/sandbox/Dockerfile -t sandbox .
-```

containers/app/Dockerfile

@@ -1,101 +0,0 @@
-ARG OPENHANDS_BUILD_VERSION=dev
-FROM node:21.7.2-bookworm-slim AS frontend-builder
-
-WORKDIR /app
-
-COPY ./frontend/package.json frontend/package-lock.json ./
-RUN npm install -g npm@10.5.1
-RUN npm ci
-
-COPY ./frontend ./
-RUN npm run build
-
-FROM python:3.12.3-slim AS backend-builder
-
-WORKDIR /app
-ENV PYTHONPATH='/app'
-
-ENV POETRY_NO_INTERACTION=1 \
-    POETRY_VIRTUALENVS_IN_PROJECT=1 \
-    POETRY_VIRTUALENVS_CREATE=1 \
-    POETRY_CACHE_DIR=/tmp/poetry_cache
-
-RUN apt-get update -y \
-    && apt-get install -y curl make git build-essential \
-    && python3 -m pip install poetry==1.8.2  --break-system-packages
-
-COPY ./pyproject.toml ./poetry.lock ./
-RUN touch README.md
-RUN export POETRY_CACHE_DIR && poetry install --without evaluation,llama-index --no-root && rm -rf $POETRY_CACHE_DIR
-
-FROM python:3.12.3-slim AS openhands-app
-
-WORKDIR /app
-
-ARG OPENHANDS_BUILD_VERSION #re-declare for this section
-
-ENV RUN_AS_OPENHANDS=true
-# A random number--we need this to be different from the user's UID on the host machine
-ENV OPENHANDS_USER_ID=42420
-ENV SANDBOX_LOCAL_RUNTIME_URL=http://host.docker.internal
-ENV USE_HOST_NETWORK=false
-ENV WORKSPACE_BASE=/opt/workspace_base
-ENV OPENHANDS_BUILD_VERSION=$OPENHANDS_BUILD_VERSION
-RUN mkdir -p $WORKSPACE_BASE
-
-RUN apt-get update -y \
-    && apt-get install -y curl ssh sudo
-
-# Install Docker - https://docs.docker.com/engine/install/debian/
-RUN apt-get install ca-certificates curl \
-    && curl -fsSL https://download.docker.com/linux/debian/gpg -o /etc/apt/keyrings/docker.asc \
-    && chmod a+r /etc/apt/keyrings/docker.asc \
-    && echo "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/debian bookworm stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null \
-    && apt-get update \
-    && apt install -y docker-ce
-
-# 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 $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 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=openhands:app --chmod=770 --from=backend-builder ${VIRTUAL_ENV} ${VIRTUAL_ENV}
-RUN playwright install --with-deps chromium
-
-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 ./openhands/agenthub ./openhands/agenthub
-COPY --chown=openhands:app ./pyproject.toml ./pyproject.toml
-COPY --chown=openhands:app ./poetry.lock ./poetry.lock
-COPY --chown=openhands:app ./README.md ./README.md
-COPY --chown=openhands:app ./MANIFEST.in ./MANIFEST.in
-COPY --chown=openhands:app ./LICENSE ./LICENSE
-
-# 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=openhands:app --chmod=770 --from=frontend-builder /app/build ./frontend/build
-COPY --chown=openhands:app --chmod=770 ./containers/app/entrypoint.sh /app/entrypoint.sh
-
-USER root
-
-WORKDIR /app
-
-ENTRYPOINT ["/app/entrypoint.sh"]
-CMD ["uvicorn", "openhands.server.listen:app", "--host", "0.0.0.0", "--port", "3000"]

containers/app/config.sh

@@ -1,4 +0,0 @@
-DOCKER_REGISTRY=ghcr.io
-DOCKER_ORG=all-hands-ai
-DOCKER_IMAGE=openhands
-DOCKER_BASE_DIR="."

containers/app/entrypoint.sh

@@ -1,64 +0,0 @@
-#!/bin/bash
-set -eo pipefail
-
-echo "Starting OpenHands..."
-if [[ $NO_SETUP == "true" ]]; then
-  echo "Skipping setup, running as $(whoami)"
-  "$@"
-  exit 0
-fi
-
-if [ "$(id -u)" -ne 0 ]; then
-  echo "The OpenHands entrypoint.sh must run as root"
-  exit 1
-fi
-
-if [ -z "$SANDBOX_USER_ID" ]; then
-  echo "SANDBOX_USER_ID is not set"
-  exit 1
-fi
-
-if [[ "$SANDBOX_USER_ID" -eq 0 ]]; then
-  echo "Running OpenHands as root"
-  export RUN_AS_OPENHANDS=false
-  mkdir -p /root/.cache/ms-playwright/
-  if [ -d "/home/openhands/.cache/ms-playwright/" ]; then
-    mv /home/openhands/.cache/ms-playwright/ /root/.cache/
-  fi
-  "$@"
-else
-  echo "Setting up enduser with id $SANDBOX_USER_ID"
-  if id "enduser" &>/dev/null; then
-    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 $SANDBOX_USER_ID. Moving openhands user."
-      incremented_id=$(($SANDBOX_USER_ID + 1))
-      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 $SANDBOX_USER_ID for a second time. Exiting."
-        exit 1
-      fi
-    fi
-  fi
-  usermod -aG app enduser
-  # get the user group of /var/run/docker.sock and set openhands to that group
-  DOCKER_SOCKET_GID=$(stat -c '%g' /var/run/docker.sock)
-  echo "Docker socket group id: $DOCKER_SOCKET_GID"
-  if getent group $DOCKER_SOCKET_GID; then
-    echo "Group with id $DOCKER_SOCKET_GID already exists"
-  else
-    echo "Creating group with id $DOCKER_SOCKET_GID"
-    groupadd -g $DOCKER_SOCKET_GID docker
-  fi
-
-  mkdir -p /home/enduser/.cache/huggingface/hub/
-  mkdir -p /home/enduser/.cache/ms-playwright/
-  if [ -d "/home/openhands/.cache/ms-playwright/" ]; then
-    mv /home/openhands/.cache/ms-playwright/ /home/enduser/.cache/
-  fi
-
-  usermod -aG $DOCKER_SOCKET_GID enduser
-  echo "Running as enduser"
-  su enduser /bin/bash -c "${*@Q}" # This magically runs any arguments passed to the script as a command
-fi

containers/build.sh

@@ -1,156 +0,0 @@
-#!/bin/bash
-set -eo pipefail
-
-# Initialize variables with default values
-image_name=""
-org_name=""
-push=0
-load=0
-tag_suffix=""
-
-# Function to display usage information
-usage() {
-    echo "Usage: $0 -i <image_name> [-o <org_name>] [--push] [--load] [-t <tag_suffix>]"
-    echo "  -i: Image name (required)"
-    echo "  -o: Organization name"
-    echo "  --push: Push the image"
-    echo "  --load: Load the image"
-    echo "  -t: Tag suffix"
-    exit 1
-}
-
-# Parse command-line options
-while [[ $# -gt 0 ]]; do
-    case $1 in
-        -i) image_name="$2"; shift 2 ;;
-        -o) org_name="$2"; shift 2 ;;
-        --push) push=1; shift ;;
-        --load) load=1; shift ;;
-        -t) tag_suffix="$2"; shift 2 ;;
-        *) usage ;;
-    esac
-done
-# Check if required arguments are provided
-if [[ -z "$image_name" ]]; then
-    echo "Error: Image name is required."
-    usage
-fi
-
-echo "Building: $image_name"
-tags=()
-
-OPENHANDS_BUILD_VERSION="dev"
-
-cache_tag_base="buildcache"
-cache_tag="$cache_tag_base"
-
-if [[ -n $RELEVANT_SHA ]]; then
-  git_hash=$(git rev-parse --short "$RELEVANT_SHA")
-  tags+=("$git_hash")
-  tags+=("$RELEVANT_SHA")
-fi
-
-if [[ -n $GITHUB_REF_NAME ]]; then
-  # check if ref name is a version number
-  if [[ $GITHUB_REF_NAME =~ ^[0-9]+\.[0-9]+\.[0-9]+$ ]]; then
-    major_version=$(echo "$GITHUB_REF_NAME" | cut -d. -f1)
-    minor_version=$(echo "$GITHUB_REF_NAME" | cut -d. -f1,2)
-    tags+=("$major_version" "$minor_version")
-    tags+=("latest")
-  fi
-  sanitized_ref_name=$(echo "$GITHUB_REF_NAME" | sed 's/[^a-zA-Z0-9.-]\+/-/g')
-  OPENHANDS_BUILD_VERSION=$sanitized_ref_name
-  sanitized_ref_name=$(echo "$sanitized_ref_name" | tr '[:upper:]' '[:lower:]') # lower case is required in tagging
-  tags+=("$sanitized_ref_name")
-  cache_tag+="-${sanitized_ref_name}"
-fi
-
-if [[ -n $tag_suffix ]]; then
-  cache_tag+="-${tag_suffix}"
-  for i in "${!tags[@]}"; do
-    tags[$i]="${tags[$i]}-$tag_suffix"
-  done
-fi
-
-echo "Tags: ${tags[@]}"
-
-if [[ "$image_name" == "openhands" ]]; then
-  dir="./containers/app"
-elif [[ "$image_name" == "runtime" ]]; then
-  dir="./containers/runtime"
-else
-  dir="./containers/$image_name"
-fi
-
-if [[ (! -f "$dir/Dockerfile") && "$image_name" != "runtime" ]]; then
-  # Allow runtime to be built without a Dockerfile
-  echo "No Dockerfile found"
-  exit 1
-fi
-if [[ ! -f "$dir/config.sh" ]]; then
-  echo "No config.sh found for Dockerfile"
-  exit 1
-fi
-
-source "$dir/config.sh"
-
-if [[ -n "$org_name" ]]; then
-  DOCKER_ORG="$org_name"
-fi
-
-# If $DOCKER_IMAGE_HASH_TAG is set, add it to the tags
-if [[ -n "$DOCKER_IMAGE_HASH_TAG" ]]; then
-  tags+=("$DOCKER_IMAGE_HASH_TAG")
-fi
-# If $DOCKER_IMAGE_TAG is set, add it to the tags
-if [[ -n "$DOCKER_IMAGE_TAG" ]]; then
-  tags+=("$DOCKER_IMAGE_TAG")
-fi
-
-DOCKER_REPOSITORY="$DOCKER_REGISTRY/$DOCKER_ORG/$DOCKER_IMAGE"
-DOCKER_REPOSITORY=${DOCKER_REPOSITORY,,} # lowercase
-echo "Repo: $DOCKER_REPOSITORY"
-echo "Base dir: $DOCKER_BASE_DIR"
-
-args=""
-for tag in "${tags[@]}"; do
-  args+=" -t $DOCKER_REPOSITORY:$tag"
-done
-
-if [[ $push -eq 1 ]]; then
-  args+=" --push"
-  args+=" --cache-to=type=registry,ref=$DOCKER_REPOSITORY:$cache_tag,mode=max"
-fi
-
-if [[ $load -eq 1 ]]; then
-  args+=" --load"
-fi
-
-echo "Args: $args"
-
-# Modify the platform selection based on --load flag
-if [[ $load -eq 1 ]]; then
-  # When loading, build only for the current platform
-  platform=$(docker version -f '{{.Server.Os}}/{{.Server.Arch}}')
-else
-  # For push or without load, build for multiple platforms
-  platform="linux/amd64,linux/arm64"
-fi
-
-echo "Building for platform(s): $platform"
-
-docker buildx build \
-  $args \
-  --build-arg OPENHANDS_BUILD_VERSION="$OPENHANDS_BUILD_VERSION" \
-  --cache-from=type=registry,ref=$DOCKER_REPOSITORY:$cache_tag \
-  --cache-from=type=registry,ref=$DOCKER_REPOSITORY:$cache_tag_base-main \
-  --platform $platform \
-  --provenance=false \
-  -f "$dir/Dockerfile" \
-  "$DOCKER_BASE_DIR"
-
-# If load was requested, print the loaded images
-if [[ $load -eq 1 ]]; then
-  echo "Local images built:"
-  docker images "$DOCKER_REPOSITORY" --format "{{.Repository}}:{{.Tag}}"
-fi

containers/dev/Dockerfile

@@ -1,124 +0,0 @@
-# syntax=docker/dockerfile:1
-
-###
-FROM ubuntu:22.04 AS dind
-
-# https://docs.docker.com/engine/install/ubuntu/
-RUN apt-get update && apt-get install -y \
-	ca-certificates \
-	curl \
-	&& install -m 0755 -d /etc/apt/keyrings \
-	&& curl -fsSL https://download.docker.com/linux/ubuntu/gpg -o /etc/apt/keyrings/docker.asc \
-	&& chmod a+r /etc/apt/keyrings/docker.asc \
-	&& echo \
-		"deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/ubuntu \
-		$(. /etc/os-release && echo "$VERSION_CODENAME") stable" | tee /etc/apt/sources.list.d/docker.list > /dev/null
-
-RUN apt-get update && apt-get install -y \
-	docker-ce \
-	docker-ce-cli \
-	containerd.io \
-	docker-buildx-plugin \
-	docker-compose-plugin \
-	&& rm -rf /var/lib/apt/lists/* \
-	&& apt-get clean \
-	&& apt-get autoremove -y
-
-###
-FROM dind AS openhands
-
-ENV DEBIAN_FRONTEND=noninteractive
-
-#
-RUN apt-get update && apt-get install -y \
-	bash \
-    build-essential \
-    curl \
-	git \
-	git-lfs \
-	software-properties-common \
-	make \
-    netcat \
-    sudo \
-	wget \
-	&& rm -rf /var/lib/apt/lists/* \
-	&& apt-get clean \
-	&& apt-get autoremove -y
-
-# https://github.com/cli/cli/blob/trunk/docs/install_linux.md
-RUN curl -fsSL https://cli.github.com/packages/githubcli-archive-keyring.gpg | dd of=/usr/share/keyrings/githubcli-archive-keyring.gpg \
-	&& chmod go+r /usr/share/keyrings/githubcli-archive-keyring.gpg \
-	&& echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/githubcli-archive-keyring.gpg] https://cli.github.com/packages stable main" | tee /etc/apt/sources.list.d/github-cli.list > /dev/null \
-	&& apt-get update && apt-get -y install \
-    gh \
-  && rm -rf /var/lib/apt/lists/* \
-  && apt-get clean \
-  && apt-get autoremove -y
-
-# Python 3.12
-RUN add-apt-repository ppa:deadsnakes/ppa \
-    && apt-get update \
-    && apt-get install -y python3.12 python3.12-venv python3.12-dev python3-pip \
-    && ln -s /usr/bin/python3.12 /usr/bin/python
-
-# NodeJS >= 18.17.1
-RUN curl -fsSL https://deb.nodesource.com/setup_18.x | bash - \
-    && apt-get install -y nodejs
-
-# Poetry >= 1.8
-RUN curl -fsSL https://install.python-poetry.org | python3.12 - \
-    && ln -s ~/.local/bin/poetry /usr/local/bin/poetry
-
-#
-RUN <<EOF
-#!/bin/bash
-printf "#!/bin/bash
-set +x
-uname -a
-docker --version
-gh --version | head -n 1
-git --version
-#
-python --version
-echo node `node --version`
-echo npm `npm --version`
-poetry --version
-netcat -h 2>&1 | head -n 1
-" > /version.sh
-chmod a+x /version.sh
-EOF
-
-###
-FROM openhands AS dev
-
-RUN apt-get update && apt-get install -y \
-	dnsutils \
-	file \
-	iproute2 \
-	jq \
-	lsof \
-	ripgrep \
-	silversearcher-ag \
-	vim \
-	&& rm -rf /var/lib/apt/lists/* \
-	&& apt-get clean \
-	&& apt-get autoremove -y
-
-WORKDIR /app
-
-# cache build dependencies
-RUN \
-  --mount=type=bind,source=./,target=/app/ \
-  <<EOF
-#!/bin/bash
-make -s clean
-make -s check-dependencies
-make -s install-python-dependencies
-
-# NOTE
-# node_modules are .dockerignore-d therefore not mountable
-# make -s install-frontend-dependencies
-EOF
-
-#
-CMD ["bash"]

containers/dev/README.md

@@ -1,54 +0,0 @@
-# Develop in Docker
-
-Install [Docker](https://docs.docker.com/engine/install/) on your host machine and run:
-
-```bash
-make docker-dev
-# same as:
-cd ./containers/dev
-./dev.sh
-```
-
-It could take some time if you are running for the first time as Docker will pull all the  tools required for building OpenHands. The next time you run again, it should be instant.
-
-## Build and run
-
-If everything goes well, you should be inside a container after Docker finishes building the `openhands:dev` image similar to the following:
-
-```bash
-Build and run in Docker ...
-root@93fc0005fcd2:/app#
-```
-
-You may now proceed with the normal [build and run](../../Development.md) workflow as if you were on the host.
-
-## Make changes
-
-The source code on the host is mounted as `/app` inside docker. You may edit the files as usual either inside the Docker container or on your host with your favorite IDE/editors.
-
-The following are also mapped as readonly from your host:
-
-```yaml
-# host credentials
-- $HOME/.git-credentials:/root/.git-credentials:ro
-- $HOME/.gitconfig:/root/.gitconfig:ro
-- $HOME/.npmrc:/root/.npmrc:ro
-```
-
-## VSCode
-
-Alternatively, if you use VSCode, you could also [attach to the running container](https://code.visualstudio.com/docs/devcontainers/attach-container).
-
-See details for [developing in docker](https://code.visualstudio.com/docs/devcontainers/containers) or simply ask `OpenHands` ;-)
-
-## Rebuild dev image
-
-You could optionally pass additional options to the build script.
-
-```bash
-make docker-dev OPTIONS="--build"
-# or
-./containers/dev/dev.sh --build
-```
-
-See [docker compose run](https://docs.docker.com/reference/cli/docker/compose/run/) for more options.

containers/dev/compose.yml

@@ -1,38 +0,0 @@
-#
-services:
-  dev:
-    privileged: true
-    build:
-      context: ${OPENHANDS_WORKSPACE:-../../}
-      dockerfile: ./containers/dev/Dockerfile
-    image: openhands:dev
-    container_name: openhands-dev
-    environment:
-      - BACKEND_HOST=${BACKEND_HOST:-"0.0.0.0"}
-      - SANDBOX_API_HOSTNAME=host.docker.internal
-      #
-      - 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
-      # source code
-      - ${OPENHANDS_WORKSPACE:-../../}:/app
-      # host credentials
-      - $HOME/.git-credentials:/root/.git-credentials:ro
-      - $HOME/.gitconfig:/root/.gitconfig:ro
-      - $HOME/.npmrc:/root/.npmrc:ro
-      # cache
-      - cache-data:/root/.cache
-    pull_policy: never
-    stdin_open: true
-    tty: true
-
-##
-volumes:
-  cache-data:

containers/dev/dev.sh

@@ -1,39 +0,0 @@
-#!/bin/bash
-set -o pipefail
-
-function get_docker() {
-    echo "Docker is required to build and run OpenHands."
-    echo "https://docs.docker.com/get-started/get-docker/"
-    exit 1
-}
-
-function check_tools() {
-	command -v docker &>/dev/null || get_docker
-}
-
-function exit_if_indocker() {
-    if [ -f /.dockerenv ]; then
-        echo "Running inside a Docker container. Exiting..."
-        exit 1
-    fi
-}
-
-#
-exit_if_indocker
-
-check_tools
-
-##
-OPENHANDS_WORKSPACE=$(git rev-parse --show-toplevel)
-
-cd "$OPENHANDS_WORKSPACE/containers/dev/" || exit 1
-
-##
-export BACKEND_HOST="0.0.0.0"
-#
-export SANDBOX_USER_ID=$(id -u)
-export WORKSPACE_BASE=${WORKSPACE_BASE:-$OPENHANDS_WORKSPACE/workspace}
-
-docker compose run --rm --service-ports "$@" dev
-
-##

containers/e2b-sandbox/Dockerfile

@@ -1,19 +0,0 @@
-FROM ubuntu:22.04
-
-# install basic packages
-RUN apt-get update && apt-get install -y \
-    curl \
-    wget \
-    git \
-    vim \
-    nano \
-    unzip \
-    zip \
-    python3 \
-    python3-pip \
-    python3-venv \
-    python3-dev \
-    build-essential \
-    openssh-server \
-    sudo \
-    && rm -rf /var/lib/apt/lists/*

containers/e2b-sandbox/README.md

@@ -1,15 +0,0 @@
-# How to build custom E2B sandbox for OpenHands
-
-[E2B](https://e2b.dev) is an [open-source](https://github.com/e2b-dev/e2b) secure cloud environment (sandbox) made for running AI-generated code and agents. E2B offers [Python](https://pypi.org/project/e2b/) and [JS/TS](https://www.npmjs.com/package/e2b) SDK to spawn and control these sandboxes.
-
-
-1. Install the CLI with NPM.
-    ```sh
-    npm install -g @e2b/cli@latest
-    ```
-    Full CLI API is [here](https://e2b.dev/docs/cli/installation).
-
-1. Build the sandbox
-  ```sh
-  e2b template build --dockerfile ./Dockerfile --name "openhands"
-  ```

containers/e2b-sandbox/e2b.toml

@@ -1,14 +0,0 @@
-# This is a config for E2B sandbox template.
-# You can use 'template_id' (785n69crgahmz0lkdw9h) or 'template_name (openhands) from this config to spawn a sandbox:
-
-# Python SDK
-# from e2b import Sandbox
-# sandbox = Sandbox(template='openhands')
-
-# JS SDK
-# import { Sandbox } from 'e2b'
-# const sandbox = await Sandbox.create({ template: 'openhands' })
-
-dockerfile = "Dockerfile"
-template_name = "openhands"
-template_id = "785n69crgahmz0lkdw9h"

containers/runtime/README.md

@@ -1,12 +0,0 @@
-# Dynamically constructed Dockerfile
-
-This folder builds a runtime image (sandbox), which will use a dynamically generated `Dockerfile`
-that depends on the `base_image` **AND** a [Python source distribution](https://docs.python.org/3.10/distutils/sourcedist.html) that is based on the current commit of `openhands`.
-
-The following command will generate a `Dockerfile` file for `nikolaik/python-nodejs:python3.12-nodejs22` (the default base image), an updated `config.sh` and the runtime source distribution files/folders into `containers/runtime`:
-
-```bash
-poetry run python3 openhands/runtime/utils/runtime_build.py \
-    --base_image nikolaik/python-nodejs:python3.12-nodejs22 \
-    --build_folder containers/runtime
-```

containers/runtime/config.sh

@@ -1,7 +0,0 @@
-DOCKER_REGISTRY=ghcr.io
-DOCKER_ORG=all-hands-ai
-DOCKER_BASE_DIR="./containers/runtime"
-DOCKER_IMAGE=runtime
-# These variables will be appended by the runtime_build.py script
-# DOCKER_IMAGE_TAG=
-# DOCKER_IMAGE_HASH_TAG=

controller/action_parser.py

@@ -3,6 +3,16 @@ from abc import ABC, abstractmethod
 from openhands.events.action import Action
 
 
+class ActionParseError(Exception):
+    """Exception raised when the response from the LLM cannot be parsed into an action."""
+
+    def __init__(self, error: str):
+        self.error = error
+
+    def __str__(self):
+        return self.error
+
+
 class ResponseParser(ABC):
     """This abstract base class is a general interface for an response parser dedicated to
     parsing the action from the response from the LLM.

controller/agent.py

@@ -11,6 +11,7 @@ from openhands.core.exceptions import (
 )
 from openhands.llm.llm import LLM
 from openhands.runtime.plugins import PluginRequirement
+from openhands.utils.prompt import PromptManager
 
 
 class Agent(ABC):
@@ -33,6 +34,7 @@ class Agent(ABC):
         self.llm = llm
         self.config = config
         self._complete = False
+        self.prompt_manager: PromptManager | None = None
 
     @property
     def complete(self) -> bool:

controller/agent_controller.py

@@ -0,0 +1,955 @@
+import asyncio
+import copy
+import os
+import traceback
+from typing import Callable, ClassVar, Type
+
+import litellm
+from litellm.exceptions import BadRequestError, ContextWindowExceededError
+
+from openhands.controller.agent import Agent
+from openhands.controller.state.state import State, TrafficControlState
+from openhands.controller.stuck import StuckDetector
+from openhands.core.config import AgentConfig, LLMConfig
+from openhands.core.exceptions import (
+    FunctionCallNotExistsError,
+    FunctionCallValidationError,
+    LLMMalformedActionError,
+    LLMNoActionError,
+    LLMResponseError,
+)
+from openhands.core.logger import LOG_ALL_EVENTS
+from openhands.core.logger import openhands_logger as logger
+from openhands.core.schema import AgentState
+from openhands.events import EventSource, EventStream, EventStreamSubscriber
+from openhands.events.action import (
+    Action,
+    ActionConfirmationStatus,
+    AddTaskAction,
+    AgentDelegateAction,
+    AgentFinishAction,
+    AgentRejectAction,
+    ChangeAgentStateAction,
+    CmdRunAction,
+    IPythonRunCellAction,
+    MessageAction,
+    ModifyTaskAction,
+    NullAction,
+)
+from openhands.events.event import Event
+from openhands.events.observation import (
+    AgentDelegateObservation,
+    AgentStateChangedObservation,
+    ErrorObservation,
+    NullObservation,
+    Observation,
+)
+from openhands.events.serialization.event import truncate_content
+from openhands.llm.llm import LLM
+from openhands.utils.shutdown_listener import should_continue
+
+# note: RESUME is only available on web GUI
+TRAFFIC_CONTROL_REMINDER = (
+    "Please click on resume button if you'd like to continue, or start a new task."
+)
+
+
+class AgentController:
+    id: str
+    agent: Agent
+    max_iterations: int
+    event_stream: EventStream
+    state: State
+    confirmation_mode: bool
+    agent_to_llm_config: dict[str, LLMConfig]
+    agent_configs: dict[str, AgentConfig]
+    agent_task: asyncio.Future | None = None
+    parent: 'AgentController | None' = None
+    delegate: 'AgentController | None' = None
+    _pending_action: Action | None = None
+    _closed: bool = False
+    filter_out: ClassVar[tuple[type[Event], ...]] = (
+        NullAction,
+        NullObservation,
+        ChangeAgentStateAction,
+        AgentStateChangedObservation,
+    )
+
+    def __init__(
+        self,
+        agent: Agent,
+        event_stream: EventStream,
+        max_iterations: int,
+        max_budget_per_task: float | None = None,
+        agent_to_llm_config: dict[str, LLMConfig] | None = None,
+        agent_configs: dict[str, AgentConfig] | None = None,
+        sid: str = 'default',
+        confirmation_mode: bool = False,
+        initial_state: State | None = None,
+        is_delegate: bool = False,
+        headless_mode: bool = True,
+        status_callback: Callable | None = None,
+    ):
+        """Initializes a new instance of the AgentController class.
+
+        Args:
+            agent: The agent instance to control.
+            event_stream: The event stream to publish events to.
+            max_iterations: The maximum number of iterations the agent can run.
+            max_budget_per_task: The maximum budget (in USD) allowed per task, beyond which the agent will stop.
+            agent_to_llm_config: A dictionary mapping agent names to LLM configurations in the case that
+                we delegate to a different agent.
+            agent_configs: A dictionary mapping agent names to agent configurations in the case that
+                we delegate to a different agent.
+            sid: The session ID of the agent.
+            confirmation_mode: Whether to enable confirmation mode for agent actions.
+            initial_state: The initial state of the controller.
+            is_delegate: Whether this controller is a delegate.
+            headless_mode: Whether the agent is run in headless mode.
+            status_callback: Optional callback function to handle status updates.
+        """
+        self._step_lock = asyncio.Lock()
+        self.id = sid
+        self.agent = agent
+        self.headless_mode = headless_mode
+
+        # subscribe to the event stream
+        self.event_stream = event_stream
+        self.event_stream.subscribe(
+            EventStreamSubscriber.AGENT_CONTROLLER, self.on_event, self.id
+        )
+
+        # state from the previous session, state from a parent agent, or a fresh state
+        self.set_initial_state(
+            state=initial_state,
+            max_iterations=max_iterations,
+            confirmation_mode=confirmation_mode,
+        )
+        self.max_budget_per_task = max_budget_per_task
+        self.agent_to_llm_config = agent_to_llm_config if agent_to_llm_config else {}
+        self.agent_configs = agent_configs if agent_configs else {}
+        self._initial_max_iterations = max_iterations
+        self._initial_max_budget_per_task = max_budget_per_task
+
+        # stuck helper
+        self._stuck_detector = StuckDetector(self.state)
+        self.status_callback = status_callback
+
+    async def close(self) -> None:
+        """Closes the agent controller, canceling any ongoing tasks and unsubscribing from the event stream.
+
+        Note that it's fairly important that this closes properly, otherwise the state is incomplete.
+        """
+        await self.set_agent_state_to(AgentState.STOPPED)
+
+        # we made history, now is the time to rewrite it!
+        # the final state.history will be used by external scripts like evals, tests, etc.
+        # history will need to be complete WITH delegates events
+        # like the regular agent history, it does not include:
+        # - 'hidden' events, events with hidden=True
+        # - backend events (the default 'filtered out' types, types in self.filter_out)
+        start_id = self.state.start_id if self.state.start_id >= 0 else 0
+        end_id = (
+            self.state.end_id
+            if self.state.end_id >= 0
+            else self.event_stream.get_latest_event_id()
+        )
+        self.state.history = list(
+            self.event_stream.get_events(
+                start_id=start_id,
+                end_id=end_id,
+                reverse=False,
+                filter_out_type=self.filter_out,
+                filter_hidden=True,
+            )
+        )
+
+        # unsubscribe from the event stream
+        self.event_stream.unsubscribe(EventStreamSubscriber.AGENT_CONTROLLER, self.id)
+        self._closed = True
+
+    def log(self, level: str, message: str, extra: dict | None = None) -> None:
+        """Logs a message to the agent controller's logger.
+
+        Args:
+            level (str): The logging level to use (e.g., 'info', 'debug', 'error').
+            message (str): The message to log.
+            extra (dict | None, optional): Additional fields to include in the log. Defaults to None.
+        """
+        message = f'[Agent Controller {self.id}] {message}'
+        getattr(logger, level)(message, extra=extra, stacklevel=2)
+
+    def update_state_before_step(self):
+        self.state.iteration += 1
+        self.state.local_iteration += 1
+
+    async def update_state_after_step(self):
+        # update metrics especially for cost. Use deepcopy to avoid it being modified by agent._reset()
+        self.state.local_metrics = copy.deepcopy(self.agent.llm.metrics)
+
+    async def _react_to_exception(
+        self,
+        e: Exception,
+    ):
+        await self.set_agent_state_to(AgentState.ERROR)
+        if self.status_callback is not None:
+            err_id = ''
+            if isinstance(e, litellm.AuthenticationError):
+                err_id = 'STATUS$ERROR_LLM_AUTHENTICATION'
+            self.status_callback('error', err_id, str(e))
+
+    async def start_step_loop(self):
+        """The main loop for the agent's step-by-step execution."""
+        self.log('info', 'Starting step loop...')
+        while True:
+            if not self._is_awaiting_observation() and not should_continue():
+                break
+            if self._closed:
+                break
+            try:
+                await self._step()
+            except asyncio.CancelledError:
+                self.log('debug', 'AgentController task was cancelled')
+                break
+            except Exception as e:
+                traceback.print_exc()
+                self.log('error', f'Error while running the agent: {e}')
+                await self._react_to_exception(e)
+
+            await asyncio.sleep(0.1)
+
+    async def on_event(self, event: Event) -> None:
+        """Callback from the event stream. Notifies the controller of incoming events.
+
+        Args:
+            event (Event): The incoming event to process.
+        """
+        if hasattr(event, 'hidden') and event.hidden:
+            return
+
+        # if the event is not filtered out, add it to the history
+        if not any(isinstance(event, filter_type) for filter_type in self.filter_out):
+            self.state.history.append(event)
+
+        if isinstance(event, Action):
+            await self._handle_action(event)
+        elif isinstance(event, Observation):
+            await self._handle_observation(event)
+
+    async def _handle_action(self, action: Action) -> None:
+        """Handles actions from the event stream.
+
+        Args:
+            action (Action): The action to handle.
+        """
+        if isinstance(action, ChangeAgentStateAction):
+            await self.set_agent_state_to(action.agent_state)  # type: ignore
+        elif isinstance(action, MessageAction):
+            await self._handle_message_action(action)
+        elif isinstance(action, AgentDelegateAction):
+            await self.start_delegate(action)
+        elif isinstance(action, AddTaskAction):
+            self.state.root_task.add_subtask(
+                action.parent, action.goal, action.subtasks
+            )
+        elif isinstance(action, ModifyTaskAction):
+            self.state.root_task.set_subtask_state(action.task_id, action.state)
+        elif isinstance(action, AgentFinishAction):
+            self.state.outputs = action.outputs
+            self.state.metrics.merge(self.state.local_metrics)
+            await self.set_agent_state_to(AgentState.FINISHED)
+        elif isinstance(action, AgentRejectAction):
+            self.state.outputs = action.outputs
+            self.state.metrics.merge(self.state.local_metrics)
+            await self.set_agent_state_to(AgentState.REJECTED)
+
+    async def _handle_observation(self, observation: Observation) -> None:
+        """Handles observation from the event stream.
+
+        Args:
+            observation (observation): The observation to handle.
+        """
+        observation_to_print = copy.deepcopy(observation)
+        if len(observation_to_print.content) > self.agent.llm.config.max_message_chars:
+            observation_to_print.content = truncate_content(
+                observation_to_print.content, self.agent.llm.config.max_message_chars
+            )
+        # Use info level if LOG_ALL_EVENTS is set
+        log_level = 'info' if os.getenv('LOG_ALL_EVENTS') in ('true', '1') else 'debug'
+        self.log(
+            log_level, str(observation_to_print), extra={'msg_type': 'OBSERVATION'}
+        )
+
+        if observation.llm_metrics is not None:
+            self.agent.llm.metrics.merge(observation.llm_metrics)
+
+        if self._pending_action and self._pending_action.id == observation.cause:
+            if self.state.agent_state == AgentState.AWAITING_USER_CONFIRMATION:
+                return
+            self._pending_action = None
+            if self.state.agent_state == AgentState.USER_CONFIRMED:
+                await self.set_agent_state_to(AgentState.RUNNING)
+            if self.state.agent_state == AgentState.USER_REJECTED:
+                await self.set_agent_state_to(AgentState.AWAITING_USER_INPUT)
+            return
+        elif isinstance(observation, ErrorObservation):
+            if self.state.agent_state == AgentState.ERROR:
+                self.state.metrics.merge(self.state.local_metrics)
+
+    async def _handle_message_action(self, action: MessageAction) -> None:
+        """Handles message actions from the event stream.
+
+        Args:
+            action (MessageAction): The message action to handle.
+        """
+        if action.source == EventSource.USER:
+            # Use info level if LOG_ALL_EVENTS is set
+            log_level = (
+                'info' if os.getenv('LOG_ALL_EVENTS') in ('true', '1') else 'debug'
+            )
+            self.log(
+                log_level,
+                str(action),
+                extra={'msg_type': 'ACTION', 'event_source': EventSource.USER},
+            )
+            # Extend max iterations when the user sends a message (only in non-headless mode)
+            if self._initial_max_iterations is not None and not self.headless_mode:
+                self.state.max_iterations = (
+                    self.state.iteration + self._initial_max_iterations
+                )
+                if (
+                    self.state.traffic_control_state == TrafficControlState.THROTTLING
+                    or self.state.traffic_control_state == TrafficControlState.PAUSED
+                ):
+                    self.state.traffic_control_state = TrafficControlState.NORMAL
+                self.log(
+                    'debug',
+                    f'Extended max iterations to {self.state.max_iterations} after user message',
+                )
+            if self.get_agent_state() != AgentState.RUNNING:
+                await self.set_agent_state_to(AgentState.RUNNING)
+        elif action.source == EventSource.AGENT and action.wait_for_response:
+            await self.set_agent_state_to(AgentState.AWAITING_USER_INPUT)
+
+    def _reset(self) -> None:
+        """Resets the agent controller"""
+
+        self._pending_action = None
+        self.agent.reset()
+
+    async def set_agent_state_to(self, new_state: AgentState) -> None:
+        """Updates the agent's state and handles side effects. Can emit events to the event stream.
+
+        Args:
+            new_state (AgentState): The new state to set for the agent.
+        """
+        self.log(
+            'info',
+            f'Setting agent({self.agent.name}) state from {self.state.agent_state} to {new_state}',
+        )
+
+        if new_state == self.state.agent_state:
+            return
+
+        if new_state in (AgentState.STOPPED, AgentState.ERROR):
+            self._reset()
+        elif (
+            new_state == AgentState.RUNNING
+            and self.state.agent_state == AgentState.PAUSED
+            # TODO: do we really need both THROTTLING and PAUSED states, or can we clean up one of them completely?
+            and self.state.traffic_control_state == TrafficControlState.THROTTLING
+        ):
+            # user intends to interrupt traffic control and let the task resume temporarily
+            self.state.traffic_control_state = TrafficControlState.PAUSED
+            # User has chosen to deliberately continue - lets double the max iterations
+            if (
+                self.state.iteration is not None
+                and self.state.max_iterations is not None
+                and self._initial_max_iterations is not None
+                and not self.headless_mode
+            ):
+                if self.state.iteration >= self.state.max_iterations:
+                    self.state.max_iterations += self._initial_max_iterations
+
+            if (
+                self.state.metrics.accumulated_cost is not None
+                and self.max_budget_per_task is not None
+                and self._initial_max_budget_per_task is not None
+            ):
+                if self.state.metrics.accumulated_cost >= self.max_budget_per_task:
+                    self.max_budget_per_task += self._initial_max_budget_per_task
+        elif self._pending_action is not None and (
+            new_state in (AgentState.USER_CONFIRMED, AgentState.USER_REJECTED)
+        ):
+            if hasattr(self._pending_action, 'thought'):
+                self._pending_action.thought = ''  # type: ignore[union-attr]
+            if new_state == AgentState.USER_CONFIRMED:
+                confirmation_state = ActionConfirmationStatus.CONFIRMED
+            else:
+                confirmation_state = ActionConfirmationStatus.REJECTED
+            self._pending_action.confirmation_state = confirmation_state  # type: ignore[attr-defined]
+            self._pending_action._id = None  # type: ignore[attr-defined]
+            self.event_stream.add_event(self._pending_action, EventSource.AGENT)
+
+        self.state.agent_state = new_state
+        self.event_stream.add_event(
+            AgentStateChangedObservation('', self.state.agent_state),
+            EventSource.ENVIRONMENT,
+        )
+
+        if new_state == AgentState.INIT and self.state.resume_state:
+            await self.set_agent_state_to(self.state.resume_state)
+            self.state.resume_state = None
+
+    def get_agent_state(self) -> AgentState:
+        """Returns the current state of the agent.
+
+        Returns:
+            AgentState: The current state of the agent.
+        """
+        return self.state.agent_state
+
+    async def start_delegate(self, action: AgentDelegateAction) -> None:
+        """Start a delegate agent to handle a subtask.
+
+        OpenHands is a multi-agentic system. A `task` is a 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 a 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`. Otherwise, a `task` consists of
+        multiple `subtasks`, each executed by one agent.
+
+        Args:
+            action (AgentDelegateAction): The action containing information about the delegate agent to start.
+        """
+        agent_cls: Type[Agent] = Agent.get_cls(action.agent)
+        agent_config = self.agent_configs.get(action.agent, self.agent.config)
+        llm_config = self.agent_to_llm_config.get(action.agent, self.agent.llm.config)
+        llm = LLM(config=llm_config)
+        delegate_agent = agent_cls(llm=llm, config=agent_config)
+        state = State(
+            inputs=action.inputs or {},
+            local_iteration=0,
+            iteration=self.state.iteration,
+            max_iterations=self.state.max_iterations,
+            delegate_level=self.state.delegate_level + 1,
+            # global metrics should be shared between parent and child
+            metrics=self.state.metrics,
+            # start on top of the stream
+            start_id=self.event_stream.get_latest_event_id() + 1,
+        )
+        self.log(
+            'debug',
+            f'start delegate, creating agent {delegate_agent.name} using LLM {llm}',
+        )
+
+        self.event_stream.unsubscribe(EventStreamSubscriber.AGENT_CONTROLLER, self.id)
+        self.delegate = AgentController(
+            sid=self.id + '-delegate',
+            agent=delegate_agent,
+            event_stream=self.event_stream,
+            max_iterations=self.state.max_iterations,
+            max_budget_per_task=self.max_budget_per_task,
+            agent_to_llm_config=self.agent_to_llm_config,
+            agent_configs=self.agent_configs,
+            initial_state=state,
+            is_delegate=True,
+            headless_mode=self.headless_mode,
+        )
+        await self.delegate.set_agent_state_to(AgentState.RUNNING)
+
+    async def _step(self) -> None:
+        """Executes a single step of the parent or delegate agent. Detects stuck agents and limits on the number of iterations and the task budget."""
+        if self.get_agent_state() != AgentState.RUNNING:
+            await asyncio.sleep(1)
+            return
+
+        if self._pending_action:
+            await asyncio.sleep(1)
+            return
+
+        if self.delegate is not None:
+            assert self.delegate != self
+            if self.delegate.get_agent_state() == AgentState.PAUSED:
+                # no need to check too often
+                await asyncio.sleep(1)
+            else:
+                await self._delegate_step()
+            return
+
+        self.log(
+            'info',
+            f'LEVEL {self.state.delegate_level} LOCAL STEP {self.state.local_iteration} GLOBAL STEP {self.state.iteration}',
+            extra={'msg_type': 'STEP'},
+        )
+
+        # check if agent hit the resources limit
+        stop_step = False
+        if self.state.iteration >= self.state.max_iterations:
+            stop_step = await self._handle_traffic_control(
+                'iteration', self.state.iteration, self.state.max_iterations
+            )
+        if self.max_budget_per_task is not None:
+            current_cost = self.state.metrics.accumulated_cost
+            if current_cost > self.max_budget_per_task:
+                stop_step = await self._handle_traffic_control(
+                    'budget', current_cost, self.max_budget_per_task
+                )
+        if stop_step:
+            return
+
+        if self._is_stuck():
+            await self._react_to_exception(RuntimeError('Agent got stuck in a loop'))
+            return
+
+        self.update_state_before_step()
+        action: Action = NullAction()
+        try:
+            action = self.agent.step(self.state)
+            if action is None:
+                raise LLMNoActionError('No action was returned')
+        except (
+            LLMMalformedActionError,
+            LLMNoActionError,
+            LLMResponseError,
+            FunctionCallValidationError,
+            FunctionCallNotExistsError,
+        ) as e:
+            self.event_stream.add_event(
+                ErrorObservation(
+                    content=str(e),
+                ),
+                EventSource.AGENT,
+            )
+            return
+        except (ContextWindowExceededError, BadRequestError) as e:
+            # FIXME: this is a hack until a litellm fix is confirmed
+            # Check if this is a nested context window error
+            error_str = str(e).lower()
+            if (
+                'contextwindowexceedederror' in error_str
+                or 'prompt is too long' in error_str
+                or isinstance(e, ContextWindowExceededError)
+            ):
+                # When context window is exceeded, keep roughly half of agent interactions
+                self.state.history = self._apply_conversation_window(self.state.history)
+
+                # Save the ID of the first event in our truncated history for future reloading
+                if self.state.history:
+                    self.state.start_id = self.state.history[0].id
+                # Don't add error event - let the agent retry with reduced context
+                return
+            raise
+
+        if action.runnable:
+            if self.state.confirmation_mode and (
+                type(action) is CmdRunAction or type(action) is IPythonRunCellAction
+            ):
+                action.confirmation_state = (
+                    ActionConfirmationStatus.AWAITING_CONFIRMATION
+                )
+            self._pending_action = action
+
+        if not isinstance(action, NullAction):
+            if (
+                hasattr(action, 'confirmation_state')
+                and action.confirmation_state
+                == ActionConfirmationStatus.AWAITING_CONFIRMATION
+            ):
+                await self.set_agent_state_to(AgentState.AWAITING_USER_CONFIRMATION)
+            self.event_stream.add_event(action, EventSource.AGENT)
+
+        await self.update_state_after_step()
+
+        log_level = 'info' if LOG_ALL_EVENTS else 'debug'
+        self.log(log_level, str(action), extra={'msg_type': 'ACTION'})
+
+    async def _delegate_step(self) -> None:
+        """Executes a single step of the delegate agent."""
+        await self.delegate._step()  # type: ignore[union-attr]
+        assert self.delegate is not None
+        delegate_state = self.delegate.get_agent_state()
+        self.log('debug', f'Delegate state: {delegate_state}')
+        if delegate_state == AgentState.ERROR:
+            # update iteration that shall be shared across agents
+            self.state.iteration = self.delegate.state.iteration
+
+            # emit AgentDelegateObservation to mark delegate termination due to error
+            delegate_outputs = (
+                self.delegate.state.outputs if self.delegate.state else {}
+            )
+            content = (
+                f'{self.delegate.agent.name} encountered an error during execution.'
+            )
+            obs = AgentDelegateObservation(outputs=delegate_outputs, content=content)
+            self.event_stream.add_event(obs, EventSource.AGENT)
+
+            # close the delegate upon error
+            await self.delegate.close()
+
+            # resubscribe parent when delegate is finished
+            self.event_stream.subscribe(
+                EventStreamSubscriber.AGENT_CONTROLLER, self.on_event, self.id
+            )
+            self.delegate = None
+            self.delegateAction = None
+
+        elif delegate_state in (AgentState.FINISHED, AgentState.REJECTED):
+            self.log('debug', 'Delegate agent has finished execution')
+            # retrieve delegate result
+            outputs = self.delegate.state.outputs if self.delegate.state else {}
+
+            # update iteration that shall be shared across agents
+            self.state.iteration = self.delegate.state.iteration
+
+            # close delegate controller: we must close the delegate controller before adding new events
+            await self.delegate.close()
+
+            # resubscribe parent when delegate is finished
+            self.event_stream.subscribe(
+                EventStreamSubscriber.AGENT_CONTROLLER, self.on_event, self.id
+            )
+
+            # update delegate result observation
+            # TODO: replace this with AI-generated summary (#2395)
+            formatted_output = ', '.join(
+                f'{key}: {value}' for key, value in outputs.items()
+            )
+            content = (
+                f'{self.delegate.agent.name} finishes task with {formatted_output}'
+            )
+            obs = AgentDelegateObservation(outputs=outputs, content=content)
+
+            # clean up delegate status
+            self.delegate = None
+            self.delegateAction = None
+            self.event_stream.add_event(obs, EventSource.AGENT)
+        return
+
+    async def _handle_traffic_control(
+        self, limit_type: str, current_value: float, max_value: float
+    ) -> bool:
+        """Handles agent state after hitting the traffic control limit.
+
+        Args:
+            limit_type (str): The type of limit that was hit.
+            current_value (float): The current value of the limit.
+            max_value (float): The maximum value of the limit.
+        """
+        stop_step = False
+        if self.state.traffic_control_state == TrafficControlState.PAUSED:
+            self.log(
+                'debug', 'Hitting traffic control, temporarily resume upon user request'
+            )
+            self.state.traffic_control_state = TrafficControlState.NORMAL
+        else:
+            self.state.traffic_control_state = TrafficControlState.THROTTLING
+            # Format values as integers for iterations, keep decimals for budget
+            if limit_type == 'iteration':
+                current_str = str(int(current_value))
+                max_str = str(int(max_value))
+            else:
+                current_str = f'{current_value:.2f}'
+                max_str = f'{max_value:.2f}'
+
+            if self.headless_mode:
+                e = RuntimeError(
+                    f'Agent reached maximum {limit_type} in headless mode. '
+                    f'Current {limit_type}: {current_str}, max {limit_type}: {max_str}'
+                )
+                await self._react_to_exception(e)
+            else:
+                e = RuntimeError(
+                    f'Agent reached maximum {limit_type}. '
+                    f'Current {limit_type}: {current_str}, max {limit_type}: {max_str}. '
+                )
+                # FIXME: this isn't really an exception--we should have a different path
+                await self._react_to_exception(e)
+            stop_step = True
+        return stop_step
+
+    def get_state(self) -> State:
+        """Returns the current running state object.
+
+        Returns:
+            State: The current state object.
+        """
+        return self.state
+
+    def set_initial_state(
+        self,
+        state: State | None,
+        max_iterations: int,
+        confirmation_mode: bool = False,
+    ) -> None:
+        """Sets the initial state for the agent, either from the previous session, or from a parent agent, or by creating a new one.
+
+        Args:
+            state: The state to initialize with, or None to create a new state.
+            max_iterations: The maximum number of iterations allowed for the task.
+            confirmation_mode: Whether to enable confirmation mode.
+        """
+        # state can come from:
+        # - the previous session, in which case it has history
+        # - from a parent agent, in which case it has no history
+        # - None / a new state
+        if state is None:
+            self.state = State(
+                inputs={},
+                max_iterations=max_iterations,
+                confirmation_mode=confirmation_mode,
+            )
+        else:
+            self.state = state
+
+            if self.state.start_id <= -1:
+                self.state.start_id = 0
+
+            self.log(
+                'debug',
+                f'AgentController {self.id} initializing history from event {self.state.start_id}',
+            )
+
+            self._init_history()
+
+    def _init_history(self) -> None:
+        """Initializes the agent's history from the event stream.
+
+        The history is a list of events that:
+        - Excludes events of types listed in self.filter_out
+        - Excludes events with hidden=True attribute
+        - For delegate events (between AgentDelegateAction and AgentDelegateObservation):
+            - Excludes all events between the action and observation
+            - Includes the delegate action and observation themselves
+
+        The history is loaded in two parts if truncation_id is set:
+        1. First user message from start_id onwards
+        2. Rest of history from truncation_id to the end
+
+        Otherwise loads normally from start_id.
+        """
+        # define range of events to fetch
+        # delegates start with a start_id and initially won't find any events
+        # otherwise we're restoring a previous session
+        start_id = self.state.start_id if self.state.start_id >= 0 else 0
+        end_id = (
+            self.state.end_id
+            if self.state.end_id >= 0
+            else self.event_stream.get_latest_event_id()
+        )
+
+        # sanity check
+        if start_id > end_id + 1:
+            self.log(
+                'warning',
+                f'start_id {start_id} is greater than end_id + 1 ({end_id + 1}). History will be empty.',
+            )
+            self.state.history = []
+            return
+
+        events: list[Event] = []
+
+        # If we have a truncation point, get first user message and then rest of history
+        if hasattr(self.state, 'truncation_id') and self.state.truncation_id > 0:
+            # Find first user message from stream
+            first_user_msg = next(
+                (
+                    e
+                    for e in self.event_stream.get_events(
+                        start_id=start_id,
+                        end_id=end_id,
+                        reverse=False,
+                        filter_out_type=self.filter_out,
+                        filter_hidden=True,
+                    )
+                    if isinstance(e, MessageAction) and e.source == EventSource.USER
+                ),
+                None,
+            )
+            if first_user_msg:
+                events.append(first_user_msg)
+
+            # the rest of the events are from the truncation point
+            start_id = self.state.truncation_id
+
+        # Get rest of history
+        events_to_add = list(
+            self.event_stream.get_events(
+                start_id=start_id,
+                end_id=end_id,
+                reverse=False,
+                filter_out_type=self.filter_out,
+                filter_hidden=True,
+            )
+        )
+        events.extend(events_to_add)
+
+        # Find all delegate action/observation pairs
+        delegate_ranges: list[tuple[int, int]] = []
+        delegate_action_ids: list[int] = []  # stack of unmatched delegate action IDs
+
+        for event in events:
+            if isinstance(event, AgentDelegateAction):
+                delegate_action_ids.append(event.id)
+                # Note: we can get agent=event.agent and task=event.inputs.get('task','')
+                # if we need to track these in the future
+
+            elif isinstance(event, AgentDelegateObservation):
+                # Match with most recent unmatched delegate action
+                if not delegate_action_ids:
+                    self.log(
+                        'warning',
+                        f'Found AgentDelegateObservation without matching action at id={event.id}',
+                    )
+                    continue
+
+                action_id = delegate_action_ids.pop()
+                delegate_ranges.append((action_id, event.id))
+
+        # Filter out events between delegate action/observation pairs
+        if delegate_ranges:
+            filtered_events: list[Event] = []
+            current_idx = 0
+
+            for start_id, end_id in sorted(delegate_ranges):
+                # Add events before delegate range
+                filtered_events.extend(
+                    event for event in events[current_idx:] if event.id < start_id
+                )
+
+                # Add delegate action and observation
+                filtered_events.extend(
+                    event for event in events if event.id in (start_id, end_id)
+                )
+
+                # Update index to after delegate range
+                current_idx = next(
+                    (i for i, e in enumerate(events) if e.id > end_id), len(events)
+                )
+
+            # Add any remaining events after last delegate range
+            filtered_events.extend(events[current_idx:])
+
+            self.state.history = filtered_events
+        else:
+            self.state.history = events
+
+        # make sure history is in sync
+        self.state.start_id = start_id
+
+    def _apply_conversation_window(self, events: list[Event]) -> list[Event]:
+        """Cuts history roughly in half when context window is exceeded, preserving action-observation pairs
+        and ensuring the first user message is always included.
+
+        The algorithm:
+        1. Cut history in half
+        2. Check first event in new history:
+           - If Observation: find and include its Action
+           - If MessageAction: ensure its related Action-Observation pair isn't split
+        3. Always include the first user message
+
+        Args:
+            events: List of events to filter
+
+        Returns:
+            Filtered list of events keeping newest half while preserving pairs
+        """
+        if not events:
+            return events
+
+        # Find first user message - we'll need to ensure it's included
+        first_user_msg = next(
+            (
+                e
+                for e in events
+                if isinstance(e, MessageAction) and e.source == EventSource.USER
+            ),
+            None,
+        )
+
+        # cut in half
+        mid_point = max(1, len(events) // 2)
+        kept_events = events[mid_point:]
+
+        # Handle first event in truncated history
+        if kept_events:
+            i = 0
+            while i < len(kept_events):
+                first_event = kept_events[i]
+                if isinstance(first_event, Observation) and first_event.cause:
+                    # Find its action and include it
+                    matching_action = next(
+                        (
+                            e
+                            for e in reversed(events[:mid_point])
+                            if isinstance(e, Action) and e.id == first_event.cause
+                        ),
+                        None,
+                    )
+                    if matching_action:
+                        kept_events = [matching_action] + kept_events
+                    else:
+                        self.log(
+                            'warning',
+                            f'Found Observation without matching Action at id={first_event.id}',
+                        )
+                        # drop this observation
+                        kept_events = kept_events[1:]
+                    break
+
+                elif isinstance(first_event, MessageAction) or (
+                    isinstance(first_event, Action)
+                    and first_event.source == EventSource.USER
+                ):
+                    # if it's a message action or a user action, keep it and continue to find the next event
+                    i += 1
+                    continue
+
+                else:
+                    # if it's an action with source == EventSource.AGENT, we're good
+                    break
+
+        # Save where to continue from in next reload
+        if kept_events:
+            self.state.truncation_id = kept_events[0].id
+
+        # Ensure first user message is included
+        if first_user_msg and first_user_msg not in kept_events:
+            kept_events = [first_user_msg] + kept_events
+
+        # start_id points to first user message
+        if first_user_msg:
+            self.state.start_id = first_user_msg.id
+
+        return kept_events
+
+    def _is_stuck(self) -> bool:
+        """Checks if the agent or its delegate is stuck in a loop.
+
+        Returns:
+            bool: True if the agent is stuck, False otherwise.
+        """
+        # check if delegate stuck
+        if self.delegate and self.delegate._is_stuck():
+            return True
+
+        return self._stuck_detector.is_stuck(self.headless_mode)
+
+    def __repr__(self):
+        return (
+            f'AgentController(id={self.id}, agent={self.agent!r}, '
+            f'event_stream={self.event_stream!r}, '
+            f'state={self.state!r}, agent_task={self.agent_task!r}, '
+            f'delegate={self.delegate!r}, _pending_action={self._pending_action!r})'
+        )
+
+    def _is_awaiting_observation(self):
+        events = self.event_stream.get_events(reverse=True)
+        for event in events:
+            if isinstance(event, AgentStateChangedObservation):
+                result = event.agent_state == AgentState.RUNNING
+                return result
+        return False

controller/state/state.py

@@ -6,13 +6,13 @@ from typing import Any
 
 from openhands.controller.state.task import RootTask
 from openhands.core.logger import openhands_logger as logger
-from openhands.core.metrics import Metrics
 from openhands.core.schema import AgentState
 from openhands.events.action import (
     MessageAction,
 )
 from openhands.events.action.agent import AgentFinishAction
-from openhands.memory.history import ShortTermHistory
+from openhands.events.event import Event, EventSource
+from openhands.llm.metrics import Metrics
 from openhands.storage.files import FileStore
 
 
@@ -77,10 +77,9 @@ class State:
     # max number of iterations for the current task
     max_iterations: int = 100
     confirmation_mode: bool = False
-    history: ShortTermHistory = field(default_factory=ShortTermHistory)
+    history: list[Event] = field(default_factory=list)
     inputs: dict = field(default_factory=dict)
     outputs: dict = field(default_factory=dict)
-    last_error: str | None = None
     agent_state: AgentState = AgentState.LOADING
     resume_state: AgentState | None = None
     traffic_control_state: TrafficControlState = TrafficControlState.NORMAL
@@ -93,10 +92,14 @@ class State:
     # start_id and end_id track the range of events in history
     start_id: int = -1
     end_id: int = -1
-    almost_stuck: int = 0
+    # truncation_id tracks where to load history after context window truncation
+    truncation_id: int = -1
+
+    delegates: dict[tuple[int, int], tuple[str, str]] = field(default_factory=dict)
     # NOTE: This will never be used by the controller, but it can be used by different
     # evaluation tasks to store extra data needed to track the progress/state of the task.
     extra_data: dict[str, Any] = field(default_factory=dict)
+    last_error: str = ''
 
     def save_to_session(self, sid: str, file_store: FileStore):
         pickled = pickle.dumps(self)
@@ -115,7 +118,7 @@ class State:
             pickled = base64.b64decode(encoded)
             state = pickle.loads(pickled)
         except Exception as e:
-            logger.error(f'Failed to restore state from session: {e}')
+            logger.debug(f'Could not restore state from session: {e}')
             raise e
 
         # update state
@@ -124,49 +127,45 @@ class State:
         else:
             state.resume_state = None
 
-        # don't carry last_error anymore after restore
-        state.last_error = None
-
         # first state after restore
         state.agent_state = AgentState.LOADING
         return state
 
     def __getstate__(self):
+        # don't pickle history, it will be restored from the event stream
         state = self.__dict__.copy()
-
-        # save the relevant data from recent history
-        # so that we can restore it when the state is restored
-        if 'history' in state:
-            state['start_id'] = state['history'].start_id
-            state['end_id'] = state['history'].end_id
-
-        # don't save history object itself
-        state.pop('history', None)
+        state['history'] = []
         return state
 
     def __setstate__(self, state):
         self.__dict__.update(state)
 
-        # recreate the history object
+        # make sure we always have the attribute history
         if not hasattr(self, 'history'):
-            self.history = ShortTermHistory()
+            self.history = []
 
-        # restore the relevant data in history from the state
-        self.history.start_id = self.start_id
-        self.history.end_id = self.end_id
-
-        # remove the restored data from the state if any
-
-    def get_current_user_intent(self):
+    def get_current_user_intent(self) -> tuple[str | None, list[str] | None]:
         """Returns the latest user message and image(if provided) that appears after a FinishAction, or the first (the task) if nothing was finished yet."""
         last_user_message = None
         last_user_message_image_urls: list[str] | None = []
-        for event in self.history.get_events(reverse=True):
+        for event in reversed(self.history):
             if isinstance(event, MessageAction) and event.source == 'user':
                 last_user_message = event.content
-                last_user_message_image_urls = event.images_urls
+                last_user_message_image_urls = event.image_urls
             elif isinstance(event, AgentFinishAction):
                 if last_user_message is not None:
-                    return last_user_message
+                    return last_user_message, None
 
         return last_user_message, last_user_message_image_urls
+
+    def get_last_agent_message(self) -> MessageAction | None:
+        for event in reversed(self.history):
+            if isinstance(event, MessageAction) and event.source == EventSource.AGENT:
+                return event
+        return None
+
+    def get_last_user_message(self) -> MessageAction | None:
+        for event in reversed(self.history):
+            if isinstance(event, MessageAction) and event.source == EventSource.USER:
+                return event
+        return None

controller/stuck.py

@@ -24,17 +24,44 @@ class StuckDetector:
     def __init__(self, state: State):
         self.state = state
 
-    def is_stuck(self):
-        # filter out MessageAction with source='user' from history
+    def is_stuck(self, headless_mode: bool = True):
+        """Checks if the agent is stuck in a loop.
+
+        Args:
+            headless_mode: Matches AgentController's headless_mode.
+                          If True: Consider all history (automated/testing)
+                          If False: Consider only history after last user message (interactive)
+
+        Returns:
+            bool: True if the agent is stuck in a loop, False otherwise.
+        """
+        if not headless_mode:
+            # In interactive mode, only look at history after the last user message
+            last_user_msg_idx = -1
+            for i, event in enumerate(reversed(self.state.history)):
+                if (
+                    isinstance(event, MessageAction)
+                    and event.source == EventSource.USER
+                ):
+                    last_user_msg_idx = len(self.state.history) - i - 1
+                    break
+
+            history_to_check = self.state.history[last_user_msg_idx + 1 :]
+        else:
+            # In headless mode, look at all history
+            history_to_check = self.state.history
+
+        # Filter out user messages and null events
         filtered_history = [
             event
-            for event in self.state.history.get_events()
+            for event in history_to_check
             if not (
+                # Filter works elegantly in both modes:
+                # - In headless: actively filters out user messages from full history
+                # - In non-headless: no-op since we already sliced after last user message
                 (isinstance(event, MessageAction) and event.source == EventSource.USER)
-                or
                 # there might be some NullAction or NullObservation in the history at least for now
-                isinstance(event, NullAction)
-                or isinstance(event, NullObservation)
+                or isinstance(event, (NullAction, NullObservation))
             )
         ]
 
@@ -82,43 +109,19 @@ class StuckDetector:
         # it takes 4 actions and 4 observations to detect a loop
         # assert len(last_actions) == 4 and len(last_observations) == 4
 
-        # reset almost_stuck reminder
-        self.state.almost_stuck = 0
-
-        # almost stuck? if two actions, obs are the same, we're almost stuck
-        if len(last_actions) >= 2 and len(last_observations) >= 2:
+        # Check for a loop of 4 identical action-observation pairs
+        if len(last_actions) == 4 and len(last_observations) == 4:
             actions_equal = all(
-                self._eq_no_pid(last_actions[0], action) for action in last_actions[:2]
+                self._eq_no_pid(last_actions[0], action) for action in last_actions
             )
             observations_equal = all(
                 self._eq_no_pid(last_observations[0], observation)
-                for observation in last_observations[:2]
+                for observation in last_observations
             )
 
-            # the last two actions and obs are the same?
             if actions_equal and observations_equal:
-                self.state.almost_stuck = 2
-
-            # the last three actions and observations are the same?
-            if len(last_actions) >= 3 and len(last_observations) >= 3:
-                if (
-                    actions_equal
-                    and observations_equal
-                    and self._eq_no_pid(last_actions[0], last_actions[2])
-                    and self._eq_no_pid(last_observations[0], last_observations[2])
-                ):
-                    self.state.almost_stuck = 1
-
-            if len(last_actions) == 4 and len(last_observations) == 4:
-                if (
-                    actions_equal
-                    and observations_equal
-                    and self._eq_no_pid(last_actions[0], last_actions[3])
-                    and self._eq_no_pid(last_observations[0], last_observations[3])
-                ):
-                    logger.warning('Action, Observation loop detected')
-                    self.state.almost_stuck = 0
-                    return True
+                logger.warning('Action, Observation loop detected')
+                return True
 
         return False
 

core/cli.py

@@ -1,6 +1,8 @@
 import asyncio
 import logging
+import sys
 from typing import Type
+from uuid import uuid4
 
 from termcolor import colored
 
@@ -9,26 +11,33 @@ from openhands import __version__
 from openhands.controller import AgentController
 from openhands.controller.agent import Agent
 from openhands.core.config import (
+    AppConfig,
     get_parser,
     load_app_config,
 )
 from openhands.core.logger import openhands_logger as logger
+from openhands.core.loop import run_agent_until_done
 from openhands.core.schema import AgentState
 from openhands.events import EventSource, EventStream, EventStreamSubscriber
 from openhands.events.action import (
     Action,
+    ActionConfirmationStatus,
     ChangeAgentStateAction,
     CmdRunAction,
+    FileEditAction,
     MessageAction,
 )
 from openhands.events.event import Event
 from openhands.events.observation import (
     AgentStateChangedObservation,
     CmdOutputObservation,
+    FileEditObservation,
+    NullObservation,
 )
 from openhands.llm.llm import LLM
 from openhands.runtime import get_runtime_cls
-from openhands.runtime.runtime import Runtime
+from openhands.runtime.base import Runtime
+from openhands.security import SecurityAnalyzer, options
 from openhands.storage import get_file_store
 
 
@@ -40,6 +49,15 @@ def display_command(command: str):
     print('❯ ' + colored(command + '\n', 'green'))
 
 
+def display_confirmation(confirmation_state: ActionConfirmationStatus):
+    if confirmation_state == ActionConfirmationStatus.CONFIRMED:
+        print(colored('✅ ' + confirmation_state + '\n', 'green'))
+    elif confirmation_state == ActionConfirmationStatus.REJECTED:
+        print(colored('❌ ' + confirmation_state + '\n', 'red'))
+    else:
+        print(colored('⏳ ' + confirmation_state + '\n', 'yellow'))
+
+
 def display_command_output(output: str):
     lines = output.split('\n')
     for line in lines:
@@ -50,17 +68,27 @@ def display_command_output(output: str):
     print('\n')
 
 
-def display_event(event: Event):
+def display_file_edit(event: FileEditAction | FileEditObservation):
+    print(colored(str(event), 'green'))
+
+
+def display_event(event: Event, config: AppConfig):
     if isinstance(event, Action):
         if hasattr(event, 'thought'):
             display_message(event.thought)
     if isinstance(event, MessageAction):
-        if event.source != EventSource.USER:
+        if event.source == EventSource.AGENT:
             display_message(event.content)
     if isinstance(event, CmdRunAction):
         display_command(event.command)
     if isinstance(event, CmdOutputObservation):
         display_command_output(event.content)
+    if isinstance(event, FileEditAction):
+        display_file_edit(event)
+    if isinstance(event, FileEditObservation):
+        display_file_edit(event)
+    if hasattr(event, 'confirmation_state') and config.security.confirmation_mode:
+        display_confirmation(event.confirmation_state)
 
 
 async def main():
@@ -103,57 +131,101 @@ async def main():
         event_stream=event_stream,
         sid=sid,
         plugins=agent_cls.sandbox_plugins,
+        headless_mode=True,
     )
 
+    if config.security.security_analyzer:
+        options.SecurityAnalyzers.get(
+            config.security.security_analyzer, SecurityAnalyzer
+        )(event_stream)
+
     controller = AgentController(
         agent=agent,
         max_iterations=config.max_iterations,
         max_budget_per_task=config.max_budget_per_task,
         agent_to_llm_config=config.get_agent_to_llm_config_map(),
         event_stream=event_stream,
+        confirmation_mode=config.security.confirmation_mode,
     )
 
-    if controller is not None:
-        controller.agent_task = asyncio.create_task(controller.start_step_loop())
-
     async def prompt_for_next_task():
-        next_message = input('How can I help? >> ')
+        # Run input() in a thread pool to avoid blocking the event loop
+        loop = asyncio.get_event_loop()
+        next_message = await loop.run_in_executor(
+            None, lambda: input('How can I help? >> ')
+        )
+        if not next_message.strip():
+            await prompt_for_next_task()
         if next_message == 'exit':
             event_stream.add_event(
-                ChangeAgentStateAction(AgentState.STOPPED), EventSource.USER
+                ChangeAgentStateAction(AgentState.STOPPED), EventSource.ENVIRONMENT
             )
             return
         action = MessageAction(content=next_message)
         event_stream.add_event(action, EventSource.USER)
 
+    async def prompt_for_user_confirmation():
+        loop = asyncio.get_event_loop()
+        user_confirmation = await loop.run_in_executor(
+            None, lambda: input('Confirm action (possible security risk)? (y/n) >> ')
+        )
+        return user_confirmation.lower() == 'y'
+
     async def on_event(event: Event):
-        display_event(event)
+        display_event(event, config)
         if isinstance(event, AgentStateChangedObservation):
-            if event.agent_state == AgentState.ERROR:
-                print('An error occurred. Please try again.')
             if event.agent_state in [
                 AgentState.AWAITING_USER_INPUT,
                 AgentState.FINISHED,
-                AgentState.ERROR,
             ]:
                 await prompt_for_next_task()
+        if (
+            isinstance(event, NullObservation)
+            and controller.state.agent_state == AgentState.AWAITING_USER_CONFIRMATION
+        ):
+            user_confirmed = await prompt_for_user_confirmation()
+            if user_confirmed:
+                event_stream.add_event(
+                    ChangeAgentStateAction(AgentState.USER_CONFIRMED), EventSource.USER
+                )
+            else:
+                event_stream.add_event(
+                    ChangeAgentStateAction(AgentState.USER_REJECTED), EventSource.USER
+                )
 
-    event_stream.subscribe(EventStreamSubscriber.MAIN, on_event)
+    event_stream.subscribe(EventStreamSubscriber.MAIN, on_event, str(uuid4()))
 
-    await prompt_for_next_task()
+    await runtime.connect()
 
-    while controller.state.agent_state not in [
-        AgentState.STOPPED,
-    ]:
-        await asyncio.sleep(1)  # Give back control for a tick, so the agent can run
+    asyncio.create_task(prompt_for_next_task())
 
-    print('Exiting...')
-    await controller.close()
+    await run_agent_until_done(
+        controller, runtime, [AgentState.STOPPED, AgentState.ERROR]
+    )
 
 
 if __name__ == '__main__':
-    loop = asyncio.get_event_loop()
+    loop = asyncio.new_event_loop()
+    asyncio.set_event_loop(loop)
     try:
         loop.run_until_complete(main())
+    except KeyboardInterrupt:
+        print('Received keyboard interrupt, shutting down...')
+    except ConnectionRefusedError as e:
+        print(f'Connection refused: {e}')
+        sys.exit(1)
+    except Exception as e:
+        print(f'An error occurred: {e}')
+        sys.exit(1)
     finally:
-        pass
+        try:
+            # Cancel all running tasks
+            pending = asyncio.all_tasks(loop)
+            for task in pending:
+                task.cancel()
+            # Wait for all tasks to complete with a timeout
+            loop.run_until_complete(asyncio.gather(*pending, return_exceptions=True))
+            loop.close()
+        except Exception as e:
+            print(f'Error during cleanup: {e}')
+            sys.exit(1)

core/config/__init__.py

@@ -3,7 +3,6 @@ from openhands.core.config.app_config import AppConfig
 from openhands.core.config.config_utils import (
     OH_DEFAULT_AGENT,
     OH_MAX_ITERATIONS,
-    UndefinedString,
     get_field_info,
 )
 from openhands.core.config.llm_config import LLMConfig
@@ -22,7 +21,6 @@ from openhands.core.config.utils import (
 __all__ = [
     'OH_DEFAULT_AGENT',
     'OH_MAX_ITERATIONS',
-    'UndefinedString',
     'AgentConfig',
     'AppConfig',
     'LLMConfig',