Fix pr #8247: Simplify delegate agents prompt and finish

Co-authored-by: openhands <openhands@all-hands.dev>
Co-authored-by: Rohit Malhotra <rohitvinodmalhotra@gmail.com>

.github/ISSUE_TEMPLATE/feature_request.md

@@ -12,3 +12,6 @@ assignees: ''
 **Describe the UX or technical implementation you have in mind**
 
 **Additional context**
+
+
+### If you find this feature request or enhancement useful, make sure to add a 👍 to the issue

.github/pull_request_template.md

@@ -1,12 +1,12 @@
 - [ ] This change is worth documenting at https://docs.all-hands.dev/
 - [ ] Include this change in the Release Notes. If checked, you **must** provide an **end-user friendly** description for your change below
 
-**End-user friendly description of the problem this fixes or functionality that this introduces.**
+**End-user friendly description of the problem this fixes or functionality this introduces.**
 
 
 ---
-**Give a summary of what the PR does, explaining any non-trivial design decisions.**
+**Summarize what the PR does, explaining any non-trivial design decisions.**
 
 
 ---
-**Link of any specific issues this addresses.**
+**Link of any specific issues this addresses:**

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

@@ -1,53 +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: blacksmith-4vcpu-ubuntu-2204
-    steps:
-      - uses: actions/checkout@v4
-      - name: Set up Docker Buildx
-        id: buildx
-        uses: docker/setup-buildx-action@v3
-      - name: Install tmux
-        run: sudo apt-get update && sudo apt-get install -y tmux
-      - name: Setup Node.js
-        uses: useblacksmith/setup-node@v5
-        with:
-          node-version: '22.x'
-      - name: Install poetry via pipx
-        run: pipx install poetry
-      - name: Set up Python
-        uses: useblacksmith/setup-python@v6
-        with:
-          python-version: '3.12'
-          cache: 'poetry'
-      - name: Install Python dependencies using Poetry
-        run: poetry install --without evaluation
-      - 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/ghcr-build.yml

@@ -174,20 +174,19 @@ jobs:
           build-args: ${{ env.DOCKER_BUILD_ARGS }}
           context: containers/runtime
           provenance: false
-      # Forked repos can't push to GHCR, so we need to upload the image as an artifact
+      # Forked repos can't push to GHCR, so we just build in order to populate the cache for rebuilding
       - name: Build runtime image ${{ matrix.base_image.image }} for fork
         if: github.event.pull_request.head.repo.fork
         uses: useblacksmith/build-push-action@v1
         with:
           tags: ghcr.io/${{ env.REPO_OWNER }}/runtime:${{ env.RELEVANT_SHA }}-${{ matrix.base_image.tag }}
-          outputs: type=docker,dest=/tmp/runtime-${{ matrix.base_image.tag }}.tar
           context: containers/runtime
-      - name: Upload runtime image for fork
+      - name: Upload runtime source for fork
         if: github.event.pull_request.head.repo.fork
         uses: actions/upload-artifact@v4
         with:
-          name: runtime-${{ matrix.base_image.tag }}
-          path: /tmp/runtime-${{ matrix.base_image.tag }}.tar
+          name: runtime-src-${{ matrix.base_image.tag }}
+          path: containers/runtime
 
   verify_hash_equivalence_in_runtime_and_app:
     name: Verify Hash Equivalence in Runtime and Docker images
@@ -258,17 +257,23 @@ jobs:
       - 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
+      - name: Download runtime source for fork
         if: github.event.pull_request.head.repo.fork
         uses: actions/download-artifact@v4
         with:
-          name: runtime-${{ matrix.base_image.tag }}
-          path: /tmp
-      - name: Load runtime image for fork
-        if: github.event.pull_request.head.repo.fork
+          name: runtime-src-${{ matrix.base_image.tag }}
+          path: containers/runtime
+      - name: Lowercase Repository Owner
         run: |
-          docker load --input /tmp/runtime-${{ matrix.base_image.tag }}.tar
+          echo REPO_OWNER=$(echo ${{ github.repository_owner }} | tr '[:upper:]' '[:lower:]') >> $GITHUB_ENV
+      # Forked repos can't push to GHCR, so we need to rebuild using cache
+      - name: Build runtime image ${{ matrix.base_image.image }} for fork
+        if: github.event.pull_request.head.repo.fork
+        uses: useblacksmith/build-push-action@v1
+        with:
+          load: true
+          tags: ghcr.io/${{ env.REPO_OWNER }}/runtime:${{ env.RELEVANT_SHA }}-${{ matrix.base_image.tag }}
+          context: containers/runtime
       - name: Cache Poetry dependencies
         uses: useblacksmith/cache@v5
         with:
@@ -287,9 +292,6 @@ jobs:
         run: pipx install poetry
       - name: Install Python dependencies using Poetry
         run: make install-python-dependencies POETRY_GROUP=main,test,runtime INSTALL_PLAYWRIGHT=0
-      - name: Lowercase Repository Owner
-        run: |
-          echo REPO_OWNER=$(echo ${{ github.repository_owner }} | tr '[:upper:]' '[:lower:]') >> $GITHUB_ENV
       - name: Run docker runtime tests
         run: |
           # We install pytest-xdist in order to run tests across CPUs
@@ -327,17 +329,23 @@ jobs:
       - name: Set up Docker Buildx
         id: buildx
         uses: docker/setup-buildx-action@v3
-      # Forked repos can't push to GHCR, so we need to download the image as an artifact
-      - name: Download runtime image for fork
+      - name: Download runtime source for fork
         if: github.event.pull_request.head.repo.fork
         uses: actions/download-artifact@v4
         with:
-          name: runtime-${{ matrix.base_image.tag }}
-          path: /tmp
-      - name: Load runtime image for fork
-        if: github.event.pull_request.head.repo.fork
+          name: runtime-src-${{ matrix.base_image.tag }}
+          path: containers/runtime
+      - name: Lowercase Repository Owner
         run: |
-          docker load --input /tmp/runtime-${{ matrix.base_image.tag }}.tar
+          echo REPO_OWNER=$(echo ${{ github.repository_owner }} | tr '[:upper:]' '[:lower:]') >> $GITHUB_ENV
+      # Forked repos can't push to GHCR, so we need to rebuild using cache
+      - name: Build runtime image ${{ matrix.base_image.image }} for fork
+        if: github.event.pull_request.head.repo.fork
+        uses: useblacksmith/build-push-action@v1
+        with:
+          load: true
+          tags: ghcr.io/${{ env.REPO_OWNER }}/runtime:${{ env.RELEVANT_SHA }}-${{ matrix.base_image.tag }}
+          context: containers/runtime
       - name: Cache Poetry dependencies
         uses: useblacksmith/cache@v5
         with:
@@ -356,9 +364,6 @@ jobs:
         run: pipx install poetry
       - name: Install Python dependencies using Poetry
         run: make install-python-dependencies POETRY_GROUP=main,test,runtime INSTALL_PLAYWRIGHT=0
-      - name: Lowercase Repository Owner
-        run: |
-          echo REPO_OWNER=$(echo ${{ github.repository_owner }} | tr '[:upper:]' '[:lower:]') >> $GITHUB_ENV
       - name: Run runtime tests
         run: |
           # We install pytest-xdist in order to run tests across CPUs

.github/workflows/openhands-resolver.yml

@@ -16,10 +16,15 @@ on:
         type: string
         default: "main"
         description: "Target branch to pull and create PR against"
+      pr_type:
+        required: false
+        type: string
+        default: "draft"
+        description: "The PR type that is going to be created (draft, ready)"
       LLM_MODEL:
         required: false
         type: string
-        default: "anthropic/claude-3-5-sonnet-20241022"
+        default: "anthropic/claude-3-7-sonnet-20250219"
       LLM_API_VERSION:
         required: false
         type: string
@@ -174,7 +179,7 @@ jobs:
 
           echo "MAX_ITERATIONS=${{ inputs.max_iterations || 50 }}" >> $GITHUB_ENV
           echo "SANDBOX_ENV_GITHUB_TOKEN=${{ secrets.PAT_TOKEN || github.token }}" >> $GITHUB_ENV
-          echo "SANDBOX_ENV_BASE_CONTAINER_IMAGE=${{ inputs.base_container_image }}" >> $GITHUB_ENV
+          echo "SANDBOX_BASE_CONTAINER_IMAGE=${{ inputs.base_container_image }}" >> $GITHUB_ENV
 
           # Set branch variables
           echo "TARGET_BRANCH=${{ inputs.target_branch || 'main' }}" >> $GITHUB_ENV
@@ -280,9 +285,9 @@ jobs:
             cd /tmp && python -m openhands.resolver.send_pull_request \
               --issue-number ${{ env.ISSUE_NUMBER }} \
               --target-branch ${{ env.TARGET_BRANCH }} \
-              --pr-type draft \
+              --pr-type ${{ inputs.pr_type || 'draft' }} \
               --reviewer ${{ github.actor }} | tee pr_result.txt && \
-              grep "draft created" pr_result.txt | sed 's/.*\///g' > pr_number.txt
+              grep "PR created" pr_result.txt | sed 's/.*\///g' > pr_number.txt
           else
             cd /tmp && python -m openhands.resolver.send_pull_request \
               --issue-number ${{ env.ISSUE_NUMBER }} \

.openhands/microagents/repo.md

@@ -54,3 +54,20 @@ Frontend:
 ## Template for Github Pull Request
 
 If you are starting a pull request (PR), please follow the template in `.github/pull_request_template.md`.
+
+## Implementation Details
+
+These details may or may not be useful for your current task.
+
+### Frontend
+
+#### Action Handling:
+- Actions are defined in `frontend/src/types/action-type.ts`
+- The `HANDLED_ACTIONS` array in `frontend/src/state/chat-slice.ts` determines which actions are displayed as collapsible UI elements
+- To add a new action type to the UI:
+  1. Add the action type to the `HANDLED_ACTIONS` array
+  2. Implement the action handling in `addAssistantAction` function in chat-slice.ts
+  3. Add a translation key in the format `ACTION_MESSAGE$ACTION_NAME` to the i18n files
+- Actions with `thought` property are displayed in the UI based on their action type:
+  - Regular actions (like "run", "edit") display the thought as a separate message
+  - Special actions (like "think") are displayed as collapsible elements only

Development.md

@@ -118,7 +118,7 @@ poetry run pytest ./tests/unit/test_*.py
 To reduce build time (e.g., if no changes were made to the client-runtime component), you can use an existing Docker container image by
 setting the SANDBOX_RUNTIME_CONTAINER_IMAGE environment variable to the desired Docker image.
 
-Example: `export SANDBOX_RUNTIME_CONTAINER_IMAGE=ghcr.io/all-hands-ai/runtime:0.32-nikolaik`
+Example: `export SANDBOX_RUNTIME_CONTAINER_IMAGE=ghcr.io/all-hands-ai/runtime:0.36-nikolaik`
 
 ## Develop inside Docker container
 

Makefile

@@ -39,6 +39,7 @@ ifeq ($(INSTALL_DOCKER),)
 	@$(MAKE) -s check-docker
 endif
 	@$(MAKE) -s check-poetry
+	@$(MAKE) -s check-tmux
 	@echo "$(GREEN)Dependencies checked successfully.$(RESET)"
 
 check-system:
@@ -101,6 +102,18 @@ check-docker:
 		exit 1; \
 	fi
 
+check-tmux:
+	@echo "$(YELLOW)Checking tmux installation...$(RESET)"
+	@if command -v tmux > /dev/null; then \
+		echo "$(BLUE)$(shell tmux -V) is already installed.$(RESET)"; \
+	else \
+		echo "$(YELLOW)╔════════════════════════════════════════════════════════════════════════════╗$(RESET)"; \
+		echo "$(YELLOW)║ OPTIONAL: tmux is not installed.                                          ║$(RESET)"; \
+		echo "$(YELLOW)║ Some advanced terminal features may not work without tmux.                ║$(RESET)"; \
+		echo "$(YELLOW)║ You can install it if needed, but it's not required for development.      ║$(RESET)"; \
+		echo "$(YELLOW)╚════════════════════════════════════════════════════════════════════════════╝$(RESET)"; \
+	fi
+
 check-poetry:
 	@echo "$(YELLOW)Checking Poetry installation...$(RESET)"
 	@if command -v poetry > /dev/null; then \
@@ -175,7 +188,7 @@ install-pre-commit-hooks:
 
 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)
+	@poetry run pre-commit run --files openhands/**/* evaluation/**/* tests/**/* --show-diff-on-failure --config $(PRE_COMMIT_CONFIG_PATH)
 
 lint-frontend:
 	@echo "$(YELLOW)Running linters for frontend...$(RESET)"

README.md

@@ -9,7 +9,6 @@
 <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/openhands-ai/shared_invite/zt-2ngejmfw6-9gW4APWOC9XUp1n~SiQ6iw"><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>
@@ -18,7 +17,7 @@
   <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>
+  <a href="https://docs.google.com/spreadsheets/d/1wOUdFCMyY6Nt0AIqF705KN4JKOWgeI4wUGUP60krXXs/edit?gid=0#gid=0"><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>
 
@@ -27,7 +26,7 @@ Welcome to OpenHands (formerly OpenDevin), a platform for software development a
 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).
+Learn more at [docs.all-hands.dev](https://docs.all-hands.dev), or [sign up for OpenHands Cloud](https://app.all-hands.dev) to get started.
 
 > [!IMPORTANT]
 > Using OpenHands for work? We'd love to chat! Fill out
@@ -36,37 +35,50 @@ Learn more at [docs.all-hands.dev](https://docs.all-hands.dev), or jump to the [
 
 ![App screenshot](./docs/static/img/screenshot.png)
 
-## ⚡ Quick Start
+## ☁️ OpenHands Cloud
+The easiest way to get started with OpenHands is on [OpenHands Cloud](https://app.all-hands.dev),
+which comes with $50 in free credits for new users.
 
-The easiest way to run OpenHands is in Docker.
+## 💻 Running OpenHands Locally
+
+OpenHands can also run on your local system using Docker.
 See the [Running OpenHands](https://docs.all-hands.dev/modules/usage/installation) guide for
 system requirements and more information.
 
+> [!WARNING]
+> On a public network? See our [Hardened Docker Installation Guide](https://docs.all-hands.dev/modules/usage/runtimes/docker#hardened-docker-installation)
+> to secure your deployment by restricting network binding and implementing additional security measures.
+
+
 ```bash
-docker pull docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik
+docker pull docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik
 
 docker run -it --rm --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e LOG_ALL_EVENTS=true \
     -v /var/run/docker.sock:/var/run/docker.sock \
     -v ~/.openhands-state:/.openhands-state \
     -p 3000:3000 \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32
+    docker.all-hands.dev/all-hands-ai/openhands:0.36
 ```
 
-> [!WARNING]
-> On a public network? See our [Hardened Docker Installation](https://docs.all-hands.dev/modules/usage/runtimes/docker#hardened-docker-installation) guide
-> to secure your deployment by restricting network binding and implementing additional security measures.
-
 You'll find OpenHands running at [http://localhost:3000](http://localhost:3000)!
 
-Finally, you'll need a model provider and API key.
-[Anthropic's Claude 3.5 Sonnet](https://www.anthropic.com/api) (`anthropic/claude-3-5-sonnet-20241022`)
+When you open the application, you'll be asked to choose an LLM provider and add an API key.
+[Anthropic's Claude 3.7 Sonnet](https://www.anthropic.com/api) (`anthropic/claude-3-7-sonnet-20250219`)
 works best, but you have [many options](https://docs.all-hands.dev/modules/usage/llms).
 
----
+## 💡 Other ways to run OpenHands
+
+> [!CAUTION]
+> OpenHands is meant to be run by a single user on their local workstation.
+> It is not appropriate for multi-tenant deployments where multiple users share the same instance. There is no built-in authentication, isolation, or scalability.
+>
+> If you're interested in running OpenHands in a multi-tenant environment, please
+> [get in touch with us](https://docs.google.com/forms/d/e/1FAIpQLSet3VbGaz8z32gW9Wm-Grl4jpt5WgMXPgJ4EDPVmCETCBpJtQ/viewform)
+> for advanced deployment options.
 
 You can also [connect OpenHands to your local filesystem](https://docs.all-hands.dev/modules/usage/runtimes/docker#connecting-to-your-filesystem),
 run OpenHands in a scriptable [headless mode](https://docs.all-hands.dev/modules/usage/how-to/headless-mode),
@@ -75,14 +87,6 @@ or run it on tagged issues with [a github action](https://docs.all-hands.dev/mod
 
 Visit [Running OpenHands](https://docs.all-hands.dev/modules/usage/installation) for more information and setup instructions.
 
-> [!CAUTION]
-> OpenHands is meant to be run by a single user on their local workstation.
-> It is not appropriate for multi-tenant deployments where multiple users share the same instance. There is no built-in isolation or scalability.
->
-> If you're interested in running OpenHands in a multi-tenant environment, please
-> [get in touch with us](https://docs.google.com/forms/d/e/1FAIpQLSet3VbGaz8z32gW9Wm-Grl4jpt5WgMXPgJ4EDPVmCETCBpJtQ/viewform)
-> for advanced deployment options.
-
 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.

config.template.toml

@@ -216,13 +216,26 @@ model = "gpt-4o"
 [agent]
 
 # Whether the browsing tool is enabled
-codeact_enable_browsing = true
+enable_browsing = true
 
 # Whether the LLM draft editor is enabled
-codeact_enable_llm_editor = false
+enable_llm_editor = false
+
+# Whether the standard editor tool (str_replace_editor) is enabled
+# Only has an effect if enable_llm_editor is False
+enable_editor = true
 
 # Whether the IPython tool is enabled
-codeact_enable_jupyter = true
+enable_jupyter = true
+
+# Whether the command tool is enabled
+enable_cmd = true
+
+# Whether the think tool is enabled
+enable_think = true
+
+# Whether the finish tool is enabled
+enable_finish = true
 
 # LLM config group to use
 #llm_config = 'your-llm-config-group'
@@ -378,7 +391,7 @@ type = "noop"
 #[llm.condenser]
 #model = "gpt-4o"
 #temperature = 0.1
-#max_tokens = 1024
+#max_input_tokens = 1024
 
 #################################### Eval ####################################
 # Configuration for the evaluation, please refer to the specific evaluation

containers/dev/Dockerfile

@@ -61,8 +61,8 @@ RUN add-apt-repository ppa:deadsnakes/ppa \
     && 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 - \
+# NodeJS >= 22.x
+RUN curl -fsSL https://deb.nodesource.com/setup_22.x | bash - \
     && apt-get install -y nodejs
 
 # Poetry >= 1.8
@@ -108,7 +108,7 @@ WORKDIR /app
 
 # cache build dependencies
 RUN \
-  --mount=type=bind,source=./,target=/app/ \
+  --mount=type=bind,source=./,target=/app/,rw \
   <<EOF
 #!/bin/bash
 make -s clean

containers/dev/compose.yml

@@ -11,7 +11,7 @@ services:
       - 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:0.32-nikolaik}
+      - SANDBOX_RUNTIME_CONTAINER_IMAGE=${SANDBOX_RUNTIME_CONTAINER_IMAGE:-ghcr.io/all-hands-ai/runtime:0.36-nikolaik}
       - SANDBOX_USER_ID=${SANDBOX_USER_ID:-1234}
       - WORKSPACE_MOUNT_PATH=${WORKSPACE_BASE:-$PWD/workspace}
     ports:

docker-compose.yml

@@ -7,7 +7,7 @@ services:
     image: openhands:latest
     container_name: openhands-app-${DATE:-}
     environment:
-      - SANDBOX_RUNTIME_CONTAINER_IMAGE=${SANDBOX_RUNTIME_CONTAINER_IMAGE:-docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik}
+      - SANDBOX_RUNTIME_CONTAINER_IMAGE=${SANDBOX_RUNTIME_CONTAINER_IMAGE:-docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik}
       #- SANDBOX_USER_ID=${SANDBOX_USER_ID:-1234} # enable this only if you want a specific non-root sandbox user but you will have to manually adjust permissions of openhands-state for this user
       - WORKSPACE_MOUNT_PATH=${WORKSPACE_BASE:-$PWD/workspace}
     ports:

docs/.gitignore

@@ -3,6 +3,7 @@
 
 # Production
 /build
+/static/swagger-ui
 
 # Generated files
 .docusaurus

docs/docusaurus.config.ts

@@ -36,6 +36,14 @@ const config: Config = {
     mermaid: true,
   },
   themes: ['@docusaurus/theme-mermaid'],
+  plugins: [
+    [
+      require.resolve('docusaurus-lunr-search'),
+      {
+        languages: ['en', 'zh', 'fr', 'ja', 'pt']
+      }
+    ]
+  ],
   presets: [
     [
       'classic',
@@ -75,10 +83,19 @@ const config: Config = {
           position: 'left',
           label: 'User Guides',
         },
+        {
+          href: 'https://docs.all-hands.dev/swagger-ui/', // FIXME: this should be a relative path, but docusarus steals the click
+          label: 'API',
+          position: 'left',
+        },
         {
           type: 'localeDropdown',
           position: 'left',
         },
+        {
+          type: 'search',
+          position: 'left',
+        },
         {
           href: 'https://all-hands.dev',
           label: 'Company',

docs/generate-swagger-ui.js

@@ -0,0 +1,102 @@
+const fs = require('fs');
+const path = require('path');
+const swaggerUiDist = require('swagger-ui-dist');
+
+/**
+ * This script manually sets up Swagger UI for the Docusaurus documentation.
+ *
+ * Why we need this approach:
+ * 1. Docusaurus doesn't have a built-in way to integrate Swagger UI
+ * 2. We need to copy the necessary files from swagger-ui-dist to our static directory
+ * 3. We need to create a custom index.html file that points to our OpenAPI spec
+ * 4. This approach allows us to customize the Swagger UI to match our documentation style
+ */
+
+// Get the absolute path to the swagger-ui-dist package
+const swaggerUiDistPath = swaggerUiDist.getAbsoluteFSPath();
+
+// Create the target directory if it doesn't exist
+const targetDir = path.join(__dirname, 'static', 'swagger-ui');
+if (!fs.existsSync(targetDir)) {
+  fs.mkdirSync(targetDir, { recursive: true });
+}
+
+// Copy all files from swagger-ui-dist to our target directory
+const files = fs.readdirSync(swaggerUiDistPath);
+files.forEach(file => {
+  const sourcePath = path.join(swaggerUiDistPath, file);
+  const targetPath = path.join(targetDir, file);
+
+  // Skip directories and non-essential files
+  if (fs.statSync(sourcePath).isDirectory() ||
+      file === 'package.json' ||
+      file === 'README.md' ||
+      file.endsWith('.map')) {
+    return;
+  }
+
+  fs.copyFileSync(sourcePath, targetPath);
+});
+
+// Create a custom index.html file that points to our OpenAPI spec
+const indexHtml = `
+<!DOCTYPE html>
+<html lang="en">
+<head>
+  <meta charset="UTF-8">
+  <title>OpenHands API Documentation</title>
+  <link rel="stylesheet" type="text/css" href="./swagger-ui.css" />
+  <link rel="icon" type="image/png" href="./favicon-32x32.png" sizes="32x32" />
+  <link rel="icon" type="image/png" href="./favicon-16x16.png" sizes="16x16" />
+  <style>
+    html {
+      box-sizing: border-box;
+      overflow: -moz-scrollbars-vertical;
+      overflow-y: scroll;
+    }
+
+    *,
+    *:before,
+    *:after {
+      box-sizing: inherit;
+    }
+
+    body {
+      margin: 0;
+      background: #fafafa;
+    }
+  </style>
+</head>
+
+<body>
+  <div id="swagger-ui"></div>
+
+  <script src="./swagger-ui-bundle.js" charset="UTF-8"> </script>
+  <script src="./swagger-ui-standalone-preset.js" charset="UTF-8"> </script>
+  <script>
+    window.onload = function() {
+      // Begin Swagger UI call region
+      const ui = SwaggerUIBundle({
+        url: "/openapi.json",
+        dom_id: '#swagger-ui',
+        deepLinking: true,
+        presets: [
+          SwaggerUIBundle.presets.apis,
+          SwaggerUIStandalonePreset
+        ],
+        plugins: [
+          SwaggerUIBundle.plugins.DownloadUrl
+        ],
+        layout: "StandaloneLayout"
+      });
+      // End Swagger UI call region
+      window.ui = ui;
+    };
+  </script>
+</body>
+</html>
+`;
+
+fs.writeFileSync(path.join(targetDir, 'index.html'), indexHtml);
+
+console.log('Swagger UI files generated successfully in static/swagger-ui/');

docs/i18n/fr/docusaurus-plugin-content-docs/current/usage/configuration-options.md

@@ -354,12 +354,12 @@ Les options de configuration de l'agent sont définies dans les sections `[agent
   - Valeur par défaut : `true`
   - Description : Si l'appel de fonction est activé
 
-- `codeact_enable_browsing`
+- `enable_browsing`
   - Type : `bool`
   - Valeur par défaut : `false`
   - Description : Si le délégué de navigation est activé dans l'espace d'action (fonctionne uniquement avec l'appel de fonction)
 
-- `codeact_enable_llm_editor`
+- `enable_llm_editor`
   - Type : `bool`
   - Valeur par défaut : `false`
   - Description : Si l'éditeur LLM est activé dans l'espace d'action (fonctionne uniquement avec l'appel de fonction)

docs/i18n/fr/docusaurus-plugin-content-docs/current/usage/how-to/cli-mode.md

@@ -52,7 +52,7 @@ LLM_API_KEY="sk_test_12345"
 ```bash
 docker run -it \
     --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e SANDBOX_USER_ID=$(id -u) \
     -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
     -e LLM_API_KEY=$LLM_API_KEY \
@@ -61,7 +61,7 @@ docker run -it \
     -v /var/run/docker.sock:/var/run/docker.sock \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app-$(date +%Y%m%d%H%M%S) \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32 \
+    docker.all-hands.dev/all-hands-ai/openhands:0.36 \
     python -m openhands.core.cli
 ```
 

docs/i18n/fr/docusaurus-plugin-content-docs/current/usage/how-to/gui-mode.md

@@ -21,14 +21,18 @@ OpenHands fournit un mode Interface Graphique (GUI) convivial pour interagir ave
 3. Entrez la `Clé API` correspondante pour le fournisseur choisi.
 4. Cliquez sur "Enregistrer" pour appliquer les paramètres.
 
-### Configuration du Jeton GitHub
+### Jetons de Contrôle de Version
+
+OpenHands prend en charge plusieurs fournisseurs de contrôle de version. Vous pouvez configurer des jetons pour plusieurs fournisseurs simultanément.
+
+#### Configuration du Jeton GitHub
 
 OpenHands exporte automatiquement un `GITHUB_TOKEN` vers l'environnement shell s'il est disponible. Cela peut se produire de deux manières :
 
 1. **Localement (OSS)** : L'utilisateur saisit directement son jeton GitHub
 2. **En ligne (SaaS)** : Le jeton est obtenu via l'authentification OAuth GitHub
 
-#### Configuration d'un Jeton GitHub Local
+##### Configuration d'un Jeton GitHub Local
 
 1. **Générer un Personal Access Token (PAT)** :
    - Allez dans Paramètres GitHub > Paramètres développeur > Personal Access Tokens > Tokens (classique)
@@ -40,11 +44,11 @@ OpenHands exporte automatiquement un `GITHUB_TOKEN` vers l'environnement shell s
 
 2. **Entrer le Jeton dans OpenHands** :
    - Cliquez sur le bouton Paramètres (icône d'engrenage) en haut à droite
-   - Accédez à la section "GitHub"
+   - Accédez à la section "Git Provider Settings"
    - Collez votre jeton dans le champ "Jeton GitHub"
    - Cliquez sur "Enregistrer" pour appliquer les modifications
 
-#### Politiques de Jetons Organisationnels
+##### Politiques de Jetons Organisationnels
 
 Si vous travaillez avec des dépôts organisationnels, une configuration supplémentaire peut être nécessaire :
 
@@ -59,7 +63,7 @@ Si vous travaillez avec des dépôts organisationnels, une configuration supplé
    - Si nécessaire, cliquez sur "Activer SSO" à côté de votre organisation
    - Terminez le processus d'autorisation SSO
 
-#### Authentification OAuth (Mode En Ligne)
+##### Authentification OAuth (Mode En Ligne)
 
 Lorsque vous utilisez OpenHands en mode en ligne, le flux OAuth GitHub :
 
@@ -74,7 +78,7 @@ Lorsque vous utilisez OpenHands en mode en ligne, le flux OAuth GitHub :
    - Autorisez OpenHands à accéder à votre compte GitHub
    - Si vous utilisez une organisation, autorisez l'accès à l'organisation si vous y êtes invité
 
-#### Dépannage
+##### Dépannage
 
 Problèmes courants et solutions :
 
@@ -95,6 +99,43 @@ Problèmes courants et solutions :
    - Vérifiez la console du navigateur pour tout message d'erreur
    - Utilisez le bouton "Tester la connexion" dans les paramètres s'il est disponible
 
+#### Configuration du Jeton GitLab
+
+OpenHands exporte automatiquement un `GITLAB_TOKEN` vers l'environnement shell, uniquement pour les installations locales, s'il est disponible.
+
+##### Configuration d'un Jeton GitLab
+
+1. **Générer un Personal Access Token (PAT)** :
+   - Sur GitLab, allez dans Paramètres utilisateur > Jetons d'accès
+   - Créez un nouveau jeton avec les portées suivantes :
+     - `api` (Accès API)
+     - `read_user` (Lecture des informations utilisateur)
+     - `read_repository` (Lecture du dépôt)
+     - `write_repository` (Écriture du dépôt)
+   - Définissez une date d'expiration ou laissez vide pour un jeton sans expiration
+
+2. **Entrer le Jeton dans OpenHands** :
+   - Cliquez sur le bouton Paramètres (icône d'engrenage)
+   - Accédez à la section `Git Provider Settings`
+   - Collez votre jeton dans le champ `Jeton GitLab`
+   - Si vous utilisez GitLab auto-hébergé, entrez l'URL de votre instance GitLab
+   - Cliquez sur `Enregistrer les modifications` pour appliquer les changements
+
+##### Dépannage
+
+Problèmes courants et solutions :
+
+1. **Jeton Non Reconnu** :
+   - Assurez-vous que le jeton est correctement enregistré dans les paramètres
+   - Vérifiez que le jeton n'a pas expiré
+   - Vérifiez que le jeton a les portées requises
+   - Pour les instances auto-hébergées, vérifiez l'URL correcte de l'instance
+
+2. **Accès Refusé** :
+   - Vérifiez les permissions d'accès au projet
+   - Vérifiez si le jeton possède les portées nécessaires
+   - Pour les dépôts de groupe/organisation, assurez-vous d'avoir les accès appropriés
+
 ### Paramètres Avancés
 
 1. Basculez sur `Options Avancées` pour accéder aux paramètres supplémentaires.

docs/i18n/fr/docusaurus-plugin-content-docs/current/usage/how-to/headless-mode.md

@@ -46,7 +46,7 @@ LLM_API_KEY="sk_test_12345"
 ```bash
 docker run -it \
     --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e SANDBOX_USER_ID=$(id -u) \
     -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
     -e LLM_API_KEY=$LLM_API_KEY \
@@ -56,6 +56,6 @@ docker run -it \
     -v /var/run/docker.sock:/var/run/docker.sock \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app-$(date +%Y%m%d%H%M%S) \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32 \
+    docker.all-hands.dev/all-hands-ai/openhands:0.36 \
     python -m openhands.core.main -t "write a bash script that prints hi" --no-auto-continue
 ```

docs/i18n/fr/docusaurus-plugin-content-docs/current/usage/installation.mdx

@@ -13,16 +13,16 @@
 La façon la plus simple d'exécuter OpenHands est avec Docker.
 
 ```bash
-docker pull docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik
+docker pull docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik
 
 docker run -it --rm --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e LOG_ALL_EVENTS=true \
     -v /var/run/docker.sock:/var/run/docker.sock \
     -p 3000:3000 \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32
+    docker.all-hands.dev/all-hands-ai/openhands:0.36
 ```
 
 Vous pouvez également exécuter OpenHands en mode [headless scriptable](https://docs.all-hands.dev/modules/usage/how-to/headless-mode), en tant que [CLI interactive](https://docs.all-hands.dev/modules/usage/how-to/cli-mode), ou en utilisant l'[Action GitHub OpenHands](https://docs.all-hands.dev/modules/usage/how-to/github-action).

docs/i18n/fr/docusaurus-plugin-content-docs/current/usage/runtimes.md

@@ -13,7 +13,7 @@ C'est le Runtime par défaut qui est utilisé lorsque vous démarrez OpenHands.
 
 ```
 docker run # ...
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -v /var/run/docker.sock:/var/run/docker.sock \
     # ...
 ```

docs/i18n/ja/docusaurus-plugin-content-docs/current/usage/configuration-options.md

@@ -348,12 +348,12 @@ dockerコマンドで使用する場合は、`-e LLM_<option>`として渡しま
   - デフォルト値: `true`
   - 説明: 関数呼び出しが有効かどうか
 
-- `codeact_enable_browsing`
+- `enable_browsing`
   - 型: `bool`
   - デフォルト値: `false`
   - 説明: アクションスペースでブラウジングデリゲートが有効かどうか（関数呼び出しでのみ機能）
 
-- `codeact_enable_llm_editor`
+- `enable_llm_editor`
   - 型: `bool`
   - デフォルト値: `false`
   - 説明: アクションスペースでLLMエディタが有効かどうか（関数呼び出しでのみ機能）

docs/i18n/ja/docusaurus-plugin-content-docs/current/usage/how-to/cli-mode.md

@@ -34,7 +34,7 @@ Docker で OpenHands を CLI モードで実行するには:
 ```bash
 docker run -it \
     --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e SANDBOX_USER_ID=$(id -u) \
     -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
     -e LLM_API_KEY=$LLM_API_KEY \
@@ -44,7 +44,7 @@ docker run -it \
     -v ~/.openhands-state:/.openhands-state \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app-$(date +%Y%m%d%H%M%S) \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32 \
+    docker.all-hands.dev/all-hands-ai/openhands:0.36 \
     python -m openhands.core.cli
 ```
 

docs/i18n/ja/docusaurus-plugin-content-docs/current/usage/how-to/gui-mode.md

@@ -16,7 +16,11 @@ OpenHandsは、AI アシスタントとやり取りするためのグラフィ
 3. 選択したプロバイダーに対応する`API Key`を入力します。
 4. `Save Changes`をクリックして設定を適用します。
 
-### GitHubトークンの設定
+### バージョン管理トークン
+
+OpenHandsは複数のバージョン管理プロバイダーをサポートしています。複数のプロバイダーのトークンを同時に設定できます。
+
+#### GitHubトークンの設定
 
 OpenHandsは、利用可能な場合、自動的に`GITHUB_TOKEN`をシェル環境にエクスポートします。これは2つの方法で行われます。
 
@@ -34,7 +38,7 @@ OpenHandsは、利用可能な場合、自動的に`GITHUB_TOKEN`をシェル環
      - Minimal Permissions（検索用に**Meta Data = Read-only**を選択し、ブランチ作成用に**Pull Requests = Read and Write**、**Content = Read and Write**を選択します）
   2. **OpenHandsにトークンを入力**:
    - 設定ボタン（歯車アイコン）をクリックします。
-   - `GitHub Settings`セクションに移動します。
+   - `Git Provider Settings`セクションに移動します。
    - `GitHub Token`フィールドにトークンを貼り付けます。
    - `Save Changes`をクリックして変更を適用します。
 </details>
@@ -94,6 +98,46 @@ OpenHandsは、利用可能な場合、自動的に`GITHUB_TOKEN`をシェル環
    - 組織を使用している場合は、プロンプトが表示されたら組織へのアクセスを承認します。
 </details>
 
+#### GitLabトークンの設定
+
+OpenHandsは、利用可能な場合、ローカルインストールのみ、自動的に`GITLAB_TOKEN`をシェル環境にエクスポートします。
+
+<details>
+  <summary>GitLabトークンの設定</summary>
+
+  1. **Personal Access Token（PAT）の生成**:
+   - GitLabで、User Settings > Access Tokensに移動します。
+   - 以下のスコープを持つ新しいトークンを作成します:
+     - `api`（APIアクセス）
+     - `read_user`（ユーザー情報の読み取り）
+     - `read_repository`（リポジトリ読み取り）
+     - `write_repository`（リポジトリ書き込み）
+   - 有効期限を設定するか、無期限トークンの場合は空白のままにします。
+  2. **OpenHandsにトークンを入力**:
+   - 設定ボタン（歯車アイコン）をクリックします。
+   - `Git Provider Settings`セクションに移動します。
+   - `GitLab Token`フィールドにトークンを貼り付けます。
+   - セルフホスト型GitLabを使用している場合は、GitLabインスタンスのURLを入力します。
+   - `Save Changes`をクリックして変更を適用します。
+</details>
+
+<details>
+  <summary>トラブルシューティング</summary>
+
+  一般的な問題と解決策:
+
+  - **トークンが認識されない**:
+     - トークンが設定に正しく保存されていることを確認します。
+     - トークンの有効期限が切れていないことを確認します。
+     - トークンに必要なスコープがあることを確認します。
+     - セルフホスト型インスタンスの場合は、正しいインスタンスURLを確認します。
+
+  - **アクセスが拒否された**:
+     - プロジェクトのアクセス権限を確認します。
+     - トークンに必要なスコープがあるかどうかを確認します。
+     - グループ/組織のリポジトリの場合は、適切なアクセス権があることを確認します。
+</details>
+
 ### 高度な設定
 
 1. 設定ページ内で、`Advanced`オプションを切り替えて追加の設定にアクセスします。

docs/i18n/ja/docusaurus-plugin-content-docs/current/usage/how-to/headless-mode.md

@@ -31,7 +31,7 @@ DockerでOpenHandsをヘッドレスモードで実行するには:
 ```bash
 docker run -it \
     --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e SANDBOX_USER_ID=$(id -u) \
     -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
     -e LLM_API_KEY=$LLM_API_KEY \
@@ -42,7 +42,7 @@ docker run -it \
     -v ~/.openhands-state:/.openhands-state \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app-$(date +%Y%m%d%H%M%S) \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32 \
+    docker.all-hands.dev/all-hands-ai/openhands:0.36 \
     python -m openhands.core.main -t "write a bash script that prints hi"
 ```
 

docs/i18n/ja/docusaurus-plugin-content-docs/current/usage/prompting/microagents-overview.md

@@ -13,7 +13,7 @@ OpenHandsがリポジトリで動作する際:
 
 1. リポジトリに`.openhands/microagents/`が存在する場合、そこからリポジトリ固有の指示を読み込みます。
 2. 会話のキーワードによってトリガーされる一般的なガイドラインを読み込みます。
-現在の[パブリックMicroagents](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/knowledge)を参照してください。
+現在の[パブリックMicroagents](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents)を参照してください。
 
 ## Microagentのフォーマット
 

docs/i18n/ja/docusaurus-plugin-content-docs/current/usage/prompting/microagents-public.md

@@ -88,4 +88,4 @@ triggers:
 - ビルド時間とイメージサイズを最適化
 ```
 
-より多くの例については、[現在のパブリックマイクロエージェント](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/knowledge)をご覧ください。
+より多くの例については、[現在のパブリックマイクロエージェント](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents)をご覧ください。

docs/i18n/ja/docusaurus-plugin-content-docs/current/usage/runtimes/docker.md

@@ -25,7 +25,7 @@ nikolaik の `SANDBOX_RUNTIME_CONTAINER_IMAGE` は、ランタイムサーバー
 
     ```bash
     docker run # ...
-        -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+        -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
         -e SANDBOX_USER_ID=$(id -u) \
         -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
         -v $WORKSPACE_BASE:/opt/workspace_base \
@@ -82,5 +82,5 @@ docker network create openhands-network
 # 分離されたネットワークで OpenHands を実行
 docker run # ... \
     --network openhands-network \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32
+    docker.all-hands.dev/all-hands-ai/openhands:0.36
 ```

docs/i18n/pt-BR/docusaurus-plugin-content-docs/current/usage/configuration-options.md

@@ -292,17 +292,17 @@ As opções de configuração do agente são definidas nas seções `[agent]` e
   - Padrão: `true`
   - Descrição: Se a chamada de função está habilitada
 
-- `codeact_enable_browsing`
+- `enable_browsing`
   - Tipo: `bool`
   - Padrão: `false`
   - Descrição: Se o delegado de navegação está habilitado no espaço de ação (funciona apenas com chamada de função)
 
-- `codeact_enable_llm_editor`
+- `enable_llm_editor`
   - Tipo: `bool`
   - Padrão: `false`
   - Descrição: Se o editor LLM está habilitado no espaço de ação (funciona apenas com chamada de função)
 
-- `codeact_enable_jupyter`
+- `enable_jupyter`
   - Tipo: `bool`
   - Padrão: `false`
   - Descrição: Se o Jupyter está habilitado no espaço de ação

docs/i18n/pt-BR/docusaurus-plugin-content-docs/current/usage/how-to/cli-mode.md

@@ -35,7 +35,7 @@ Para executar o OpenHands no modo CLI com Docker:
 ```bash
 docker run -it \
     --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e SANDBOX_USER_ID=$(id -u) \
     -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
     -e LLM_API_KEY=$LLM_API_KEY \
@@ -45,7 +45,7 @@ docker run -it \
     -v ~/.openhands-state:/.openhands-state \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app-$(date +%Y%m%d%H%M%S) \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32 \
+    docker.all-hands.dev/all-hands-ai/openhands:0.36 \
     python -m openhands.core.cli
 ```
 

docs/i18n/pt-BR/docusaurus-plugin-content-docs/current/usage/how-to/gui-mode.md

@@ -17,7 +17,11 @@ O OpenHands fornece um modo de Interface Gráfica do Usuário (GUI) para interag
 3. Insira a `Chave de API` correspondente para o provedor escolhido.
 4. Clique em `Salvar Alterações` para aplicar as configurações.
 
-### Configuração do Token do GitHub
+### Tokens de Controle de Versão
+
+O OpenHands suporta múltiplos provedores de controle de versão. Você pode configurar tokens para vários provedores simultaneamente.
+
+#### Configuração do Token do GitHub
 
 O OpenHands exporta automaticamente um `GITHUB_TOKEN` para o ambiente shell se ele estiver disponível. Isso pode acontecer de duas maneiras:
 
@@ -35,7 +39,7 @@ O OpenHands exporta automaticamente um `GITHUB_TOKEN` para o ambiente shell se e
      - Minimal Permissions (Selecione **Meta Data = Read-only** para pesquisa, **Pull Requests = Read and Write**, **Content = Read and Write** para criação de branches)
   2. **Insira o Token no OpenHands**:
    - Clique no botão Settings (ícone de engrenagem).
-   - Navegue até a seção `GitHub Settings`.
+   - Navegue até a seção `Git Provider Settings`.
    - Cole seu token no campo `GitHub Token`.
    - Clique em `Save Changes` para aplicar as alterações.
 </details>
@@ -95,6 +99,46 @@ O OpenHands exporta automaticamente um `GITHUB_TOKEN` para o ambiente shell se e
    - Se estiver usando uma organização, autorize o acesso à organização se solicitado.
 </details>
 
+#### Configuração do Token do GitLab
+
+O OpenHands exporta automaticamente um `GITLAB_TOKEN` para o ambiente shell, apenas para instalações locais, se ele estiver disponível.
+
+<details>
+  <summary>Configurando um Token do GitLab</summary>
+
+  1. **Gere um Personal Access Token (PAT)**:
+   - No GitLab, vá para User Settings > Access Tokens.
+   - Crie um novo token com os seguintes escopos:
+     - `api` (Acesso à API)
+     - `read_user` (Leitura de informações do usuário)
+     - `read_repository` (Leitura do repositório)
+     - `write_repository` (Escrita no repositório)
+   - Defina uma data de expiração ou deixe em branco para um token sem expiração.
+  2. **Insira o Token no OpenHands**:
+   - Clique no botão Settings (ícone de engrenagem).
+   - Navegue até a seção `Git Provider Settings`.
+   - Cole seu token no campo `GitLab Token`.
+   - Se estiver usando GitLab auto-hospedado, insira a URL da sua instância GitLab.
+   - Clique em `Save Changes` para aplicar as alterações.
+</details>
+
+<details>
+  <summary>Solução de Problemas</summary>
+
+  Problemas comuns e soluções:
+
+  - **Token Não Reconhecido**:
+     - Certifique-se de que o token esteja salvo corretamente nas configurações.
+     - Verifique se o token não expirou.
+     - Verifique se o token possui os escopos necessários.
+     - Para instâncias auto-hospedadas, verifique a URL correta da instância.
+
+  - **Acesso Negado**:
+     - Verifique as permissões de acesso ao projeto.
+     - Verifique se o token possui os escopos necessários.
+     - Para repositórios de grupo/organização, certifique-se de ter o acesso adequado.
+</details>
+
 ### Configurações Avançadas
 
 1. Dentro da página Settings, ative as opções `Advanced` para acessar configurações adicionais.

docs/i18n/pt-BR/docusaurus-plugin-content-docs/current/usage/how-to/headless-mode.md

@@ -32,7 +32,7 @@ Para executar o OpenHands no modo Headless com Docker:
 ```bash
 docker run -it \
     --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e SANDBOX_USER_ID=$(id -u) \
     -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
     -e LLM_API_KEY=$LLM_API_KEY \
@@ -43,7 +43,7 @@ docker run -it \
     -v ~/.openhands-state:/.openhands-state \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app-$(date +%Y%m%d%H%M%S) \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32 \
+    docker.all-hands.dev/all-hands-ai/openhands:0.36 \
     python -m openhands.core.main -t "escreva um script bash que imprima oi"
 ```
 

docs/i18n/pt-BR/docusaurus-plugin-content-docs/current/usage/installation.mdx

@@ -58,17 +58,17 @@
 A maneira mais fácil de executar o OpenHands é no Docker.
 
 ```bash
-docker pull docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik
+docker pull docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik
 
 docker run -it --rm --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e LOG_ALL_EVENTS=true \
     -v /var/run/docker.sock:/var/run/docker.sock \
     -v ~/.openhands-state:/.openhands-state \
     -p 3000:3000 \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32
+    docker.all-hands.dev/all-hands-ai/openhands:0.36
 ```
 
 Você encontrará o OpenHands em execução em http://localhost:3000!

docs/i18n/pt-BR/docusaurus-plugin-content-docs/current/usage/prompting/microagents-overview.md

@@ -13,7 +13,7 @@ Quando o OpenHands trabalha com um repositório, ele:
 
 1. Carrega instruções específicas do repositório de `.openhands/microagents/`, se presentes no repositório.
 2. Carrega diretrizes gerais acionadas por palavras-chave nas conversas.
-Veja os [Microagentes Públicos](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/knowledge) atuais.
+Veja os [Microagentes Públicos](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents) atuais.
 
 ## Formato do Microagente
 

docs/i18n/pt-BR/docusaurus-plugin-content-docs/current/usage/prompting/microagents-public.md

@@ -4,7 +4,7 @@
 
 Microagentes públicos são diretrizes especializadas acionadas por palavras-chave para todos os usuários do OpenHands.
 Eles são definidos em arquivos markdown no diretório
-[`microagents/knowledge/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/knowledge).
+[`microagents/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents).
 
 Microagentes públicos:
 - Monitoram comandos recebidos em busca de suas palavras-chave de acionamento.
@@ -15,7 +15,7 @@ Microagentes públicos:
 ## Microagentes Públicos Atuais
 
 Para mais informações sobre microagentes específicos, consulte seus arquivos de documentação individuais no
-diretório [`microagents/knowledge/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/knowledge/).
+diretório [`microagents/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/).
 
 ### Agente GitHub
 **Arquivo**: `github.md`
@@ -59,7 +59,7 @@ yes | npm install package-name
 ## Contribuindo com um Microagente Público
 
 Você pode criar seus próprios microagentes públicos adicionando novos arquivos markdown ao
-diretório [`microagents/knowledge/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/knowledge/).
+diretório [`microagents/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/).
 
 ### Melhores Práticas para Microagentes Públicos
 
@@ -81,7 +81,7 @@ Antes de criar um microagente público, considere:
 
 #### 2. Crie o Arquivo
 
-Crie um novo arquivo markdown em [`microagents/knowledge/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/knowledge/)
+Crie um novo arquivo markdown em [`microagents/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/)
 com um nome descritivo (por exemplo, `docker.md` para um agente focado em Docker).
 
 Atualize o arquivo com o frontmatter necessário [de acordo com o formato exigido](./microagents-overview#microagent-format)
@@ -149,5 +149,5 @@ Lembre-se de:
 - Otimizar para tempo de build e tamanho da imagem
 ```
 
-Veja os [microagentes públicos atuais](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/knowledge) para
+Veja os [microagentes públicos atuais](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents) para
 mais exemplos.

docs/i18n/pt-BR/docusaurus-plugin-content-docs/current/usage/runtimes.md

@@ -13,7 +13,7 @@ Este é o Runtime padrão que é usado quando você inicia o OpenHands. Você po
 
 ```
 docker run # ...
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -v /var/run/docker.sock:/var/run/docker.sock \
     # ...
 ```

docs/i18n/zh-Hans/docusaurus-plugin-content-docs/current/usage/configuration-options.md

@@ -349,17 +349,17 @@ Agent 配置选项在 `config.toml` 文件的 `[agent]` 和 `[agent.<agent_name>
   - 默认值: `true`
   - 描述: 是否启用函数调用
 
-- `codeact_enable_browsing`
+- `enable_browsing`
   - 类型: `bool`
   - 默认值: `false`
   - 描述: 是否在 action space 中启用浏览代理(仅适用于函数调用)
 
-- `codeact_enable_llm_editor`
+- `enable_llm_editor`
   - 类型: `bool`
   - 默认值: `false`
   - 描述: 是否在 action space 中启用 LLM 编辑器(仅适用于函数调用)
 
-- `codeact_enable_jupyter`
+- `enable_jupyter`
   - 类型: `bool`
   - 默认值: `false`
   - 描述: 是否在 action space 中启用 Jupyter

docs/i18n/zh-Hans/docusaurus-plugin-content-docs/current/usage/how-to/cli-mode.md

@@ -50,7 +50,7 @@ LLM_API_KEY="sk_test_12345"
 ```bash
 docker run -it \
     --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e SANDBOX_USER_ID=$(id -u) \
     -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
     -e LLM_API_KEY=$LLM_API_KEY \
@@ -59,7 +59,7 @@ docker run -it \
     -v /var/run/docker.sock:/var/run/docker.sock \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app-$(date +%Y%m%d%H%M%S) \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32 \
+    docker.all-hands.dev/all-hands-ai/openhands:0.36 \
     python -m openhands.core.cli
 ```
 

docs/i18n/zh-Hans/docusaurus-plugin-content-docs/current/usage/how-to/gui-mode.md

@@ -19,14 +19,18 @@ OpenHands 提供了一个用户友好的图形用户界面（GUI）模式，用
 3. 输入所选提供商对应的 `API Key`。
 4. 点击"保存"应用设置。
 
-### GitHub Token 设置
+### 版本控制令牌
+
+OpenHands 支持多个版本控制提供商。您可以同时配置多个提供商的令牌。
+
+#### GitHub Token 设置
 
 如果可用，OpenHands 会自动将 `GITHUB_TOKEN` 导出到 shell 环境中。这可以通过两种方式实现：
 
 1. **本地（OSS）**：用户直接输入他们的 GitHub token
 2. **在线（SaaS）**：通过 GitHub OAuth 身份验证获取 token
 
-#### 设置本地 GitHub Token
+##### 设置本地 GitHub Token
 
 1. **生成个人访问令牌（PAT）**：
    - 转到 GitHub 设置 > 开发者设置 > 个人访问令牌 > 令牌（经典）
@@ -38,11 +42,11 @@ OpenHands 提供了一个用户友好的图形用户界面（GUI）模式，用
 
 2. **在 OpenHands 中输入令牌**：
    - 点击右上角的设置按钮（齿轮图标）
-   - 导航到"GitHub"部分
+   - 导航到"Git Provider Settings"部分
    - 将令牌粘贴到"GitHub Token"字段中
    - 点击"保存"应用更改
 
-#### 组织令牌策略
+##### 组织令牌策略
 
 如果您使用组织仓库，可能需要额外的设置：
 
@@ -57,7 +61,7 @@ OpenHands 提供了一个用户友好的图形用户界面（GUI）模式，用
    - 如果需要，点击组织旁边的"启用 SSO"
    - 完成 SSO 授权过程
 
-#### OAuth 身份验证（在线模式）
+##### OAuth 身份验证（在线模式）
 
 在在线模式下使用 OpenHands 时，GitHub OAuth 流程：
 
@@ -72,7 +76,7 @@ OpenHands 提供了一个用户友好的图形用户界面（GUI）模式，用
    - 授权 OpenHands 访问您的 GitHub 帐户
    - 如果使用组织，在出现提示时授权组织访问
 
-#### 故障排除
+##### 故障排除
 
 常见问题和解决方案：
 
@@ -93,6 +97,43 @@ OpenHands 提供了一个用户友好的图形用户界面（GUI）模式，用
    - 检查浏览器控制台中是否有任何错误消息
    - 如果可用，使用设置中的"测试连接"按钮
 
+#### GitLab Token 设置
+
+OpenHands 会自动将 `GITLAB_TOKEN` 导出到 shell 环境中，仅适用于本地安装，如果它可用的话。
+
+##### 设置 GitLab Token
+
+1. **生成个人访问令牌（PAT）**：
+   - 在 GitLab 中，转到用户设置 > 访问令牌
+   - 创建具有以下范围的新令牌：
+     - `api`（API 访问）
+     - `read_user`（读取用户信息）
+     - `read_repository`（读取仓库）
+     - `write_repository`（写入仓库）
+   - 设置过期日期或留空以获取永不过期的令牌
+
+2. **在 OpenHands 中输入令牌**：
+   - 点击设置按钮（齿轮图标）
+   - 导航到 `Git Provider Settings` 部分
+   - 将令牌粘贴到 `GitLab Token` 字段中
+   - 如果使用自托管 GitLab，请输入您的 GitLab 实例 URL
+   - 点击 `Save Changes` 应用更改
+
+##### 故障排除
+
+常见问题和解决方案：
+
+1. **令牌无法识别**：
+   - 确保令牌已正确保存在设置中
+   - 检查令牌是否已过期
+   - 验证令牌是否具有所需的范围
+   - 对于自托管实例，验证正确的实例 URL
+
+2. **访问被拒绝**：
+   - 验证项目访问权限
+   - 检查令牌是否具有必要的范围
+   - 对于组/组织仓库，确保您拥有适当的访问权限
+
 ### 高级设置
 
 1. 切换`高级选项`以访问其他设置。

docs/i18n/zh-Hans/docusaurus-plugin-content-docs/current/usage/how-to/headless-mode.md

@@ -47,7 +47,7 @@ LLM_API_KEY="sk_test_12345"
 ```bash
 docker run -it \
     --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e SANDBOX_USER_ID=$(id -u) \
     -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
     -e LLM_API_KEY=$LLM_API_KEY \
@@ -57,6 +57,6 @@ docker run -it \
     -v /var/run/docker.sock:/var/run/docker.sock \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app-$(date +%Y%m%d%H%M%S) \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32 \
+    docker.all-hands.dev/all-hands-ai/openhands:0.36 \
     python -m openhands.core.main -t "write a bash script that prints hi" --no-auto-continue
 ```

docs/i18n/zh-Hans/docusaurus-plugin-content-docs/current/usage/installation.mdx

@@ -11,16 +11,16 @@
 在 Docker 中运行 OpenHands 是最简单的方式。
 
 ```bash
-docker pull docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik
+docker pull docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik
 
 docker run -it --rm --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e LOG_ALL_EVENTS=true \
     -v /var/run/docker.sock:/var/run/docker.sock \
     -p 3000:3000 \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32
+    docker.all-hands.dev/all-hands-ai/openhands:0.36
 ```
 
 你也可以在可脚本化的[无头模式](https://docs.all-hands.dev/modules/usage/how-to/headless-mode)下运行 OpenHands，作为[交互式 CLI](https://docs.all-hands.dev/modules/usage/how-to/cli-mode)，或使用 [OpenHands GitHub Action](https://docs.all-hands.dev/modules/usage/how-to/github-action)。

docs/i18n/zh-Hans/docusaurus-plugin-content-docs/current/usage/runtimes.md

@@ -11,7 +11,7 @@
 
 ```
 docker run # ...
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -v /var/run/docker.sock:/var/run/docker.sock \
     # ...
 ```

docs/modules/usage/cloud/_category_.json

@@ -0,0 +1,8 @@
+{
+  "label": "OpenHands Cloud",
+  "position": 9,
+  "link": {
+    "type": "generated-index",
+    "description": "Documentation for OpenHands Cloud features and services."
+  }
+}

docs/modules/usage/cloud/cloud-api.md

@@ -0,0 +1,177 @@
+# OpenHands Cloud API
+
+OpenHands Cloud provides a REST API that allows you to programmatically interact with the service. This is useful if you easily want to kick off your own jobs from your programs in a flexible way.
+
+This guide explains how to obtain an API key and use the API to start conversations.
+For more detailed information about the API, refer to the [OpenHands API Reference](https://docs.all-hands.dev/swagger-ui/).
+
+## Obtaining an API Key
+
+To use the OpenHands Cloud API, you'll need to generate an API key:
+
+1. Log in to your [OpenHands Cloud](https://app.all-hands.dev) account
+2. Navigate to the [Settings page](https://app.all-hands.dev/settings)
+3. Locate the "API Keys" section
+4. Click "Generate New Key"
+5. Give your key a descriptive name (e.g., "Development", "Production")
+6. Copy the generated API key and store it securely - it will only be shown once
+
+![API Key Generation](/img/docs/api-key-generation.png)
+
+## API Usage
+
+### Starting a New Conversation
+
+To start a new conversation with OpenHands performing a task, you'll need to make a POST request to the conversation endpoint.
+
+#### Request Parameters
+
+| Parameter | Type | Required | Description |
+|-----------|------|----------|-------------|
+| `initial_user_msg` | string | Yes | The initial message to start the conversation |
+| `repository` | string | No | Git repository name to provide context in the format `owner/repo`. You must have access to the repo. |
+
+#### Examples
+
+<details>
+<summary>cURL</summary>
+
+```bash
+curl -X POST "https://app.all-hands.dev/api/conversations" \
+  -H "Authorization: Bearer YOUR_API_KEY" \
+  -H "Content-Type: application/json" \
+  -d '{
+    "initial_user_msg": "Check whether there is any incorrect information in the README.md file and send a PR to fix it if so.",
+    "repository": "yourusername/your-repo"
+  }'
+```
+</details>
+
+<details>
+<summary>Python (with requests)</summary>
+
+```python
+import requests
+
+api_key = "YOUR_API_KEY"
+url = "https://app.all-hands.dev/api/conversations"
+
+headers = {
+    "Authorization": f"Bearer {api_key}",
+    "Content-Type": "application/json"
+}
+
+data = {
+    "initial_user_msg": "Check whether there is any incorrect information in the README.md file and send a PR to fix it if so.",
+    "repository": "yourusername/your-repo"
+}
+
+response = requests.post(url, headers=headers, json=data)
+conversation = response.json()
+
+print(f"Conversation Link: https://app.all-hands.dev/conversations/{conversation['id']}")
+print(f"Status: {conversation['status']}")
+```
+</details>
+
+<details>
+<summary>TypeScript/JavaScript (with fetch)</summary>
+
+```typescript
+const apiKey = "YOUR_API_KEY";
+const url = "https://app.all-hands.dev/api/conversations";
+
+const headers = {
+  "Authorization": `Bearer ${apiKey}`,
+  "Content-Type": "application/json"
+};
+
+const data = {
+  initial_user_msg: "Check whether there is any incorrect information in the README.md file and send a PR to fix it if so.",
+  repository: "yourusername/your-repo"
+};
+
+async function startConversation() {
+  try {
+    const response = await fetch(url, {
+      method: "POST",
+      headers: headers,
+      body: JSON.stringify(data)
+    });
+
+    const conversation = await response.json();
+
+    console.log(`Conversation Link: https://app.all-hands.dev/conversations/${conversation.id}`);
+    console.log(`Status: ${conversation.status}`);
+
+    return conversation;
+  } catch (error) {
+    console.error("Error starting conversation:", error);
+  }
+}
+
+startConversation();
+```
+
+</details>
+
+#### Response
+
+The API will return a JSON object with details about the created conversation:
+
+```json
+{
+  "status": "ok",
+  "conversation_id": "abc1234",
+}
+```
+
+You may also receive an `AuthenticationError` if:
+
+1. You provided an invalid API key
+2. You provided the wrong repo name
+3. You don't have access to the repo
+
+
+### Retrieving Conversation Status
+
+You can check the status of a conversation by making a GET request to the conversation endpoint.
+
+#### Endpoint
+
+```
+GET https://app.all-hands.dev/api/conversations/{conversation_id}
+```
+
+#### Example
+
+<details>
+<summary>cURL</summary>
+
+```bash
+curl -X GET "https://app.all-hands.dev/api/conversations/{conversation_id}" \
+  -H "Authorization: Bearer YOUR_API_KEY"
+```
+</details>
+
+#### Response
+
+The response is formatted as follows:
+
+```json
+{
+  "conversation_id":"abc1234",
+  "title":"Update README.md",
+  "created_at":"2025-04-29T15:13:51.370706Z",
+  "last_updated_at":"2025-04-29T15:13:57.199210Z",
+  "status":"RUNNING",
+  "selected_repository":"yourusername/your-repo",
+  "trigger":"gui"
+}
+```
+
+## Rate Limits
+
+The API has a limit of 10 simultaneous conversations per account. If you need a higher limit for your use case, please contact us at [contact@all-hands.dev](mailto:contact@all-hands.dev).
+
+If you exceed this limit, the API will return a 429 Too Many Requests response.

docs/modules/usage/cloud/openhands-cloud.mdx

@@ -6,20 +6,26 @@ OpenHands Cloud is the cloud hosted version of OpenHands by All Hands AI.
 
 OpenHands Cloud can be accessed at https://app.all-hands.dev/.
 
+You can also interact with OpenHands Cloud programmatically using the [API](./cloud-api).
+
 ## Getting Started
 
-After visiting OpenHands Cloud, you will be asked to connect with your GitHub account:
-1. After reading and accepting the terms of service, click `Connect to GitHub`.
+After visiting OpenHands Cloud, you will be asked to connect with your GitHub or GitLab account:
+
+1. After reading and accepting the terms of service, click `Log in with GitHub` or `Log in with GitLab`.
 2. Review the permissions requested by OpenHands and then click `Authorize OpenHands AI`.
-   - OpenHands will require some permissions from your GitHub account. To read more about these permissions,
-     you can click the `Learn more` link on the GitHub authorize page.
+   - OpenHands will require some permissions from your GitHub or GitLab account. To read more about these permissions:
+     - GitHub: You can click the `Learn more` link on the GitHub authorize page.
+     - GitLab: You can expand each permission request on the GitLab authorize page.
 
 ## Repository Access
 
-### Adding Repository Access
+### GitHub
+
+#### Adding Repository Access
 
 You can grant OpenHands specific repository access:
-1. Click the `Select a GitHub project` dropdown, select `Add more repositories...`.
+1. Click `Add GitHub repos` on the Home page.
 2. Select the organization, then choose the specific repositories to grant OpenHands access to.
    <details>
      <summary>Permission Details for Repository Access</summary>
@@ -42,11 +48,15 @@ You can grant OpenHands specific repository access:
 
 3. Click on `Install & Authorize`.
 
-### Modifying Repository Access
+#### Modifying Repository Access
 
-You can modify repository access at any time by:
-* Using the same `Select a GitHub project > Add more repositories` workflow, or
-* Visiting the Settings page and selecting `Configure GitHub Repositories` under the `GitHub Settings` section.
+You can modify GitHub repository access at any time by:
+* Using the same `Add GitHub repos` workflow, or
+* Visiting the Settings page and selecting `Configure GitHub Repositories` under the `Git Settings` section.
+
+### GitLab
+
+When using your GitLab account, OpenHands will automatically have access to your repositories.
 
 ## Conversation Persistence
 

docs/modules/usage/configuration-options.md

@@ -291,17 +291,17 @@ The agent configuration options are defined in the `[agent]` and `[agent.<agent_
   - Default: `true`
   - Description: Whether function calling is enabled
 
-- `codeact_enable_browsing`
+- `enable_browsing`
   - Type: `bool`
   - Default: `false`
   - Description: Whether browsing delegate is enabled in the action space (only works with function calling)
 
-- `codeact_enable_llm_editor`
+- `enable_llm_editor`
   - Type: `bool`
   - Default: `false`
   - Description: Whether LLM editor is enabled in the action space (only works with function calling)
 
-- `codeact_enable_jupyter`
+- `enable_jupyter`
   - Type: `bool`
   - Default: `false`
   - Description: Whether Jupyter is enabled in the action space

docs/modules/usage/customization/repository.md

@@ -1,18 +1,17 @@
 # Repository Customization
 
-You can customize how OpenHands works with your repository by creating a
+You can customize how OpenHands interacts with your repository by creating a
 `.openhands` directory at the root level.
 
 ## Microagents
-You can use microagents to extend the OpenHands prompts with information
-about your project and how you want OpenHands to work. See
-[Repository Microagents](../prompting/microagents-repo) for more information.
+
+Microagents allow you to extend OpenHands prompts with information specific to your project and define how OpenHands
+should function. See [Microagents Overview](../prompting/microagents-overview) for more information.
 
 
 ## Setup Script
-You can add `.openhands/setup.sh`, which will be run every time OpenHands begins
-working with your repository. This is a good place to install dependencies, set
-environment variables, etc.
+You can add a `.openhands/setup.sh` file, which will run every time OpenHands begins working with your repository.
+This is an ideal location for installing dependencies, setting environment variables, and performing other setup tasks.
 
 For example:
 ```bash

docs/modules/usage/how-to/cli-mode.md

@@ -35,7 +35,7 @@ To run OpenHands in CLI mode with Docker:
 ```bash
 docker run -it \
     --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e SANDBOX_USER_ID=$(id -u) \
     -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
     -e LLM_API_KEY=$LLM_API_KEY \
@@ -45,8 +45,11 @@ docker run -it \
     -v ~/.openhands-state:/.openhands-state \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app-$(date +%Y%m%d%H%M%S) \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32 \
+    docker.all-hands.dev/all-hands-ai/openhands:0.36 \
     python -m openhands.core.cli
 ```
 
 This command will start an interactive session in Docker where you can input tasks and receive responses from OpenHands.
+
+The `-e SANDBOX_USER_ID=$(id -u)` is passed to the Docker command to ensure the sandbox user matches the host user’s
+permissions. This prevents the agent from creating root-owned files in the mounted workspace.

docs/modules/usage/how-to/custom-sandbox-guide.md

@@ -1,7 +1,8 @@
 # Custom Sandbox
 
 :::note
-This guide is for users that would like to use their own custom Docker image for the runtime, e.g. with certain tools or programming languages pre-installed
+This guide is for users that would like to use their own custom Docker image for the runtime. For example
+with certain tools or programming languages pre-installed.
 :::
 
 The sandbox is where the agent performs its tasks. Instead of running commands directly on your computer

docs/modules/usage/how-to/gui-mode.md

@@ -18,11 +18,14 @@ OpenHands provides a Graphical User Interface (GUI) mode for interacting with th
 3. Enter the corresponding `API Key` for your chosen provider.
 4. Click `Save Changes` to apply the settings.
 
-### GitHub Token Setup
+### Version Control Tokens
 
-OpenHands automatically exports a `GITHUB_TOKEN` to the shell environment if it is available. This can happen in two ways:
+OpenHands supports multiple version control providers. You can configure tokens for multiple providers simultaneously.
+
+#### GitHub Token Setup
+
+OpenHands automatically exports a `GITHUB_TOKEN` to the shell environment if provided:
 
-**Local Installation**: The user directly inputs their GitHub token.
 <details>
   <summary>Setting Up a GitHub Token</summary>
 
@@ -36,9 +39,8 @@ OpenHands automatically exports a `GITHUB_TOKEN` to the shell environment if it
      - Minimal Permissions ( Select `Meta Data = Read-only` read for search, `Pull Requests = Read and Write` and `Content = Read and Write` for branch creation)
   2. **Enter Token in OpenHands**:
    - Click the Settings button (gear icon).
-   - Navigate to the `GitHub Settings` section.
    - Paste your token in the `GitHub Token` field.
-   - Click `Save Changes` to apply the changes.
+   - Click `Save` to apply the changes.
 </details>
 
 <details>
@@ -79,21 +81,43 @@ OpenHands automatically exports a `GITHUB_TOKEN` to the shell environment if it
      - Check the browser console for any error messages.
 </details>
 
-**OpenHands Cloud**: The token is obtained through GitHub OAuth authentication.
+#### GitLab Token Setup
+
+OpenHands automatically exports a `GITLAB_TOKEN` to the shell environment if provided:
 
 <details>
-  <summary>OAuth Authentication</summary>
+  <summary>Setting Up a GitLab Token</summary>
 
-  When using OpenHands Cloud, the GitHub OAuth flow requests the following permissions:
-   - Repository access (read/write)
-   - Workflow management
-   - Organization read access
+  1. **Generate a Personal Access Token (PAT)**:
+   - On GitLab, go to User Settings > Access Tokens.
+   - Create a new token with the following scopes:
+     - `api` (API access)
+     - `read_user` (Read user information)
+     - `read_repository` (Read repository)
+     - `write_repository` (Write repository)
+   - Set an expiration date or leave it blank for a non-expiring token.
+  2. **Enter Token in OpenHands**:
+   - Click the Settings button (gear icon).
+   - Paste your token in the `GitLab Token` field.
+   - Enter your GitLab instance URL if using self-hosted GitLab.
+   - Click `Save` to apply the changes.
+</details>
 
-  To authenticate OpenHands:
-   - Click `Sign in with GitHub` when prompted.
-   - Review the requested permissions.
-   - Authorize OpenHands to access your GitHub account.
-   - If using an organization, authorize organization access if prompted.
+<details>
+  <summary>Troubleshooting</summary>
+
+  Common issues and solutions:
+
+  - **Token Not Recognized**:
+     - Ensure the token is properly saved in settings.
+     - Check that the token hasn't expired.
+     - Verify the token has the required scopes.
+     - For self-hosted instances, verify the correct instance URL.
+
+  - **Access Denied**:
+     - Verify project access permissions.
+     - Check if the token has the necessary scopes.
+     - For group/organization repositories, ensure you have proper access.
 </details>
 
 ### Advanced Settings
@@ -112,7 +136,6 @@ OpenHands automatically exports a `GITHUB_TOKEN` to the shell environment if it
 ## Tips for Effective Use
 
 - Be specific in your requests to get the most accurate and helpful responses, as described in the [prompting best practices](../prompting/prompting-best-practices).
-- Use the workspace panel to explore your project structure.
 - Use one of the recommended models, as described in the [LLMs section](usage/llms/llms.md).
 
 Remember, the GUI mode of OpenHands is designed to make your interaction with the AI assistant as smooth and intuitive

docs/modules/usage/how-to/headless-mode.md

@@ -32,7 +32,7 @@ To run OpenHands in Headless mode with Docker:
 ```bash
 docker run -it \
     --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e SANDBOX_USER_ID=$(id -u) \
     -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
     -e LLM_API_KEY=$LLM_API_KEY \
@@ -43,10 +43,13 @@ docker run -it \
     -v ~/.openhands-state:/.openhands-state \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app-$(date +%Y%m%d%H%M%S) \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32 \
+    docker.all-hands.dev/all-hands-ai/openhands:0.36 \
     python -m openhands.core.main -t "write a bash script that prints hi"
 ```
 
+The `-e SANDBOX_USER_ID=$(id -u)` is passed to the Docker command to ensure the sandbox user matches the host user’s
+permissions. This prevents the agent from creating root-owned files in the mounted workspace.
+
 ## Advanced Headless Configurations
 
 To view all available configuration options for headless mode, run the Python command with the `--help` flag.

docs/modules/usage/installation.mdx

@@ -58,17 +58,17 @@ A system with a modern processor and a minimum of **4GB RAM** is recommended to
 The easiest way to run OpenHands is in Docker.
 
 ```bash
-docker pull docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik
+docker pull docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik
 
 docker run -it --rm --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.32-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.36-nikolaik \
     -e LOG_ALL_EVENTS=true \
     -v /var/run/docker.sock:/var/run/docker.sock \
     -v ~/.openhands-state:/.openhands-state \
     -p 3000:3000 \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app \
-    docker.all-hands.dev/all-hands-ai/openhands:0.32
+    docker.all-hands.dev/all-hands-ai/openhands:0.36
 ```
 
 You'll find OpenHands running at http://localhost:3000!

docs/modules/usage/key-features.md

@@ -6,23 +6,24 @@
 - Displays the conversation between the user and OpenHands.
 - OpenHands explains its actions in this panel.
 
-### Workspace
-- Browse project files and directories.
-- Use the `Open in VS Code` option to:
-  * Modify files
-  * Upload and download files
+### Changes
+- Shows the file changes performed by OpenHands.
+
+### VS Code
+- Embedded VS Code for browsing and modifying files.
+- Can also be used to upload and download files.
+
+### Terminal
+- A space for OpenHands and users to run terminal commands.
 
 ### Jupyter
 - Shows all Python commands that were executed by OpenHands.
 - Particularly handy when using OpenHands to perform data visualization tasks.
 
 ### App
-- Shows the web server when OpenHands runs an application.
+- Displays the web server when OpenHands runs an application.
 - Users can interact with the running application.
 
 ### Browser
 - Used by OpenHands to browse websites.
 - The browser is non-interactive.
-
-### Terminal
-- A space for OpenHands and users to run terminal commands.

docs/modules/usage/llms/llms.md

@@ -17,6 +17,8 @@ Based on these findings and community feedback, the following models have been v
 - [gemini/gemini-2.5-pro](https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/)
 - [deepseek/deepseek-chat](https://api-docs.deepseek.com/)
 - [openai/o3-mini](https://openai.com/index/openai-o3-mini/)
+- [openai/o3](https://openai.com/index/introducing-o3-and-o4-mini/)
+- [openai/o4-mini](https://openai.com/index/introducing-o3-and-o4-mini/)
 - [all-hands/openhands-lm-32b-v0.1](https://www.all-hands.dev/blog/introducing-openhands-lm-32b----a-strong-open-coding-agent-model) -- available through [OpenRouter](https://openrouter.ai/all-hands/openhands-lm-32b-v0.1)
 
 

docs/modules/usage/llms/local-llms.md

@@ -15,7 +15,7 @@ It is highly recommended that you use GPUs to serve local models for optimal exp
 For example, to download [OpenHands LM 32B v0.1](https://huggingface.co/all-hands/openhands-lm-32b-v0.1):
 
 ```bash
-huggingface-cli download all-hands/openhands-lm-32b-v0.1 --local-dir my_folder/openhands-lm-32b-v0.1
+huggingface-cli download all-hands/openhands-lm-32b-v0.1 --local-dir all-hands/openhands-lm-32b-v0.1
 ```
 
 ## Create an OpenAI-Compatible Endpoint With a Model Serving Framework
@@ -27,7 +27,7 @@ huggingface-cli download all-hands/openhands-lm-32b-v0.1 --local-dir my_folder/o
 
 ```bash
 SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1 python3 -m sglang.launch_server \
-    --model my_folder/openhands-lm-32b-v0.1 \
+    --model all-hands/openhands-lm-32b-v0.1 \
     --served-model-name openhands-lm-32b-v0.1 \
     --port 8000 \
     --tp 2 --dp 1 \
@@ -41,7 +41,7 @@ SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1 python3 -m sglang.launch_server \
 - Example launch command for OpenHands LM 32B (with at least 2 GPUs):
 
 ```bash
-vllm serve my_folder/openhands-lm-32b-v0.1 \
+vllm serve all-hands/openhands-lm-32b-v0.1 \
     --host 0.0.0.0 --port 8000 \
     --api-key mykey \
     --tensor-parallel-size 2 \
@@ -67,7 +67,7 @@ Ensure `config.toml` exists by running `make setup-config` which will create one
 workspace_base="/path/to/your/workspace"
 
 [llm]
-embedding_model="local"
+model="openhands-lm-32b-v0.1"
 ollama_base_url="http://localhost:8000"
 ```
 

docs/modules/usage/mcp.md

@@ -0,0 +1,96 @@
+# Model Context Protocol (MCP)
+
+:::note
+This page outlines how to configure and use the Model Context Protocol (MCP) in OpenHands, allowing you to extend the agent's capabilities with custom tools.
+:::
+
+## Overview
+
+Model Context Protocol (MCP) is a mechanism that allows OpenHands to communicate with external tool servers. These servers can provide additional functionality to the agent, such as specialized data processing, external API access, or custom tools. MCP is based on the open standard defined at [modelcontextprotocol.io](https://modelcontextprotocol.io).
+
+## Configuration
+
+MCP configuration is defined in the `[mcp]` section of your `config.toml` file.
+
+### Configuration Example
+
+```toml
+[mcp]
+# SSE Servers - External servers that communicate via Server-Sent Events
+sse_servers = [
+    # Basic SSE server with just a URL
+    "http://example.com:8080/mcp",
+
+    # SSE server with API key authentication
+    {url="https://secure-example.com/mcp", api_key="your-api-key"}
+]
+
+# Stdio Servers - Local processes that communicate via standard input/output
+stdio_servers = [
+    # Basic stdio server
+    {name="fetch", command="uvx", args=["mcp-server-fetch"]},
+
+    # Stdio server with environment variables
+    {
+        name="data-processor",
+        command="python",
+        args=["-m", "my_mcp_server"],
+        env={
+            "DEBUG": "true",
+            "PORT": "8080"
+        }
+    }
+]
+```
+
+## Configuration Options
+
+### SSE Servers
+
+SSE servers are configured using either a string URL or an object with the following properties:
+
+- `url` (required)
+  - Type: `str`
+  - Description: The URL of the SSE server
+
+- `api_key` (optional)
+  - Type: `str`
+  - Default: `None`
+  - Description: API key for authentication with the SSE server
+
+### Stdio Servers
+
+Stdio servers are configured using an object with the following properties:
+
+- `name` (required)
+  - Type: `str`
+  - Description: A unique name for the server
+
+- `command` (required)
+  - Type: `str`
+  - Description: The command to run the server
+
+- `args` (optional)
+  - Type: `list of str`
+  - Default: `[]`
+  - Description: Command-line arguments to pass to the server
+
+- `env` (optional)
+  - Type: `dict of str to str`
+  - Default: `{}`
+  - Description: Environment variables to set for the server process
+
+## How MCP Works
+
+When OpenHands starts, it:
+
+1. Reads the MCP configuration from `config.toml`
+2. Connects to any configured SSE servers
+3. Starts any configured stdio servers
+4. Registers the tools provided by these servers with the agent
+
+The agent can then use these tools just like any built-in tool. When the agent calls an MCP tool:
+
+1. OpenHands routes the call to the appropriate MCP server
+2. The server processes the request and returns a response
+3. OpenHands converts the response to an observation and presents it to the agent

docs/modules/usage/prompting/microagents-keyword.md

@@ -0,0 +1,38 @@
+# Keyword-Triggered Microagents
+
+## Purpose
+
+Keyword-triggered microagents provide OpenHands with specific instructions that are activated when certain keywords
+appear in the prompt. This is useful for tailoring behavior based on particular tools, languages, or frameworks.
+
+## Usage
+
+These microagents are only loaded when a prompt includes one of the trigger words.
+
+## Frontmatter Syntax
+
+Frontmatter is required for keyword-triggered microagents. It must be placed at the top of the file,
+above the guidelines.
+
+Enclose the frontmatter in triple dashes (---) and include the following fields:
+
+| Field      | Description                                      | Required | Default          |
+|------------|--------------------------------------------------|----------|------------------|
+| `triggers` | A list of keywords that activate the microagent. | Yes      | None             |
+| `agent`    | The agent this microagent applies to.            | No       | 'CodeActAgent'   |
+
+
+## Example
+
+Keyword-triggered microagent file example located at `.openhands/microagents/yummy.md`:
+```
+---
+triggers:
+- yummyhappy
+- happyyummy
+---
+
+The user has said the magic word. Respond with "That was delicious!"
+```
+
+[See examples of microagents triggered by keywords in the official OpenHands repository](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents)

docs/modules/usage/prompting/microagents-overview.md

@@ -1,31 +1,40 @@
 # Microagents Overview
 
-Microagents are specialized prompts that enhance OpenHands with domain-specific knowledge, repository-specific context
-and task-specific workflows. They help by providing expert guidance, automating common tasks, and ensuring
-consistent practices across projects.
+Microagents are specialized prompts that enhance OpenHands with domain-specific knowledge.
+They provide expert guidance, automate common tasks, and ensure consistent practices across projects.
 
-## Microagent Categories
+## Microagent Types
 
-Currently OpenHands supports two categories of microagents:
+Currently OpenHands supports the following types of microagents:
 
-- [Repository-specific Microagents](./microagents-repo): Repository-specific context and guidelines for OpenHands.
-- [Public Microagents](./microagents-public): General guidelines triggered by keywords for all OpenHands users.
+- [General Repository Microagents](./microagents-repo): General guidelines for OpenHands about the repository.
+- [Keyword-Triggered Microagents](./microagents-keyword): Guidelines activated by specific keywords in prompts.
 
-A microagent is classified as repository-specific or public depending on its location:
+To customize OpenHands' behavior, create a .openhands/microagents/ directory in the root of your repository and
+add `<microagent_name>.md` files inside.
 
-- Repository-specific microagents are located in a repository's `.openhands/microagents/` directory
-- Public microagents are located in the official OpenHands repository inside the `/microagents` folder
+:::note
+Loaded microagents take up space in the context window.
+These microagents, alongside user messages, inform OpenHands about the task and the environment.
+:::
 
-When OpenHands works with a repository, it:
+Example repository structure:
 
-1. Loads **repository-specific** microagents from `.openhands/microagents/` if present in the repository.
-2. Loads **public knowledge** microagents triggered by keywords in conversations
-3. Loads **public tasks** microagents when explicitly requested by the user
+```
+some-repository/
+└── .openhands/
+    └── microagents/
+        └── repo.md            # General repository guidelines
+        └── trigger_this.md    # Microagent triggered by specific keywords
+        └── trigger_that.md    # Microagent triggered by specific keywords
+```
 
-You can check out the existing public microagents at the [official OpenHands repository](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/).
+## Microagents Frontmatter Requirements
 
-## Microagent Format
+Each microagent file may include frontmatter that provides additional information. In some cases, this frontmatter
+is required:
 
-All microagents use markdown files with YAML frontmatter that have special instructions to help OpenHands activate them.
-
-Check out the [syntax documentation](./microagents-syntax) for a comprehensive guide on how to configure your microagents.
+| Microagent Type                  | Required |
+|----------------------------------|----------|
+| `General Repository Microagents` | No       |
+| `Keyword-Triggered Microagents`  | Yes      |

docs/modules/usage/prompting/microagents-public.md

@@ -1,35 +1,17 @@
-# Public Microagents
+# Global Microagents
 
 ## Overview
 
-Public microagents provide specialized context and capabilities for all OpenHands users, regardless of their repository configuration. Unlike repository-specific microagents, public microagents are globally available across all repositories.
+Global microagents are [keyword-triggered microagents](./microagents-keyword) that apply to all OpenHands users. A list of the current
+global microagents can be found [in the OpenHands repository](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents).
 
-Public microagents come in two types:
+## Contributing a Global Microagent
 
-- **Knowledge microagents**: Automatically activated when keywords in conversations match their triggers
-- **Task microagents**: Explicitly invoked by users to guide through specific workflows
-
-Both types follow the same syntax and structure as repository-specific microagents, using markdown files with YAML frontmatter that define their behavior and capabilities. They are located in the official OpenHands repository under:
-
-- [`microagents/knowledge/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/knowledge) for knowledge microagents
-- [`microagents/tasks/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/tasks) for task microagents
-
-Public microagents:
-
-- Monitor incoming commands for their trigger words.
-- Activate when relevant triggers are detected.
-- Apply their specialized knowledge and capabilities.
-- Follow their specific guidelines and restrictions.
-
-When loading public microagents, OpenHands scans the official repository's microagents directories recursively, processing all markdown files except README.md. The system categorizes each microagent based on its `type` field in the YAML frontmatter, regardless of its exact file location within the knowledge or tasks directories.
-
-## Contributing a Public Microagent
-
-You can create public microagents and share with the community by opening a pull request to the official repository.
+You can create global microagents and share with the community by opening a pull request to the official repository.
 
 See the [CONTRIBUTING.md](https://github.com/All-Hands-AI/OpenHands/blob/main/CONTRIBUTING.md) for specific instructions on how to contribute to OpenHands.
 
-### Public Microagents Best Practices
+### Global Microagents Best Practices
 
 - **Clear Scope**: Keep the microagent focused on a specific domain or task.
 - **Explicit Instructions**: Provide clear, unambiguous guidelines.
@@ -37,11 +19,11 @@ See the [CONTRIBUTING.md](https://github.com/All-Hands-AI/OpenHands/blob/main/CO
 - **Safety First**: Include necessary warnings and constraints.
 - **Integration Awareness**: Consider how the microagent interacts with other components.
 
-### Steps to Contribute a Public Microagent
+### Steps to Contribute a Global Microagent
 
-#### 1. Plan the Public Microagent
+#### 1. Plan the Global Microagent
 
-Before creating a public microagent, consider:
+Before creating a global microagent, consider:
 
 - What specific problem or use case will it address?
 - What unique capabilities or knowledge should it have?
@@ -51,23 +33,19 @@ Before creating a public microagent, consider:
 #### 2. Create File
 
 Create a new Markdown file with a descriptive name in the appropriate directory:
+[`microagents/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents)
 
-- [`microagents/knowledge/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/knowledge) for knowledge microagents
-- [`microagents/tasks/`](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/tasks) for task microagents
+#### 3. Testing the Global Microagent
 
-Ensure it follows the correct [syntax](./microagents-syntax.md) and [best practices](./microagents-syntax.md#markdown-content-best-practices).
-
-#### 3. Testing the Public Microagent
-
-- Test the agent with various prompts
-- Verify trigger words activate the agent correctly
-- Ensure instructions are clear and comprehensive
-- Check for potential conflicts and overlaps with existing agents
+- Test the agent with various prompts.
+- Verify trigger words activate the agent correctly.
+- Ensure instructions are clear and comprehensive.
+- Check for potential conflicts and overlaps with existing agents.
 
 #### 4. Submission Process
 
 Submit a pull request with:
 
-- The new microagent file
-- Updated documentation if needed
-- Description of the agent's purpose and capabilities
+- The new microagent file.
+- Updated documentation if needed.
+- Description of the agent's purpose and capabilities.

docs/modules/usage/prompting/microagents-repo.md

@@ -1,117 +1,31 @@
-# Repository-specific Microagents
+# General Repository Microagents
 
-## Overview
+## Purpose
 
-OpenHands can be customized to work more effectively with specific repositories by providing repository-specific context and guidelines.
+General guidelines for OpenHands to work more effectively with the repository.
 
-This section explains how to optimize OpenHands for your project.
+## Usage
 
-## Creating Repository Microagents
+These microagents are always loaded as part of the context.
 
-You can customize OpenHands' behavior for your repository by creating a `.openhands/microagents/` directory in your repository's root.
+## Frontmatter Syntax
 
-You can enhance OpenHands' performance by adding custom microagents to your repository:
+The frontmatter for this type of microagent is optional.
 
-1. For overall repository-specific instructions, create a `.openhands/microagents/repo.md` file
-2. For reusable domain knowledge triggered by keywords, add multiple `.md` files to `.openhands/microagents/knowledge/`
-3. For common workflows and tasks, create multiple `.md` files to `.openhands/microagents/tasks/`
+Frontmatter should be enclosed in triple dashes (---) and may include the following fields:
 
-Check out the [best practices](./microagents-syntax.md#markdown-content-best-practices) for formatting the content of your custom microagent.
+| Field     | Description                             | Required | Default        |
+|-----------|-----------------------------------------|----------|----------------|
+| `agent`   | The agent this microagent applies to    | No       | 'CodeActAgent' |
 
-Keep in mind that loaded microagents take up space in the context window. It's crucial to strike a balance between the additional context provided by microagents and the instructions provided in the user's inputs.
-
-Note that you can use OpenHands to create new microagents. The public microagent [`add_agent`](https://github.com/All-Hands-AI/OpenHands/blob/main/microagents/knowledge/add_agent.md) is loaded to all OpenHands instance and can support you on this.
-
-## Types of Microagents
-
-OpenHands supports three primary types of microagents, each with specific purposes and features to enhance agent performance:
-
-- [repository](#repository-microagents)
-- [knowledge](#knowledge-microagents)
-- [tasks](#tasks-microagents)
-
-The standard directory structure within a repository is:
-
-- One main `repo.md` file containing repository-specific instructions
-- Additional `Knowledge` agents in `.openhands/microagents/knowledge/` directory
-- Additional `Task` agents in `.openhands/microagents/tasks/` directory
-
-When processing the `.openhands/microagents/` directory, OpenHands will recursively scan all subfolders and process any `.md` files (except `README.md`) it finds. The system determines the microagent type based on the `type` field in the YAML frontmatter, not by the file's location. However, for organizational clarity, it's recommended to follow the standard directory structure.
-
-### Repository Microagents
-
-The `Repository` microagent is loaded specifically from `.openhands/microagents/repo.md` and serves as the main
-repository-specific instruction file. This single file is automatically loaded whenever OpenHands works with that repository
-without requiring any keyword matching or explicit call from the user.
-
-OpenHands does not support multiple `repo.md` files in different locations or multiple microagents with type `repo`.
-
-If you need to organize different types of repository information, the recommended approach is to use a single `repo.md` file with well-structured sections rather than trying to create multiple microagents with the type `repo`.
-
-The best practice is to include project-specific instructions, team practices, coding standards, and architectural guidelines that are relevant for **all** prompts in that repository.
-
-Example structure:
+## Example
 
+General repository microagent file example located at `.openhands/microagents/repo.md`:
 ```
-your-repository/
-└── .openhands/
-    └── microagents/
-        └── repo.md    # Repository-specific instructions
+This project is a TODO application that allows users to track TODO items.
+
+To set it up, you can run `npm run build`.
+Always make sure the tests are passing before committing changes. You can run the tests by running `npm run test`.
 ```
 
-[See the example in the official OpenHands repository](https://github.com/All-Hands-AI/OpenHands/blob/main/.openhands/microagents/repo.md?plain=1)
-
-### Knowledge Microagents
-
-Knowledge microagents provide specialized domain expertise:
-
-- Recommended to be located in `.openhands/microagents/knowledge/`
-- Triggered by specific keywords in conversations
-- Contain expertise on tools, languages, frameworks, and common practices
-
-Use knowledge microagents to trigger additional context relevant to specific technologies, tools, or workflows. For example, mentioning "git" in your conversation will automatically trigger git-related expertise to help with Git operations.
-
-Examples structure:
-
-```
-your-repository/
-└── .openhands/
-    └── microagents/
-        └── knowledge/
-            └── git.md
-            └── docker.md
-            └── python.md
-            └── ...
-        └── repo.md
-```
-
-You can find several real examples of `Knowledge` microagents in the [offical OpenHands repository](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/knowledge)
-
-### Tasks Microagents
-
-Task microagents guide users through interactive workflows:
-
-- Recommended to be located in `.openhands/microagents/tasks/`
-- Provide step-by-step processes for common development tasks
-- Accept inputs and adapt to different scenarios
-- Ensure consistent outcomes for complex operations
-
-Task microagents are a convenient way to store multi-step processes you perform regularly. For instance, you can create a `update_pr_description.md` microagent to automatically generate better pull request descriptions based on code changes.
-
-Examples structure:
-
-```
-your-repository/
-└── .openhands/
-    └── microagents/
-        └── tasks/
-            └── update_pr_description.md
-            └── address_pr_comments.md
-            └── get_test_to_pass.md
-            └── ...
-        └── knowledge/
-            └── ...
-        └── repo.md
-```
-
-You can find several real examples of `Tasks` microagents in the [offical OpenHands repository](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/tasks)
+[See more examples of general repository microagents here.](https://github.com/All-Hands-AI/OpenHands/tree/main/.openhands/microagents)

docs/modules/usage/prompting/microagents-syntax.md

@@ -1,128 +0,0 @@
-# Microagents Syntax
-
-Microagents are defined using markdown files with YAML frontmatter that specify their behavior, triggers, and capabilities.
-
-Find below a comprehensive description of the frontmatter syntax and other details about how to use each type of microagent available at OpenHands.
-
-## Frontmatter Schema
-
-Every microagent requires a YAML frontmatter section at the beginning of the file, enclosed by triple dashes (`---`). The fields are:
-
-| Field      | Description                                        | Required                 | Used By          |
-| ---------- | -------------------------------------------------- | ------------------------ | ---------------- |
-| `name`     | Unique identifier for the microagent               | Yes                      | All types        |
-| `type`     | Type of microagent: `repo`, `knowledge`, or `task` | Yes                      | All types        |
-| `version`  | Version number (Semantic versioning recommended)   | Yes                      | All types        |
-| `agent`    | The agent type (typically `CodeActAgent`)          | Yes                      | All types        |
-| `author`   | Creator of the microagent                          | No                       | All types        |
-| `triggers` | List of keywords that activate the microagent      | Yes for knowledge agents | Knowledge agents |
-| `inputs`   | Defines required user inputs for task execution    | Yes for task agents      | Task agents      |
-
-## Core Fields
-
-### `agent`
-
-**Purpose**: Specifies which agent implementation processes the microagent (typically `CodeActAgent`).
-
-- Defines a single agent responsible for processing the microagent
-- Must be available in the OpenHands system (see the [agent hub](https://github.com/All-Hands-AI/OpenHands/tree/main/openhands/agenthub))
-- If the specified agent is not active, the microagent will not be used
-
-### `triggers`
-
-**Purpose**: Defines keywords that activate the `knowledge` microagent.
-
-**Example**:
-
-```yaml
-triggers:
-  - kubernetes
-  - k8s
-  - docker
-  - security
-  - containers cluster
-```
-
-**Key points**:
-
-- Can include both single words and multi-word phrases
-- Case-insensitive matching is typically used
-- More specific triggers (like "docker compose") prevent false activations
-- Multiple triggers increase the chance of activation in relevant contexts
-- Unique triggers like "flarglebargle" can be used for testing or special functionality
-- Triggers should be carefully chosen to avoid unwanted activations or conflicts with other microagents
-- Common terms used in many conversations may cause the microagent to be activated too frequently
-
-When using multiple triggers, the microagent will be activated if any of the trigger words or phrases appear in the
-conversation.
-
-### `inputs`
-
-**Purpose**: Defines parameters required from the user when a `task` microagent is activated.
-
-**Schema**:
-
-```yaml
-inputs:
-  - name: INPUT_NAME # Used with {{ INPUT_NAME }}
-    description: 'Description of what this input is for'
-    required: true # Optional, defaults to true
-```
-
-**Key points**:
-
-- The `name` and `description` properties are required for each input
-- The `required` property is optional and defaults to `true`
-- Input values are referenced in the microagent body using double curly braces (e.g., `{{ INPUT_NAME }}`)
-- All inputs defined will be collected from the user before the task microagent executes
-
-**Variable Usage**: Reference input values using double curly braces `{{ INPUT_NAME }}`.
-
-## Example Formats
-
-### Repository Microagent
-
-Repository microagents provide context and guidelines for a specific repository.
-
-- Located at: `.openhands/microagents/repo.md`
-- Automatically loaded when working with the repository
-- Only one per repository
-
-The `Repository` microagent is loaded specifically from `.openhands/microagents/repo.md` and serves as the main
-repository-specific instruction file. This single file is automatically loaded whenever OpenHands works with that repository
-without requiring any keyword matching or explicit call from the user.
-
-[See the example in the official OpenHands repository](https://github.com/All-Hands-AI/OpenHands/blob/main/.openhands/microagents/repo.md?plain=1)
-
-### Knowledge Microagent
-
-Provides specialized domain expertise triggered by keywords.
-
-You can find several real examples of `Knowledge` microagents in the [offical OpenHands repository](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/knowledge)
-
-### Task Microagent
-
-When explicitly asked by the user, will guide through interactive workflows with specific inputs.
-
-You can find several real examples of `Tasks` microagents in the [offical OpenHands repository](https://github.com/All-Hands-AI/OpenHands/tree/main/microagents/tasks)
-
-## Markdown Content Best Practices
-
-After the frontmatter, compose the microagent body using Markdown syntax. Examples of elements you can include are:
-
-- Clear, concise instructions outlining the microagent's purpose and responsibilities
-- Specific guidelines and constraints the microagent should adhere to
-- Relevant code snippets and practical examples to illustrate key points
-- Step-by-step procedures for task agents, guiding users through workflows
-
-**Design Tips**:
-
-- Keep microagents focused with a clear purpose
-- Provide specific guidelines rather than general advice
-- Use distinctive triggers for knowledge agents
-- Keep content concise to minimize context window usage
-- Break large microagents into smaller, focused ones
-
-Aim for clarity, brevity, and practicality in your writing. Use formatting like bullet points, code blocks, and emphasis to enhance readability and comprehension.
-
-Remember that balancing microagents details with user input space is important for maintaining effective interactions.

docs/modules/usage/runtimes/docker.md

@@ -35,8 +35,8 @@ A useful feature is the ability to connect to your local filesystem. To mount yo
 Be careful! There's nothing stopping the OpenHands agent from deleting or modifying
 any files that are mounted into its workspace.
 
-This setup can cause some issues with file permissions (hence the `SANDBOX_USER_ID` variable)
-but seems to work well on most systems.
+The `-e SANDBOX_USER_ID=$(id -u)` is passed to the Docker command to ensure the sandbox user matches the host user’s
+permissions. This prevents the agent from creating root-owned files in the mounted workspace.
 
 ## Hardened Docker Installation
 

docs/modules/usage/troubleshooting/troubleshooting.md

@@ -20,3 +20,18 @@ Try these in order:
 * If using Docker Desktop, ensure `Settings > Advanced > Allow the default Docker socket to be used` is enabled.
 * Depending on your configuration you may need `Settings > Resources > Network > Enable host networking` enabled in Docker Desktop.
 * Reinstall Docker Desktop.
+
+### Permission Error
+
+**Description**
+
+On initial prompt, an error is seen with `Permission Denied` or `PermissionError`.
+
+**Resolution**
+
+* Check if the `~/.openhands-state` is owned by `root`. If so, you can:
+  * Change the directory's ownership: `sudo chown <user>:<user> ~/.openhands-state`.
+  * or update permissions on the directory: `sudo chmod 777 ~/.openhands-state`
+  * or delete it if you don’t need previous data. OpenHands will recreate it. You'll need to re-enter LLM settings.
+* If mounting a local directory, ensure your `WORKSPACE_BASE` has the necessary permissions for the user running
+  OpenHands.

docs/package.json

@@ -4,23 +4,27 @@
   "private": true,
   "scripts": {
     "docusaurus": "docusaurus",
-    "start": "docusaurus start",
-    "build": "docusaurus build",
+    "start": "node generate-swagger-ui.js && docusaurus start",
+    "build": "node generate-swagger-ui.js && docusaurus build",
     "swizzle": "docusaurus swizzle",
     "deploy": "docusaurus deploy",
     "clear": "docusaurus clear",
     "serve": "docusaurus serve",
     "write-translations": "docusaurus write-translations",
     "write-heading-ids": "docusaurus write-heading-ids",
-    "typecheck": "tsc"
+    "typecheck": "tsc",
+    "generate-swagger-ui": "node generate-swagger-ui.js"
   },
+  "// Note": "The OpenAPI spec is stored in docs/static/openapi.json so it's accessible at /openapi.json in the deployed site",
   "dependencies": {
     "@docusaurus/core": "^3.7.0",
     "@docusaurus/plugin-content-pages": "^3.7.0",
     "@docusaurus/preset-classic": "^3.7.0",
     "@docusaurus/theme-mermaid": "^3.7.0",
     "@mdx-js/react": "^3.1.0",
+    "@node-rs/jieba": "^2.0.1",
     "clsx": "^2.0.0",
+    "docusaurus-lunr-search": "^3.6.0",
     "prism-react-renderer": "^2.4.1",
     "react": "^19.1.0",
     "react-dom": "^19.1.0",
@@ -31,6 +35,8 @@
     "@docusaurus/module-type-aliases": "^3.5.1",
     "@docusaurus/tsconfig": "^3.7.0",
     "@docusaurus/types": "^3.5.1",
+    "swagger-cli": "^4.0.4",
+    "swagger-ui-dist": "^5.21.0",
     "typescript": "~5.8.3"
   },
   "browserslist": {
@@ -47,5 +53,6 @@
   },
   "engines": {
     "node": ">=18.0"
-  }
+  },
+  "packageManager": "npm@10.5.0"
 }

docs/sidebars.ts

@@ -27,7 +27,11 @@ const sidebars: SidebarsConfig = {
           label: 'Openhands Cloud',
           id: 'usage/cloud/openhands-cloud',
         },
-
+        {
+          type: 'doc',
+          label: 'Cloud API',
+          id: 'usage/cloud/cloud-api',
+        },
         {
           type: 'doc',
           label: 'Cloud GitHub Resolver',
@@ -66,18 +70,18 @@ const sidebars: SidebarsConfig = {
             },
             {
               type: 'doc',
-              label: 'Repository-specific',
+              label: 'General Repository Microagents',
               id: 'usage/prompting/microagents-repo',
             },
             {
               type: 'doc',
-              label: 'Public',
-              id: 'usage/prompting/microagents-public',
+              label: 'Keyword-Triggered Microagents',
+              id: 'usage/prompting/microagents-keyword',
             },
             {
               type: 'doc',
-              label: 'Syntax',
-              id: 'usage/prompting/microagents-syntax',
+              label: 'Global Microagents',
+              id: 'usage/prompting/microagents-public',
             },
           ],
         },

docs/src/components/HomepageHeader/HomepageHeader.tsx

@@ -45,7 +45,6 @@ export function HomepageHeader() {
         <div align="center" className="header-links">
           <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/openhands-ai/shared_invite/zt-2ngejmfw6-9gW4APWOC9XUp1n~SiQ6iw"><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>
@@ -53,7 +52,7 @@ export function HomepageHeader() {
           <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://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>
+          <a href="https://docs.google.com/spreadsheets/d/1wOUdFCMyY6Nt0AIqF705KN4JKOWgeI4wUGUP60krXXs/edit?gid=0#gid=0"><img src="https://img.shields.io/badge/Benchmark%20score-000?logoColor=FFE165&logo=huggingface&style=for-the-badge" alt="Evaluation Benchmark Score" /></a>
         </div>
       </div>
     </div>

docs/static/README.md

@@ -0,0 +1,15 @@
+# Static Files for OpenHands Documentation
+
+This directory contains static files that are copied directly to the build output of the Docusaurus documentation.
+
+## OpenAPI Specification
+
+The `openapi.json` file in this directory is the OpenAPI specification for the OpenHands API. It is copied to the build output and is accessible at `/openapi.json` in the deployed site.
+
+This file is used by the Swagger UI interface, which is accessible at `/swagger-ui/` in the deployed site.
+
+## Why is the OpenAPI spec in the static directory?
+
+The OpenAPI specification is placed in the static directory so that it's accessible at a predictable URL in the deployed site. This allows the Swagger UI to reference it directly.
+
+We only need one copy of the OpenAPI spec file, which is this one in the static directory.

evaluation/benchmarks/commit0/run_infer.py

@@ -131,9 +131,9 @@ def get_config(
         )
     )
     agent_config = AgentConfig(
-        codeact_enable_jupyter=False,
-        codeact_enable_browsing=RUN_WITH_BROWSING,
-        codeact_enable_llm_editor=False,
+        enable_jupyter=False,
+        enable_browsing=RUN_WITH_BROWSING,
+        enable_llm_editor=False,
     )
     config.set_agent_config(agent_config)
     return config

evaluation/benchmarks/discoverybench/run_infer.py

@@ -79,8 +79,8 @@ def get_config(
     agent_config.enable_prompt_extensions = False
     agent_config = AgentConfig(
         function_calling=False,
-        codeact_enable_jupyter=True,
-        codeact_enable_browsing_delegate=True,
+        enable_jupyter=True,
+        enable_browsing=True,
     )
     config.set_agent_config(agent_config)
     return config

evaluation/benchmarks/lca_ci_build_repair/.gitignore

@@ -0,0 +1 @@
+config.yaml

evaluation/benchmarks/lca_ci_build_repair/README.MD

@@ -0,0 +1,35 @@
+# CI Builds Repair Benchmark Integration
+
+This module integrates the CI Builds Repair benchmark developed by [JetBrains-Research](https://github.com/JetBrains-Research/lca-baselines/tree/main/ci-builds-repair/ci-builds-repair-benchmark).
+
+For more information, refer to the [GitHub repository](https://github.com/JetBrains-Research/lca-baselines/tree/main/ci-builds-repair/ci-builds-repair-benchmark) and the associated [research paper](https://arxiv.org/abs/2406.11612).
+See notice below for details
+
+## Setup
+
+Before running any scripts, make sure to configure the benchmark by setting up `config.yaml`.
+This benchmark pushes to JetBrains' private GitHub repository. You will to request a `token_gh` provided by their team, to run this benchmark.
+
+## Inference
+
+To run inference with your model:
+
+```bash
+./evaluation/benchmarks/lca_ci_build_repair/scripts/run_infer.sh llm.yourmodel
+```
+
+## Evaluation
+
+To evaluate the predictions:
+
+```bash
+./evaluation/benchmarks/lca_ci_build_repair/scripts/eval_infer.sh predictions_path_containing_output
+```
+
+## Results
+The benchmark contains 68 instances, we skip instances #126 and #145, and only run 66 instances due to dockerization errors.
+
+Due to running in live GitHub machines, the benchmark is sensitive to the date it is run. Even the golden patches in the dataset might present failures due to updates.
+For example, on 2025-04-09, running the benchmark against the golden patches gave 57/67 successes, with 1 job left in the waiting list.
+
+On 2025-04-10, running the benchmark full with OH and no oracle, 37 succeeded. That is 54% of the complete set of 68 instances and 64% of the 57 that succeed with golden patches.

evaluation/benchmarks/lca_ci_build_repair/config_template.yaml

@@ -0,0 +1,11 @@
+LCA_PATH: path #where to clone lca-ci rep
+model_name: OpenHands
+benchmark_owner: ICML-25-BenchName-builds-repair
+token_gh: your_token
+#for lca-ci-repo
+repos_folder: /path/to/repos # here the cloned repos would be stored
+out_folder: /out/folder # here the result files would be stored
+data_cache_dir: /data/cache/dir/ # here the cached dataset would be stored
+username_gh: username-gh # your GitHub username
+# test_username: test_user # username that would be displayed in the benchmark. Optional. If ommitted, username_gh would be used
+language: Python # dataset language (now only Python is available)

evaluation/benchmarks/lca_ci_build_repair/eval_infer.py

@@ -0,0 +1,242 @@
+"""Implements evaluation on JetBrains CI builds repair baselines
+
+Please see https://github.com/JetBrains-Research/lca-baselines/tree/main/ci-builds-repair
+and https://huggingface.co/datasets/JetBrains-Research/lca-ci-builds-repair
+
+TODOs:
+- Add more flags
+"""
+
+import json
+import os
+from pathlib import Path
+
+import ruamel.yaml
+
+from evaluation.utils.shared import (
+    EvalMetadata,
+    get_default_sandbox_config_for_eval,
+    make_metadata,
+)
+from openhands.core.config import (
+    AppConfig,
+    LLMConfig,
+    get_parser,
+    load_app_config,
+)
+from openhands.core.logger import openhands_logger as logger
+from openhands.core.main import create_runtime
+from openhands.events.action import CmdRunAction
+from openhands.events.observation import CmdOutputObservation
+from openhands.runtime.base import Runtime
+from openhands.utils.async_utils import call_async_from_sync
+
+
+def get_config(
+    metadata: EvalMetadata,
+) -> AppConfig:
+    sandbox_config = get_default_sandbox_config_for_eval()
+    sandbox_config.base_container_image = 'python:3.12-bookworm'
+    config = AppConfig(
+        default_agent=metadata.agent_class,
+        run_as_openhands=False,
+        runtime='docker',
+        max_iterations=metadata.max_iterations,
+        sandbox=sandbox_config,
+        # do not mount workspace
+        workspace_base=None,
+        workspace_mount_path=None,
+    )
+    config.set_llm_config(metadata.llm_config)
+    agent_config = config.get_agent_config(metadata.agent_class)
+    agent_config.enable_prompt_extensions = False
+    return config
+
+
+config = load_app_config()
+
+
+def load_bench_config():
+    script_dir = os.path.dirname(
+        os.path.abspath(__file__)
+    )  # Get the absolute path of the script
+    config_path = os.path.join(script_dir, 'config.yaml')
+    yaml = ruamel.yaml.YAML(typ='rt')
+    with open(config_path, 'r') as file:
+        return yaml.load(file)
+
+
+bench_config = load_bench_config()
+
+
+def run_eval(
+    runtime: Runtime,
+):
+    """Run the evaluation and create report"""
+    logger.info(f"{'-' * 50} BEGIN Runtime Initialization Fn {'-' * 50}")
+    obs: CmdOutputObservation
+
+    lca_path = bench_config['LCA_PATH']
+    lca_ci_path = os.path.join(
+        lca_path, 'lca-baselines', 'ci-builds-repair', 'ci-builds-repair-benchmark'
+    )
+
+    model_name = bench_config['model_name']
+
+    action = CmdRunAction(command=f'mkdir {lca_path}')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    assert obs.exit_code == 0
+
+    action = CmdRunAction(command=f'cd {lca_path}')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    assert obs.exit_code == 0
+
+    lca_repo_url = 'https://github.com/juanmichelini/lca-baselines'
+    action = CmdRunAction(command=f'git clone {lca_repo_url}')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    assert obs.exit_code == 0
+
+    action = CmdRunAction(command=f'cd {lca_ci_path}')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    assert obs.exit_code == 0
+
+    action = CmdRunAction(command='git switch open-hands-integration')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    assert obs.exit_code == 0
+
+    script_dir = os.path.dirname(
+        os.path.abspath(__file__)
+    )  # Get the absolute path of the script
+    config_path = os.path.join(script_dir, 'config.yaml')
+    runtime.copy_to(config_path, lca_ci_path)
+
+    token_gh = bench_config['token_gh']
+    commandf = f'export TOKEN_GH={token_gh}'
+    action = CmdRunAction(command=commandf)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+
+    action = CmdRunAction(command='poetry install')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+
+    # Set up the task environment
+    commandf = f'poetry run python run_eval_jobs.py --model-name "{model_name}" --config-path "{lca_ci_path}/config.yaml" --job-ids-file "/tmp/output_lca.jsonl" --result-filename "testfile.jsonl"  > /tmp/single_output.txt'
+    action = CmdRunAction(command=commandf)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    logger.info(f'run_eval_jobs.py gave {obs.content} !')
+    # assert obs.exit_code == 0
+
+    commandf = 'cat /tmp/single_output.txt'
+    action = CmdRunAction(command=commandf)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    logger.info(f' {commandf} gave {obs.content}!')
+
+    testfile_path = os.path.join(bench_config['out_folder'], 'testfile.jsonl')
+    commandf = f'cat {testfile_path}'
+    action = CmdRunAction(command=commandf)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    report_str = obs.content
+
+    logger.info(f"{'-' * 50} END Runtime Initialization Fn {'-' * 50}")
+    return report_str
+
+
+def process_predictions(predictions_path: str):
+    output_path = Path(predictions_path)
+    if output_path.suffix != '.jsonl':
+        raise ValueError('output_path must end in .jsonl')
+
+    output_lca_path = output_path.with_name(output_path.stem + '_lca.jsonl')
+
+    with output_path.open() as infile, output_lca_path.open('w') as outfile:
+        for line in infile:
+            data = json.loads(line)
+            json.dump(data.get('test_result'), outfile)
+            outfile.write('\n')
+
+    return str(output_lca_path)
+
+
+if __name__ == '__main__':
+    parser = get_parser()
+    parser.add_argument(
+        '-s',
+        '--eval-split',
+        type=str,
+        default='test',
+        choices=['test'],
+        help='data split to evaluate on, must be test',
+    )
+    parser.add_argument(
+        '--predictions-path',
+        type=str,
+        help='Path to the directory containing the output.jsonl with the predictions.',
+    )
+    args, _ = parser.parse_known_args()
+
+    data_split = args.eval_split
+
+    llm_config = LLMConfig(model='dummy_model')
+
+    metadata = make_metadata(
+        llm_config,
+        f'jetbrains-lca-ci--{data_split}',
+        args.agent_cls,
+        args.max_iterations,
+        args.eval_note,
+        args.predictions_path,
+    )
+
+    # prepare image
+    config = get_config(metadata)
+    runtime = create_runtime(config)
+    call_async_from_sync(runtime.connect)
+    logger.info('Converting output.jsonl into output_lca.jsonl')
+    predictions_lca_path = process_predictions(
+        os.path.join(args.predictions_path, 'output.jsonl')
+    )
+    runtime.copy_to(predictions_lca_path, '/tmp')
+
+    # get results
+    results_str = run_eval(runtime)
+    results_path = os.path.join(args.predictions_path, 'results.jsonl')
+    with open(results_path, 'w') as file:
+        file.write(results_str)
+    logger.info(f'Saved results to {results_path}')
+
+    # make a summary
+    resolved_instances = []
+    unresolved_instances = []
+    for line in results_str.strip().splitlines():
+        data = json.loads(line)
+        conclusion = data.get('conclusion')
+        if conclusion == 'success':
+            resolved_instances.append(data)
+        elif conclusion == 'failure':
+            unresolved_instances.append(data)
+
+    completed_instances = resolved_instances + unresolved_instances
+
+    report = {
+        'success': len(resolved_instances),
+        'failure': len(unresolved_instances),
+        'resolved_instances': resolved_instances,
+        'unresolved_instances': unresolved_instances,
+        'completed_instances': completed_instances,
+    }
+
+    print(f'Results: {report}')
+    report_path = os.path.join(args.predictions_path, 'report.jsonl')
+    with open(report_path, 'w') as out_f:
+        out_f.write(json.dumps(report) + '\n')
+
+    logger.info(f'Saved report of results in swebench format to {report_path}')

evaluation/benchmarks/lca_ci_build_repair/run_infer.py

@@ -0,0 +1,406 @@
+"""Implements inference on JetBrains CI builds repair baselines
+
+Please see https://github.com/JetBrains-Research/lca-baselines/tree/main/ci-builds-repair
+and https://huggingface.co/datasets/JetBrains-Research/lca-ci-builds-repair
+
+TODOs:
+- Add EXP_NAME
+"""
+
+import asyncio
+import json
+import os
+from typing import Any
+
+import pandas as pd
+import ruamel.yaml
+from datasets import load_dataset
+
+from evaluation.utils.shared import (
+    EvalMetadata,
+    EvalOutput,
+    codeact_user_response,
+    compatibility_for_eval_history_pairs,
+    get_default_sandbox_config_for_eval,
+    make_metadata,
+    prepare_dataset,
+    reset_logger_for_multiprocessing,
+    run_evaluation,
+)
+from openhands.controller.state.state import State
+from openhands.core.config import (
+    AppConfig,
+    get_llm_config_arg,
+    get_parser,
+    load_app_config,
+)
+from openhands.core.logger import openhands_logger as logger
+from openhands.core.main import create_runtime, run_controller
+from openhands.events.action import CmdRunAction, MessageAction
+from openhands.events.observation import CmdOutputObservation
+from openhands.runtime.base import Runtime
+from openhands.utils.async_utils import call_async_from_sync
+
+
+def get_config(
+    metadata: EvalMetadata,
+) -> AppConfig:
+    sandbox_config = get_default_sandbox_config_for_eval()
+    sandbox_config.base_container_image = 'python:3.12-bookworm'
+    config = AppConfig(
+        default_agent=metadata.agent_class,
+        run_as_openhands=False,
+        runtime='docker',
+        max_iterations=metadata.max_iterations,
+        sandbox=sandbox_config,
+        # do not mount workspace
+        workspace_base=None,
+        workspace_mount_path=None,
+    )
+    config.set_llm_config(metadata.llm_config)
+    agent_config = config.get_agent_config(metadata.agent_class)
+    agent_config.enable_prompt_extensions = False
+    return config
+
+
+config = load_app_config()
+
+
+def load_bench_config():
+    script_dir = os.path.dirname(
+        os.path.abspath(__file__)
+    )  # Get the absolute path of the script
+    config_path = os.path.join(script_dir, 'config.yaml')
+    yaml = ruamel.yaml.YAML(typ='rt')
+    with open(config_path, 'r') as file:
+        return yaml.load(file)
+
+
+bench_config = load_bench_config()
+
+AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {
+    'CodeActAgent': codeact_user_response,
+}
+
+AGENT_CLS_TO_INST_SUFFIX = {
+    'CodeActAgent': 'When you think you have completed the task, please finish the interaction using the "finish" tool.\n'
+}
+
+
+def initialize_runtime(
+    runtime: Runtime,
+    instance: pd.Series,
+):
+    """Initialize the runtime for the agent.
+
+    This function is called before the runtime is used to run the agent.
+    """
+    logger.info(f"{'-' * 50} BEGIN Runtime Initialization Fn {'-' * 50}")
+    obs: CmdOutputObservation
+
+    lca_path = bench_config['LCA_PATH']
+    lca_ci_path = os.path.join(
+        lca_path, 'lca-baselines', 'ci-builds-repair', 'ci-builds-repair-benchmark'
+    )
+
+    repo_name = instance['repo_name']
+    repos_path = bench_config['repos_folder']
+    repo_owner = instance['repo_owner']
+    repo_path = os.path.join(repos_path, f'{repo_owner}__{repo_name}')
+    model_name = bench_config['model_name']
+
+    action = CmdRunAction(command=f'mkdir {lca_path}')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    assert obs.exit_code == 0
+
+    action = CmdRunAction(command=f'cd {lca_path}')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    assert obs.exit_code == 0
+
+    lca_repo_url = 'https://github.com/juanmichelini/lca-baselines'
+    action = CmdRunAction(command=f'git clone {lca_repo_url}')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    assert obs.exit_code == 0
+
+    action = CmdRunAction(command=f'cd {lca_ci_path}')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    assert obs.exit_code == 0
+
+    action = CmdRunAction(command='git switch open-hands-integration')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    assert obs.exit_code == 0
+
+    script_dir = os.path.dirname(
+        os.path.abspath(__file__)
+    )  # Get the absolute path of the script
+    config_path = os.path.join(script_dir, 'config.yaml')
+    with open(config_path, 'r') as file:
+        config_as_text = file.read()
+
+    commandf = f"echo '{config_as_text}' > config.yaml"
+    action = CmdRunAction(command=commandf)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+
+    token_gh = bench_config['token_gh']
+    commandf = f'export TOKEN_GH={token_gh}'
+    action = CmdRunAction(command=commandf)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+
+    action = CmdRunAction(command='poetry install')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+
+    # Set up the task environment
+    commandf = f'poetry run python run_get_datapoint.py --model-name {model_name} --id {instance["id"]} > branch_name.txt'
+    action = CmdRunAction(command=commandf)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    if obs.exit_code != 0:
+        print(f'run_get_datapoint.py failed at {instance["id"]} with {obs.content}')
+    assert obs.exit_code == 0
+
+    commandf = 'cat branch_name.txt'
+    action = CmdRunAction(command=commandf)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    bench_config['user_branch_name'] = obs.content
+
+    # Navigate to the task's code path
+    action = CmdRunAction(command=f'cd {repo_path}')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+
+    logger.info(f"{'-' * 50} END Runtime Initialization Fn {'-' * 50}")
+
+
+def complete_runtime(
+    runtime: Runtime,
+    instance: pd.Series,
+) -> dict[str, Any]:
+    """Complete the runtime for the agent.
+
+    This function is called before the runtime is used to run the agent.
+    If you need to do something in the sandbox to get the correctness metric after
+    the agent has run, modify this function.
+    """
+    logger.info(f"{'-' * 50} BEGIN Runtime Completion Fn {'-' * 50}")
+    obs: CmdOutputObservation
+
+    model_name = bench_config['model_name']
+
+    lca_path = bench_config['LCA_PATH']
+    lca_ci_path = os.path.join(
+        lca_path, 'lca-baselines', 'ci-builds-repair', 'ci-builds-repair-benchmark'
+    )
+
+    user_branch_name = bench_config['user_branch_name']
+
+    token_gh = bench_config['token_gh']
+    commandf = f'export TOKEN_GH={token_gh}'
+    action = CmdRunAction(command=commandf)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+
+    # Navigate to the lca-baseslines scripts path
+    action = CmdRunAction(command=f'cd {lca_ci_path}')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    assert obs.exit_code == 0
+
+    commandf = f'poetry run python run_push_datapoint.py --id {instance["id"]} --model-name {model_name} --user-branch-name {user_branch_name} > single_output.json'
+    logger.info(f'Running push script: {commandf}')
+    action = CmdRunAction(command=commandf)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    # assert obs.exit_code == 0
+
+    commandf = 'cat single_output.json'
+    action = CmdRunAction(command=commandf)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    result = json.loads(obs.content)
+
+    logger.info(f"{'-' * 50} END Runtime Completion Fn {'-' * 50}")
+
+    return result
+
+
+def process_instance(instance: Any, metadata: EvalMetadata, reset_logger: bool = True):
+    config = get_config(metadata)
+
+    # Setup the logger properly, so you can run multi-processing to parallelize the evaluation
+    if reset_logger:
+        log_dir = os.path.join(metadata.eval_output_dir, 'infer_logs')
+        reset_logger_for_multiprocessing(logger, instance['instance_id'], log_dir)
+    else:
+        logger.info(f'Starting evaluation for instance {instance["instance_id"]}.')
+
+    repo_name = instance['repo_name']
+    repo_workflow = instance['workflow_path']
+    repo_logs = instance['logs']
+    repos_path = bench_config['repos_folder']
+    repo_owner = instance['repo_owner']
+    repo_path = os.path.join(repos_path, f'{repo_owner}__{repo_name}')
+
+    # Prepare the task instruction
+    instruction_no_oracle = f"""
+<uploaded_files>
+{repo_path}
+</uploaded_files>
+
+I've uploaded a python code repository in the directory {repo_path}, Consider the following issue:
+
+<issue_description>
+The repository must pass the CI workflow {repo_workflow}.
+but it gave the following error
+{repo_logs}
+</issue_description>
+
+Can you help me implement the necessary changes to the repository so that the requirements specified in the <issue_description> are met?
+I've already taken care of all changes to any of the test files described in the <issue_description>. This means you DON'T have to modify the testing logic or any of the tests in any way!
+Also the development Python environment is already set up for you (i.e., all dependencies already installed), so you don't need to install other packages.
+Your task is to make the minimal changes to non-test files in the {repo_path} directory to ensure the <issue_description> is satisfied.
+
+Follow these phases to resolve the issue:
+
+Phase 1. READING: read the problem and reword it in clearer terms
+   1.1 If there are code or config snippets. Express in words any best practices or conventions in them.
+   1.2 Hightlight message errors, method names, variables, file names, stack traces, and technical details.
+   1.3 Explain the problem in clear terms.
+   1.4 Enumerate the steps to reproduce the problem.
+   1.5 Hightlight any best practices to take into account when testing and fixing the issue
+
+Phase 2. RUNNING: install and run the tests on the repository
+   2.1 Follow the readme
+   2.2 Install the environment and anything needed
+   2.2 Iterate and figure out how to run the tests
+
+Phase 3. EXPLORATION: find the files that are related to the problem and possible solutions
+   3.1 Use `grep` to search for relevant methods, classes, keywords and error messages.
+   3.2 Identify all files related to the problem statement.
+   3.3 Propose the methods and files to fix the issue and explain why.
+   3.4 From the possible file locations, select the most likely location to fix the issue.
+
+Phase 4. TEST CREATION: before implementing any fix, create a script to reproduce and verify the issue.
+   4.1 Look at existing test files in the repository to understand the test format/structure.
+   4.2 Create a minimal reproduction script that reproduces the located issue.
+   4.3 Run the reproduction script to confirm you are reproducing the issue.
+   4.4 Adjust the reproduction script as necessary.
+
+Phase 5. FIX ANALYSIS: state clearly the problem and how to fix it
+   5.1 State clearly what the problem is.
+   5.2 State clearly where the problem is located.
+   5.3 State clearly how the test reproduces the issue.
+   5.4 State clearly the best practices to take into account in the fix.
+   5.5 State clearly how to fix the problem.
+
+Phase 6. FIX IMPLEMENTATION: Edit the source code to implement your chosen solution.
+   6.1 Make minimal, focused changes to fix the issue.
+
+Phase 7. VERIFICATION: Test your implementation thoroughly.
+   7.1 Run your reproduction script to verify the fix works.
+   7.2 Add edge cases to your test script to ensure comprehensive coverage.
+   7.3 Run existing tests related to the modified code to ensure you haven't broken anything. Run any tests in the repository related to:
+     7.2.1 The issue you are fixing
+     7.2.2 The files you modified
+     7.2.3 The functions you changed
+   7.4 If any tests fail, revise your implementation until all tests pass
+
+Phase 8. REVIEW: Carefully re-read the problem description and compare your changes with the base commit {instance["sha_fail"]}.
+   8.1 Ensure you've fully addressed all requirements.
+
+Once all phases are done, announce: 'Agent Task Complete'.
+Be thorough in your exploration, testing, and reasoning. It's fine if your thinking process is lengthy - quality and completeness are more important than brevity.
+"""
+    runtime = create_runtime(config)
+    call_async_from_sync(runtime.connect)
+    initialize_runtime(runtime, instance)
+
+    # Run the agent
+    state: State | None = asyncio.run(
+        run_controller(
+            config=config,
+            initial_user_action=MessageAction(content=instruction_no_oracle),
+            runtime=runtime,
+            fake_user_response_fn=AGENT_CLS_TO_FAKE_USER_RESPONSE_FN.get(
+                metadata.agent_class
+            ),
+        )
+    )
+    assert state is not None
+    metrics = state.metrics.get() if state.metrics else {}
+
+    test_result = complete_runtime(runtime, instance)
+
+    # history is now available as a stream of events, rather than list of pairs of (Action, Observation)
+    # for compatibility with the existing output format, we can remake the pairs here
+    # remove when it becomes unnecessary
+    histories = compatibility_for_eval_history_pairs(state.history)
+
+    # Save the output
+    output = EvalOutput(
+        instance_id=instance['instance_id'],
+        # instance=instance.to_dict(orient='recorods'),
+        instruction=instruction_no_oracle,
+        metadata=metadata,
+        history=histories,
+        test_result=test_result,
+        metrics=metrics,
+    )
+    return output
+
+
+if __name__ == '__main__':
+    parser = get_parser()
+    parser.add_argument(
+        '-s',
+        '--eval-split',
+        type=str,
+        default='test',
+        choices=['test'],
+        help='data split to evaluate on, must be test',
+    )
+    args, _ = parser.parse_known_args()
+
+    data_split = args.eval_split
+
+    bench = load_dataset(
+        'JetBrains-Research/lca-ci-builds-repair', split=data_split
+    ).to_pandas()
+    # todo: see why 126 is giving problems on inference
+    # todo: see why 145 is giving problems on eval
+    bench = bench[bench['id'] != 126]
+    bench = bench[bench['id'] != 145]
+    # bench = bench.iloc[0:56]
+    # add column instnace_id for compatibility with oh repo, old id column must be kept for lca repo
+    bench['instance_id'] = bench['id'].astype(str)
+
+    llm_config = None
+    if args.llm_config:
+        llm_config = get_llm_config_arg(args.llm_config)
+        # modify_params must be False for evaluation purpose, for reproducibility and accurancy of results
+        llm_config.modify_params = False
+    if llm_config is None:
+        raise ValueError(f'Could not find LLM config: --llm_config {args.llm_config}')
+
+    metadata = make_metadata(
+        llm_config,
+        f'jetbrains-lca-ci--{data_split}',
+        args.agent_cls,
+        args.max_iterations,
+        args.eval_note,
+        args.eval_output_dir,
+    )
+    output_file = os.path.join(metadata.eval_output_dir, 'output.jsonl')
+    instances = prepare_dataset(bench, output_file, args.eval_n_limit)
+
+    run_evaluation(
+        instances, metadata, output_file, args.eval_num_workers, process_instance
+    )

evaluation/benchmarks/lca_ci_build_repair/scripts/eval_infer.sh

@@ -0,0 +1,33 @@
+#!/usr/bin/env bash
+set -eo pipefail
+
+source "evaluation/utils/version_control.sh"
+
+PROCESS_FILEPATH=$1
+if [ -z "$PROCESS_FILEPATH" ]; then
+    echo "Error: PROCESS_FILEPATH is empty. Usage: ./eval_infer.sh <output_file> [instance_id] [dataset_name] [split]"
+    exit 1
+fi
+
+get_openhands_version
+
+PROCESS_FILEPATH=$(realpath $PROCESS_FILEPATH)
+echo "OPENHANDS_VERSION: $OPENHANDS_VERSION"
+echo "PROCESS_FILEPATH: $PROCESS_FILEPATH"
+
+EVAL_NOTE="$OPENHANDS_VERSION"
+if [ -n "$EXP_NAME" ]; then
+  EVAL_NOTE="$EVAL_NOTE-$EXP_NAME"
+fi
+
+function run_eval() {
+  COMMAND="poetry run python ./evaluation/benchmarks/lca_ci_build_repair/eval_infer.py \
+    --predictions-path $PROCESS_FILEPATH "
+
+  echo "RUNNING: $COMMAND"
+  # Run the command
+  eval $COMMAND
+}
+
+unset SANDBOX_ENV_GITHUB_TOKEN # prevent the agent from using the github token to push
+run_eval

evaluation/benchmarks/lca_ci_build_repair/scripts/run_infer.sh

@@ -0,0 +1,27 @@
+#!/usr/bin/env bash
+set -eo pipefail
+
+source "evaluation/utils/version_control.sh"
+
+MODEL_CONFIG=$1
+
+get_openhands_version
+
+echo "OPENHANDS_VERSION: $OPENHANDS_VERSION"
+echo "MODEL_CONFIG: $MODEL_CONFIG"
+
+EVAL_NOTE="$OPENHANDS_VERSION"
+if [ -n "$EXP_NAME" ]; then
+  EVAL_NOTE="$EVAL_NOTE-$EXP_NAME"
+fi
+
+function run_eval() {
+  COMMAND="poetry run python ./evaluation/benchmarks/lca_ci_build_repair/run_infer.py \
+    --llm-config $MODEL_CONFIG "
+
+  # Run the command
+  eval $COMMAND
+}
+
+#unset SANDBOX_ENV_GITHUB_TOKEN # prevent the agent from using the github token to push
+run_eval

evaluation/benchmarks/lca_ci_build_repair/setup.py

@@ -0,0 +1,60 @@
+"""Installs LCA CI Build Repair benchmark with scripts for OH integration."""
+
+import os
+import shutil
+import subprocess
+
+import yaml
+
+
+def setup():
+    # Read config.yaml
+    print('Reading config.yaml')
+    script_dir = os.path.dirname(
+        os.path.abspath(__file__)
+    )  # Get the absolute path of the script
+    config_path = os.path.join(script_dir, 'config.yaml')
+    with open(config_path, 'r') as f:
+        config = yaml.safe_load(f)
+
+    lca_path = config['LCA_PATH']
+    lca_ci_path = os.path.join(
+        lca_path, 'lca-baselines', 'ci-builds-repair', 'ci-builds-repair-benchmark'
+    )
+    repo_url = 'https://github.com/juanmichelini/lca-baselines'
+
+    # Clone the repository to LCA_CI_PATH
+    print(f'Cloning lca-baselines repository from {repo_url} into {lca_path}')
+    result = subprocess.run(
+        ['git', 'clone', repo_url], cwd=lca_path, capture_output=True, text=True
+    )
+    if result.returncode != 0:
+        print(f'Warning cloning repository: {result.stderr}')
+
+    # Clone the repository to LCA_CI_PATH
+    print('Switching branches')
+    result = subprocess.run(
+        ['git', 'switch', 'open-hands-integration'],
+        cwd=lca_ci_path,
+        capture_output=True,
+        text=True,
+    )
+    if result.returncode != 0:
+        print(f'Warning switching repository: {result.stderr}')
+
+    # Move and rename config_lca.yaml (overwrite if exists)
+    lca_ci_config_path = os.path.join(lca_ci_path, 'config.yaml')
+    print(f'Copying config.yaml to {lca_ci_config_path}')
+    shutil.copy(config_path, lca_ci_config_path)
+
+    # Run poetry install in LCA_CI_PATH
+    print(f"Running 'poetry install' in {lca_ci_path}")
+    result = subprocess.run(
+        ['poetry', 'install'], cwd=lca_ci_path, capture_output=True, text=True
+    )
+    if result.returncode != 0:
+        print(f'Warning during poetry install: {result.stderr}')
+
+
+if __name__ == '__main__':
+    setup()

evaluation/benchmarks/multi_swe_bench/README.md

@@ -0,0 +1,65 @@
+# Multi-swe-bench Evaluation with OpenHands
+
+## LLM Setup
+
+Please follow [here](../../README.md#setup).
+
+## Dataset Preparing
+
+Please download the [**Multi-SWE-Bench** dataset](https://huggingface.co/datasets/bytedance-research/Multi-SWE-Bench).
+And change the dataset following [script](scripts/data/data_change.py).
+
+```bash
+python evaluation/benchmarks/multi_swe_bench/scripts/data/data_change.py
+```
+
+## Docker image download
+
+Please download the multi-swe-bench dokcer images from [here](https://github.com/multi-swe-bench/multi-swe-bench?tab=readme-ov-file#run-evaluation).
+
+## Generate patch
+
+Please edit the [script](infer.sh) and run it.
+
+```bash
+bash evaluation/benchmarks/multi_swe_bench/infer.sh
+```
+
+Script variable explanation:
+
+- `models`, e.g. `llm.eval_gpt4_1106_preview`, is the config group name for your
+LLM settings, as defined in your `config.toml`.
+- `git-version`, e.g. `HEAD`, is the git commit hash of the OpenHands version you would
+like to evaluate. It could also be a release tag like `0.6.2`.
+- `agent`, e.g. `CodeActAgent`, is the name of the agent for benchmarks, defaulting to `CodeActAgent`.
+- `eval_limit`, e.g. `10`, limits the evaluation to the first `eval_limit` instances. By
+default, the script evaluates the (500 issues), which will no exceed the maximum of the dataset number.
+- `max_iter`, e.g. `20`, is the maximum number of iterations for the agent to run. By
+default, it is set to 50.
+- `num_workers`, e.g. `3`, is the number of parallel workers to run the evaluation. By
+default, it is set to 1.
+- `language`, the language of your evaluating dataset.
+- `dataset`, the absolute position of the dataset jsonl.
+
+The results will be generated in evaluation/evaluation_outputs/outputs/XXX/CodeActAgent/YYY/output.jsonl, you can refer to the [example](examples/output.jsonl).
+
+## Runing evaluation
+
+First, install [multi-swe-bench](https://github.com/multi-swe-bench/multi-swe-bench).
+
+```bash
+pip install multi-swe-bench
+```
+
+Second, convert the output.jsonl to patch.jsonl with [script](scripts/eval/convert.py), you can refer to the [example](examples/patch.jsonl).
+
+```bash
+python evaluation/benchmarks/multi_swe_bench/scripts/eval/convert.py
+```
+
+Finally, evaluate with multi-swe-bench.
+The config file config.json can be refer to the [example](examples/config.json) or [github](https://github.com/multi-swe-bench/multi-swe-bench/tree/main?tab=readme-ov-file#configuration-file-example).
+
+```bash
+python -m multi_swe_bench.harness.run_evaluation --config config.json
+```

evaluation/benchmarks/multi_swe_bench/eval_infer.py

@@ -0,0 +1,456 @@
+import json
+import os
+import subprocess
+import tempfile
+import time
+from functools import partial
+
+import pandas as pd
+from swebench.harness.grading import get_eval_report
+from swebench.harness.run_evaluation import (
+    APPLY_PATCH_FAIL,
+    APPLY_PATCH_PASS,
+)
+from swebench.harness.test_spec import SWEbenchInstance, TestSpec, make_test_spec
+from swebench.harness.utils import load_swebench_dataset
+from tqdm import tqdm
+
+from evaluation.benchmarks.swe_bench.resource.mapping import (
+    get_instance_resource_factor,
+)
+from evaluation.benchmarks.swe_bench.run_infer import get_instance_docker_image
+from evaluation.utils.shared import (
+    EvalMetadata,
+    EvalOutput,
+    get_default_sandbox_config_for_eval,
+    prepare_dataset,
+    reset_logger_for_multiprocessing,
+    run_evaluation,
+)
+from openhands.core.config import (
+    AppConfig,
+    LLMConfig,
+    get_parser,
+)
+from openhands.core.logger import openhands_logger as logger
+from openhands.core.main import create_runtime
+from openhands.events.action import CmdRunAction
+from openhands.events.observation import CmdOutputObservation
+from openhands.utils.async_utils import call_async_from_sync
+
+# TODO: migrate all swe-bench docker to ghcr.io/openhands
+DOCKER_IMAGE_PREFIX = os.environ.get('EVAL_DOCKER_IMAGE_PREFIX', 'docker.io/xingyaoww/')
+logger.info(f'Using docker image prefix: {DOCKER_IMAGE_PREFIX}')
+
+
+def process_git_patch(patch):
+    if not isinstance(patch, str):
+        return ''
+
+    if not patch.strip():
+        # skip empty patches
+        return ''
+
+    patch = patch.replace('\r\n', '\n')
+    # There might be some weird characters at the beginning of the patch
+    # due to some OpenHands inference command outputs
+
+    # FOR EXAMPLE:
+    # git diff --no-color --cached 895f28f9cbed817c00ab68770433170d83132d90
+    # 0
+    # diff --git a/django/db/models/sql/.backup.query.py b/django/db/models/sql/.backup.query.py
+    # new file mode 100644
+    # index 0000000000..fc13db5948
+
+    # We "find" the first line that starts with "diff" and then we remove lines before it
+    lines = patch.split('\n')
+    for i, line in enumerate(lines):
+        if line.startswith('diff --git'):
+            patch = '\n'.join(lines[i:])
+            break
+
+    patch = patch.rstrip() + '\n'  # Make sure the last line ends with a newline
+    return patch
+
+
+def get_config(metadata: EvalMetadata, instance: pd.Series) -> AppConfig:
+    # We use a different instance image for the each instance of swe-bench eval
+    base_container_image = get_instance_docker_image(instance['instance_id'])
+    logger.info(
+        f'Using instance container image: {base_container_image}. '
+        f'Please make sure this image exists. '
+        f'Submit an issue on https://github.com/All-Hands-AI/OpenHands if you run into any issues.'
+    )
+    sandbox_config = get_default_sandbox_config_for_eval()
+    sandbox_config.base_container_image = base_container_image
+    sandbox_config.remote_runtime_resource_factor = get_instance_resource_factor(
+        dataset_name=metadata.dataset,
+        instance_id=instance['instance_id'],
+    )
+    config = AppConfig(
+        run_as_openhands=False,
+        runtime=os.environ.get('RUNTIME', 'docker'),
+        sandbox=sandbox_config,
+        # do not mount workspace
+        workspace_base=None,
+        workspace_mount_path=None,
+    )
+    return config
+
+
+def process_instance(
+    instance: pd.Series,
+    metadata: EvalMetadata,
+    reset_logger: bool = True,
+    log_dir: str | None = None,
+    runtime_failure_count: int = 0,
+) -> EvalOutput:
+    """
+    Evaluate agent performance on a SWE-bench problem instance.
+
+    Note that this signature differs from the expected input to `run_evaluation`. Use
+    `functools.partial` to provide optional arguments before passing to the evaluation harness.
+
+    Args:
+        log_dir (str | None, default=None): Path to directory where log files will be written. Must
+        be provided if `reset_logger` is set.
+
+    Raises:
+        AssertionError: if the `reset_logger` flag is set without a provided log directory.
+    """
+    # Setup the logger properly, so you can run multi-processing to parallelize the evaluation
+    if reset_logger:
+        assert (
+            log_dir is not None
+        ), "Can't reset logger without a provided log directory."
+        os.makedirs(log_dir, exist_ok=True)
+        reset_logger_for_multiprocessing(logger, instance.instance_id, log_dir)
+    else:
+        logger.info(f'Starting evaluation for instance {instance.instance_id}.')
+
+    config = get_config(metadata, instance)
+    instance_id = instance.instance_id
+    model_patch = instance['model_patch']
+    test_spec: TestSpec = instance['test_spec']
+    logger.info(f'Starting evaluation for instance {instance_id}.')
+
+    if 'test_result' not in instance.keys():
+        instance['test_result'] = {}
+    instance['test_result']['report'] = {
+        'empty_generation': False,
+        'resolved': False,
+        'failed_apply_patch': False,
+        'error_eval': False,
+        'test_timeout': False,
+    }
+
+    if model_patch == '':
+        instance['test_result']['report']['empty_generation'] = True
+        return EvalOutput(
+            instance_id=instance_id,
+            test_result=instance['test_result'],
+            metadata=metadata,
+        )
+
+    # Increase resource_factor with increasing attempt_id
+    if runtime_failure_count > 0:
+        config.sandbox.remote_runtime_resource_factor = min(
+            config.sandbox.remote_runtime_resource_factor * (2**runtime_failure_count),
+            8,
+        )
+        logger.warning(
+            f'This is the {runtime_failure_count + 1}th attempt for instance {instance.instance_id}, setting resource factor to {config.sandbox.remote_runtime_resource_factor}'
+        )
+
+    try:
+        runtime = create_runtime(config)
+        call_async_from_sync(runtime.connect)
+        # Get patch and save it to /tmp/patch.diff
+        with tempfile.TemporaryDirectory() as temp_dir:
+            # Patch file
+            patch_file_path = os.path.join(temp_dir, 'patch.diff')
+            with open(patch_file_path, 'w') as f:
+                f.write(model_patch)
+            runtime.copy_to(patch_file_path, '/tmp')
+            # Eval script
+            eval_script_path = os.path.join(temp_dir, 'eval.sh')
+            with open(eval_script_path, 'w') as f:
+                f.write(test_spec.eval_script)
+            runtime.copy_to(eval_script_path, '/tmp')
+
+        # Set +x
+        action = CmdRunAction(command='chmod +x /tmp/eval.sh')
+        action.set_hard_timeout(600)
+        logger.info(action, extra={'msg_type': 'ACTION'})
+        obs = runtime.run_action(action)
+        logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+        assert obs.exit_code == 0
+
+        # Apply patch
+        exec_command = (
+            'cd /testbed && '
+            "(git apply -v /tmp/patch.diff && echo 'APPLY_PATCH_PASS' || "
+            "(echo 'Failed to apply patch with git apply, trying with patch command...' && "
+            "(patch --batch --fuzz=5 -p1 -i /tmp/patch.diff && echo 'APPLY_PATCH_PASS' || "
+            "echo 'APPLY_PATCH_FAIL')))"
+        )
+        action = CmdRunAction(command=exec_command)
+        action.set_hard_timeout(600)
+        obs = runtime.run_action(action)
+        assert isinstance(obs, CmdOutputObservation)
+        apply_patch_output = obs.content
+        assert isinstance(apply_patch_output, str)
+        instance['test_result']['apply_patch_output'] = apply_patch_output
+
+        if 'APPLY_PATCH_FAIL' in apply_patch_output:
+            logger.info(f'[{instance_id}] {APPLY_PATCH_FAIL}:\n{apply_patch_output}')
+            instance['test_result']['report']['failed_apply_patch'] = True
+
+            return EvalOutput(
+                instance_id=instance_id,
+                test_result=instance['test_result'],
+                metadata=metadata,
+            )
+        elif 'APPLY_PATCH_PASS' in apply_patch_output:
+            logger.info(f'[{instance_id}] {APPLY_PATCH_PASS}:\n{apply_patch_output}')
+
+            # Run eval script in background and save output to log file
+            log_file = '/tmp/eval_output.log'
+            action = CmdRunAction(command=f'/tmp/eval.sh > {log_file} 2>&1 & echo $!')
+            action.set_hard_timeout(300)  # Short timeout just to get the process ID
+            obs = runtime.run_action(action)
+
+            if isinstance(obs, CmdOutputObservation) and obs.exit_code == 0:
+                pid = obs.content.split()[-1].strip()
+                logger.info(
+                    f'[{instance_id}] Evaluation process started with PID: {pid}'
+                )
+
+                # Poll for completion
+                start_time = time.time()
+                timeout = 1800  # 30 minutes
+                while True:
+                    seconds_elapsed = time.time() - start_time
+                    if seconds_elapsed > timeout:
+                        logger.info(
+                            f'[{instance_id}] Evaluation timed out after {timeout} seconds'
+                        )
+                        instance['test_result']['report']['test_timeout'] = True
+                        break
+                    check_action = CmdRunAction(
+                        command=f'ps -p {pid} > /dev/null; echo $?'
+                    )
+                    check_action.set_hard_timeout(300)
+                    check_obs = runtime.run_action(check_action)
+                    if (
+                        isinstance(check_obs, CmdOutputObservation)
+                        and check_obs.content.split()[-1].strip() == '1'
+                    ):
+                        logger.info(
+                            f'[{instance_id}] Evaluation process completed after {seconds_elapsed} seconds'
+                        )
+                        break
+                    logger.info(
+                        f'[{instance_id}] [{seconds_elapsed:.0f}s] Evaluation still running, waiting...'
+                    )
+                    time.sleep(30)  # Wait for 30 seconds before checking again
+
+                # Read the log file
+                cat_action = CmdRunAction(command=f'cat {log_file}')
+                cat_action.set_hard_timeout(300)
+                cat_obs = runtime.run_action(cat_action)
+
+                # Grade answer
+                if isinstance(cat_obs, CmdOutputObservation) and cat_obs.exit_code == 0:
+                    test_output = cat_obs.content
+                    assert isinstance(test_output, str)
+                    instance['test_result']['test_output'] = test_output
+
+                    # Get report from test output
+                    logger.info(f'[{instance_id}] Grading answer...')
+                    with tempfile.TemporaryDirectory() as temp_dir:
+                        # Create a directory structure that matches the expected format
+                        # NOTE: this is a hack to make the eval report format consistent
+                        # with the original SWE-Bench eval script
+                        log_dir = os.path.join(temp_dir, 'logs', instance_id.lower())
+                        os.makedirs(log_dir, exist_ok=True)
+                        test_output_path = os.path.join(log_dir, 'test_output.txt')
+                        with open(test_output_path, 'w') as f:
+                            f.write(test_output)
+                        try:
+                            _report = get_eval_report(
+                                test_spec=test_spec,
+                                prediction={
+                                    'model_patch': model_patch,
+                                    'instance_id': instance_id,
+                                },
+                                log_path=test_output_path,
+                                include_tests_status=True,
+                            )
+                            report = _report[instance_id]
+                            logger.info(
+                                f"[{instance_id}] report: {report}\nResult for {instance_id}: resolved: {report['resolved']}"
+                            )
+                            instance['test_result']['report']['resolved'] = report[
+                                'resolved'
+                            ]
+                        except Exception as e:
+                            logger.error(
+                                f'[{instance_id}] Error when getting eval report: {e}'
+                            )
+                            instance['test_result']['report']['resolved'] = False
+                            instance['test_result']['report']['error_eval'] = True
+            else:
+                logger.info(f'[{instance_id}] Error when starting eval:\n{obs.content}')
+                instance['test_result']['report']['error_eval'] = True
+
+            return EvalOutput(
+                instance_id=instance_id,
+                test_result=instance['test_result'],
+                metadata=metadata,
+            )
+        else:
+            logger.info(
+                f'[{instance_id}] Unexpected output when applying patch:\n{apply_patch_output}'
+            )
+            raise RuntimeError(
+                instance_id,
+                f'Unexpected output when applying patch:\n{apply_patch_output}',
+                logger,
+            )
+    finally:
+        runtime.close()
+
+
+if __name__ == '__main__':
+    parser = get_parser()
+    parser.add_argument(
+        '--input-file',
+        type=str,
+        help='Path to input predictions file',
+        required=True,
+    )
+    parser.add_argument(
+        '--dataset',
+        type=str,
+        default='princeton-nlp/SWE-bench',
+        help='data set to evaluate on, either full-test or lite-test',
+    )
+    parser.add_argument(
+        '--split',
+        type=str,
+        default='test',
+        help='split to evaluate on',
+    )
+    args, _ = parser.parse_known_args()
+
+    # Load SWE-Bench dataset
+    full_dataset: list[SWEbenchInstance] = load_swebench_dataset(
+        args.dataset, args.split
+    )
+    instance_id_to_instance = {
+        instance['instance_id']: instance for instance in full_dataset
+    }
+    logger.info(
+        f'Loaded dataset {args.dataset} with split {args.split} to run inference on.'
+    )
+
+    # Load predictions
+    assert args.input_file.endswith('.jsonl'), 'Input file must be a jsonl file.'
+    required_fields = ['instance_id', 'model_patch', 'test_result']
+    with open(args.input_file) as f:
+        predictions = pd.DataFrame.from_records(
+            [
+                {k: v for k, v in json.loads(line).items() if k in required_fields}
+                for line in tqdm(f, desc='Loading predictions')
+            ]
+        )
+    assert (
+        'instance_id' in predictions.columns
+    ), 'Input file must contain instance_id column.'
+
+    if 'model_patch' not in predictions.columns and (
+        'test_result' in predictions.columns
+        and 'model_patch' in predictions['test_result'].iloc[0]
+    ):
+        raise ValueError(
+            'Input file must contain model_patch column OR test_result column with model_patch field.'
+        )
+    assert len(predictions['instance_id'].unique()) == len(
+        predictions
+    ), 'instance_id column must be unique.'
+
+    if 'model_patch' not in predictions.columns:
+        predictions['model_patch'] = predictions['test_result'].apply(
+            lambda x: x.get('git_patch', '')
+        )
+    assert {'instance_id', 'model_patch'}.issubset(
+        set(predictions.columns)
+    ), 'Input file must contain instance_id and model_patch columns.'
+
+    # Process model_patch
+    predictions['model_patch'] = predictions['model_patch'].apply(process_git_patch)
+
+    # Merge predictions with dataset
+    predictions['instance'] = predictions['instance_id'].apply(
+        lambda x: instance_id_to_instance[x]
+    )
+    predictions['test_spec'] = predictions['instance'].apply(make_test_spec)
+
+    # Prepare dataset
+    output_file = args.input_file.replace('.jsonl', '.swebench_eval.jsonl')
+    instances = prepare_dataset(predictions, output_file, args.eval_n_limit)
+
+    # If possible, load the relevant metadata to avoid issues with `run_evaluation`.
+    metadata: EvalMetadata | None = None
+    metadata_filepath = os.path.join(os.path.dirname(args.input_file), 'metadata.json')
+    if os.path.exists(metadata_filepath):
+        with open(metadata_filepath, 'r') as metadata_file:
+            data = metadata_file.read()
+            metadata = EvalMetadata.model_validate_json(data)
+    else:
+        # Initialize with a dummy metadata when file doesn't exist
+        metadata = EvalMetadata(
+            agent_class='dummy_agent',  # Placeholder agent class
+            llm_config=LLMConfig(model='dummy_model'),  # Minimal LLM config
+            max_iterations=1,  # Minimal iterations
+            eval_output_dir=os.path.dirname(
+                args.input_file
+            ),  # Use input file dir as output dir
+            start_time=time.strftime('%Y-%m-%d %H:%M:%S'),  # Current time
+            git_commit=subprocess.check_output(['git', 'rev-parse', 'HEAD'])
+            .decode('utf-8')
+            .strip(),  # Current commit
+            dataset=args.dataset,  # Dataset name from args
+        )
+
+    # The evaluation harness constrains the signature of `process_instance_func` but we need to
+    # pass extra information. Build a new function object to avoid issues with multiprocessing.
+    process_instance_func = partial(
+        process_instance, log_dir=output_file.replace('.jsonl', '.logs')
+    )
+
+    run_evaluation(
+        instances,
+        metadata=metadata,
+        output_file=output_file,
+        num_workers=args.eval_num_workers,
+        process_instance_func=process_instance_func,
+    )
+
+    # Load evaluated predictions & print number of resolved predictions
+    evaluated_predictions = pd.read_json(output_file, lines=True)
+    fields = ['resolved', 'failed_apply_patch', 'error_eval', 'empty_generation']
+
+    def count_report_field(row, field):
+        return row['test_result']['report'][field]
+
+    report = {}
+    for field in fields:
+        count = evaluated_predictions.apply(
+            count_report_field, args=(field,), axis=1
+        ).sum()
+        report[field] = count
+        logger.info(
+            f'# {field}: {count} / {len(evaluated_predictions)}. ({count / len(evaluated_predictions):.2%})'
+        )

evaluation/benchmarks/multi_swe_bench/examples/config.json

@@ -0,0 +1,24 @@
+{
+    "mode": "evaluation",
+    "workdir": "./data/workdir",
+    "patch_files": [
+        "./data/patches/<your_patch_file>.jsonl"
+    ],
+    "dataset_files": [
+        "./data/patches/<to_evaluate_dataset_file>.jsonl"
+    ],
+    "force_build": false,
+    "output_dir": "./data/dataset",
+    "specifics": [],
+    "skips": [],
+    "repo_dir": "./data/repos",
+    "need_clone": false,
+    "global_env": [],
+    "clear_env": true,
+    "stop_on_error": true,
+    "max_workers": 8,
+    "max_workers_build_image": 8,
+    "max_workers_run_instance": 8,
+    "log_dir": "./data/logs",
+    "log_level": "DEBUG"
+}

evaluation/benchmarks/multi_swe_bench/examples/patch.jsonl

@@ -0,0 +1,3 @@
+{"org": "ponylang", "repo": "ponyc", "number": "4595", "fix_patch": "diff --git a/src/libponyc/ast/parser.c b/src/libponyc/ast/parser.c\nindex 9852922f..2c37d6b8 100644\n--- a/src/libponyc/ast/parser.c\n+++ b/src/libponyc/ast/parser.c\n@@ -693,6 +693,7 @@ DEF(idseqsingle);\n   AST_NODE(TK_LET);\n   TOKEN(\"variable name\", TK_ID);\n   AST_NODE(TK_NONE);  // Type\n+  SET_FLAG(AST_FLAG_IN_PARENS);\n   DONE();\n \n // idseq"}
+{"org": "ponylang", "repo": "ponyc", "number": "4593", "fix_patch": "diff --git a/packages/cli/command_parser.pony b/packages/cli/command_parser.pony\nindex a5acce8e..fa97808b 100644\n--- a/packages/cli/command_parser.pony\n+++ b/packages/cli/command_parser.pony\n@@ -100,6 +100,7 @@ class CommandParser\n             | let cs: CommandSpec box =>\n               return CommandParser._sub(cs, this).\n                 _parse_command(tokens, options, args, envsmap, opt_stop)\n+// Correctly handle parent default options\n             end\n           else\n             return SyntaxError(token, \"unknown command\")"}
+{"org": "ponylang", "repo": "ponyc", "number": "4588", "fix_patch": "diff --git a/src/libponyc/expr/match.c b/src/libponyc/expr/match.c\nindex 7d16066f..c2ec7056 100644\n--- a/src/libponyc/expr/match.c\n+++ b/src/libponyc/expr/match.c\n@@ -314,8 +314,10 @@ static ast_t* make_pattern_type(pass_opt_t* opt, ast_t* pattern)\n     case TK_DONTCAREREF:\n     case TK_MATCH_CAPTURE:\n     case TK_MATCH_DONTCARE:\n+      if (ast_id(pattern_type) == TK_ISO) pattern_type = set_cap_and_ephemeral(pattern_type, TK_TRN, TK_EPHEMERAL);\n       return pattern_type;\n \n+\n     case TK_TUPLE:\n     {\n       ast_t* pattern_child = ast_child(pattern);"}

evaluation/benchmarks/multi_swe_bench/infer.sh

@@ -0,0 +1,32 @@
+#!/bin/bash
+
+
+BASE_SCRIPT="./evaluation/benchmarks/multi_swe_bench/scripts/run_infer.sh"
+
+MODELS=("aaa" "bbb" "ccc" "ddd" "fff")
+GIT_VERSION="HEAD"
+AGENT_NAME="CodeActAgent"
+EVAL_LIMIT="500"
+MAX_ITER="50"
+NUM_WORKERS="1"
+LANGUAGE="XXX"
+DATASET="XXX"
+
+
+for MODEL in "${MODELS[@]}"; do
+    echo "=============================="
+    echo "Running benchmark for MODEL: $MODEL"
+    echo "=============================="
+
+    $BASE_SCRIPT \
+        "$MODEL" \
+        "$GIT_VERSION" \
+        "$AGENT_NAME" \
+        "$EVAL_LIMIT" \
+        "$MAX_ITER" \
+        "$NUM_WORKERS" \
+        "$DATASET" \
+        "$LANGUAGE"
+
+    echo "Completed $MODEL"
+done

evaluation/benchmarks/multi_swe_bench/resource/mapping.py

@@ -0,0 +1,39 @@
+"""Mapping instance_id to resource_factor.
+
+Different instances may have different resource requirements.
+e.g., some instances may require more memory/CPU to run inference.
+This file tracks the resource requirements of different instances.
+"""
+
+import json
+import os
+
+from openhands.core.logger import openhands_logger as logger
+
+CUR_DIR = os.path.dirname(os.path.abspath(__file__))
+DEFAULT_RUNTIME_RESOURCE_FACTOR = int(
+    os.environ.get('DEFAULT_RUNTIME_RESOURCE_FACTOR', 1)
+)
+
+# dataset to resource mapping
+_global_resource_mapping: dict[str, dict[str, float]] = {}
+
+
+def get_resource_mapping(dataset_name: str) -> dict[str, float]:
+    if dataset_name not in _global_resource_mapping:
+        file_path = os.path.join(CUR_DIR, f'{dataset_name}.json')
+        if not os.path.exists(file_path):
+            logger.warning(f'Resource mapping for {dataset_name} not found.')
+            return None
+
+        with open(file_path, 'r') as f:
+            _global_resource_mapping[dataset_name] = json.load(f)
+        logger.info(f'Loaded resource mapping for {dataset_name}')
+    return _global_resource_mapping[dataset_name]
+
+
+def get_instance_resource_factor(dataset_name: str, instance_id: str) -> int:
+    resource_mapping = get_resource_mapping(dataset_name)
+    if resource_mapping is None:
+        return DEFAULT_RUNTIME_RESOURCE_FACTOR
+    return int(resource_mapping.get(instance_id, DEFAULT_RUNTIME_RESOURCE_FACTOR))

evaluation/benchmarks/multi_swe_bench/run_infer.py

@@ -0,0 +1,847 @@
+import asyncio
+import json
+import os
+import tempfile
+from typing import Any
+
+import pandas as pd
+import toml
+from datasets import load_dataset
+
+import openhands.agenthub
+from evaluation.benchmarks.swe_bench.resource.mapping import (
+    get_instance_resource_factor,
+)
+from evaluation.utils.shared import (
+    EvalException,
+    EvalMetadata,
+    EvalOutput,
+    assert_and_raise,
+    codeact_user_response,
+    get_default_sandbox_config_for_eval,
+    get_metrics,
+    is_fatal_evaluation_error,
+    make_metadata,
+    prepare_dataset,
+    reset_logger_for_multiprocessing,
+    run_evaluation,
+    update_llm_config_for_completions_logging,
+)
+from openhands.controller.state.state import State
+from openhands.core.config import (
+    AgentConfig,
+    AppConfig,
+    get_llm_config_arg,
+    get_parser,
+)
+from openhands.core.logger import openhands_logger as logger
+from openhands.core.main import create_runtime, run_controller
+from openhands.events.action import CmdRunAction, FileReadAction, MessageAction
+from openhands.events.observation import CmdOutputObservation, ErrorObservation
+from openhands.events.serialization.event import event_to_dict
+from openhands.runtime.base import Runtime
+from openhands.utils.async_utils import call_async_from_sync
+from openhands.utils.shutdown_listener import sleep_if_should_continue
+
+USE_HINT_TEXT = os.environ.get('USE_HINT_TEXT', 'false').lower() == 'true'
+USE_INSTANCE_IMAGE = os.environ.get('USE_INSTANCE_IMAGE', 'true').lower() == 'true'
+RUN_WITH_BROWSING = os.environ.get('RUN_WITH_BROWSING', 'false').lower() == 'true'
+
+# TODO: migrate all swe-bench docker to ghcr.io/openhands
+# TODO: 适应所有的语言
+DOCKER_IMAGE_PREFIX = os.environ.get('EVAL_DOCKER_IMAGE_PREFIX', '')
+LANGUAGE = os.environ.get('LANGUAGE', 'python')
+logger.info(f'Using docker image prefix: {DOCKER_IMAGE_PREFIX}')
+
+
+AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {
+    'CodeActAgent': codeact_user_response,
+}
+
+
+def _get_swebench_workspace_dir_name(instance: pd.Series) -> str:
+    return f'{instance.repo}__{instance.version}'.replace('/', '__')
+
+
+def get_instruction(instance: pd.Series, metadata: EvalMetadata):
+    workspace_dir_name = _get_swebench_workspace_dir_name(instance)
+    # Prepare instruction
+
+    # Instruction based on Anthropic's official trajectory
+    # https://github.com/eschluntz/swe-bench-experiments/tree/main/evaluation/verified/20241022_tools_claude-3-5-sonnet-updated/trajs
+    instructions = {
+        'python': (
+            '<uploaded_files>\n'
+            f'/workspace/{workspace_dir_name}\n'
+            '</uploaded_files>\n'
+            f"I've uploaded a python code repository in the directory {workspace_dir_name}. Consider the following issue description:\n\n"
+            f'<issue_description>\n'
+            f'{instance.problem_statement}\n'
+            '</issue_description>\n\n'
+            'Can you help me implement the necessary changes to the repository so that the requirements specified in the <issue_description> are met?\n'
+            "I've already taken care of all changes to any of the test files described in the <issue_description>. This means you DON'T have to modify the testing logic or any of the tests in any way!\n"
+            "Also the development Python environment is already set up for you (i.e., all dependencies already installed), so you don't need to install other packages.\n"
+            'Your task is to make the minimal changes to non-test files in the /workspace directory to ensure the <issue_description> is satisfied.\n'
+            'Follow these steps to resolve the issue:\n'
+            '1. As a first step, it might be a good idea to explore the repo to familiarize yourself with its structure.\n'
+            '2. Create a script to reproduce the error and execute it with `python <filename.py>` using the BashTool, to confirm the error.\n'
+            '3. Edit the sourcecode of the repo to resolve the issue.\n'
+            '4. Rerun your reproduce script and confirm that the error is fixed!\n'
+            '5. Think about edgecases, add comprehensive tests for them in your reproduce script, and run them to make sure your fix handles them as well.\n'
+            f'6. Once you are done with the initial implementation, please carefully re-read the problem description and check the difference between the current code and the base commit {instance["base_commit"]}. Do you think that the issue has been completely and comprehensively solved? Write tests to check the correctness of the solution, specifically focusing on tests that may point out any remaining problems that are not yet solved. Run all of the tests in the repo and check if any of them fail, and if they do fix the code. Repeat this process of carefully reading the problem description and current implementation, testing, and fixing any problems until you are confident that the current implementation is correct. Find and run any tests in the repo that are related to:\n'
+            '   - The issue you are fixing\n'
+            '   - The files you modified\n'
+            '   - The functions you changed\n'
+            '   Make sure all these tests pass with your changes.\n'
+            "Your thinking should be thorough and so it's fine if it's very long.\n"
+        ),
+        'java': (
+            '<uploaded_files>\n'
+            f'/workspace/{workspace_dir_name}\n'
+            '</uploaded_files>\n'
+            f"I've uploaded a Java code repository in the directory {workspace_dir_name}. Consider the following issue description:\n\n"
+            f'<issue_description>\n'
+            f'{instance.problem_statement}\n'
+            '</issue_description>\n\n'
+            "Can you help me implement the necessary changes to the repository so that the requirements specified in the <issue_description> are met?\n"
+            "I've already taken care of all changes to any of the test files described in the <issue_description>. This means you DON'T have to modify the testing logic or any of the tests in any way!\n"
+            "Also the development Java environment is already set up for you (i.e., all dependencies already installed), so you don't need to install other packages.\n"
+            "Your task is to make the minimal changes to non-test files in the /workspace directory to ensure the <issue_description> is satisfied.\n"
+            "Follow these steps to resolve the issue:\n"
+            "1. As a first step, it might be a good idea to explore the repo to familiarize yourself with its structure.\n"
+            '2. Create a Java class to reproduce the error and execute it by first compiling with `javac <classname>.java` and then running with `java <classname>` using the BashTool, to confirm the error\n'
+            "3. Edit the sourcecode of the repo to resolve the issue.\n"
+            "4. Rerun your reproduce script or class and confirm that the error is fixed!\n"
+            "5. Think about edgecases, add comprehensive tests for them in your reproduce class or script, and run them to make sure your fix handles these cases as well.\n"
+            f"6. Once you are done with the initial implementation, please carefully re-read the problem description and check the difference between the current code and the base commit {instance['base_commit']}. Do you think that the issue has been completely and comprehensively solved? Write tests to check the correctness of the solution, specifically focusing on tests that may point out any remaining problems that are not yet solved. Run all of the tests in the repo and check if any of them fail, and if they do fix the code. Repeat this process of carefully reading the problem description and current implementation, testing, and fixing any problems until you are confident that the current implementation is correct. Find and run any tests in the repo that are related to:\n"
+            "   - The issue you are fixing\n"
+            "   - The files you modified\n"
+            "   - The functions or classes you changed\n"
+            "   Make sure all these tests pass with your changes.\n"
+            "Your thinking should be thorough and so it's fine if it's very long.\n"
+        ),
+        'go': (
+            '<uploaded_files>\n'
+            f'/workspace/{workspace_dir_name}\n'
+            '</uploaded_files>\n'
+            f"I've uploaded a Go code repository in the directory {workspace_dir_name}. Consider the following issue description:\n\n"
+            f'<issue_description>\n'
+            f'{instance.problem_statement}\n'
+            '</issue_description>\n\n'
+            'Can you help me implement the necessary changes to the repository so that the requirements specified in the <issue_description> are met?\n'
+            "I've already taken care of all changes to any of the test files described in the <issue_description>. This means you DON'T have to modify the testing logic or any of the tests in any way!\n"
+            "Also the development Go environment is already set up for you (i.e., all dependencies already installed), so you don't need to install other packages.\n"
+            'Your task is to make the minimal changes to non-test files in the /workspace directory to ensure the <issue_description> is satisfied.\n'
+            'Follow these steps to resolve the issue:\n'
+            '1. As a first step, it might be a good idea to explore the repo to familiarize yourself with its structure.\n'
+            '2. Create a script or a function to reproduce the error and execute it with `go run <filename.go>` using the BashTool, to confirm the error.\n'
+            '3. Edit the sourcecode of the repo to resolve the issue.\n'
+            '4. Rerun your reproduce script and confirm that the error is fixed!\n'
+            '5. Think about edgecases, add comprehensive tests for them in your reproduce script, and run them to make sure your fix handles them as well.\n'
+            f'6. Once you are done with the initial implementation, please carefully re-read the problem description and check the difference between the current code and the base commit {instance["base_commit"]}. Do you think that the issue has been completely and comprehensively solved? Write tests to check the correctness of the solution, specifically focusing on tests that may point out any remaining problems that are not yet solved. Run all of the tests in the repo and check if any of them fail, and if they do fix the code. Repeat this process of carefully reading the problem description and current implementation, testing, and fixing any problems until you are confident that the current implementation is correct. Find and run any tests in the repo that are related to:\n'
+            '   - The issue you are fixing\n'
+            '   - The files you modified\n'
+            '   - The functions you changed\n'
+            '   Make sure all these tests pass with your changes.\n'
+            "Your thinking should be thorough and so it's fine if it's very long.\n"
+        ),
+        'c': (
+            '<uploaded_files>\n'
+            f'/workspace/{workspace_dir_name}\n'
+            '</uploaded_files>\n'
+            f"I've uploaded a C code repository in the directory {workspace_dir_name}. Consider the following issue description:\n\n"
+            f'<issue_description>\n'
+            f'{instance.problem_statement}\n'
+            '</issue_description>\n\n'
+            'Can you help me implement the necessary changes to the repository so that the requirements specified in the <issue_description> are met?\n'
+            "I've already taken care of all changes to any of the test files described in the <issue_description>. This means you DON'T have to modify the testing logic or any of the tests in any way!\n"
+            "Also the development C environment is already set up for you (i.e., all dependencies already installed), so you don't need to install other packages.\n"
+            'Your task is to make the minimal changes to non-test files in the /workspace directory to ensure the <issue_description> is satisfied.\n'
+            'Follow these steps to resolve the issue:\n'
+            '1. As a first step, it might be a good idea to explore the repo to familiarize yourself with its structure.\n'
+            '2. Create a script to reproduce the error by compiling your C code (for example, using `gcc <filename.c> -o <executable>`) and then running the executable using the BashTool, to confirm the error.\n'
+            '3. Edit the sourcecode of the repo to resolve the issue.\n'
+            '4. Rerun your reproduce script and confirm that the error is fixed!\n'
+            '5. Think about edgecases, add comprehensive tests for them in your reproduce script, and run them to make sure your fix handles them as well.\n'
+            f'6. Once you are done with the initial implementation, please carefully re-read the problem description and check the difference between the current code and the base commit {instance["base_commit"]}. Do you think that the issue has been completely and comprehensively solved? Write tests to check the correctness of the solution, specifically focusing on tests that may point out any remaining problems that are not yet solved. Run all of the tests in the repo and check if any of them fail, and if they do fix the code. Repeat this process of carefully reading the problem description and current implementation, testing, and fixing any problems until you are confident that the current implementation is correct. Find and run any tests in the repo that are related to:\n'
+            '   - The issue you are fixing\n'
+            '   - The files you modified\n'
+            '   - The functions you changed\n'
+            '   Make sure all these tests pass with your changes.\n'
+            "Your thinking should be thorough and so it's fine if it's very long.\n"
+        ),
+        'cpp': (
+            '<uploaded_files>\n'
+            f'/workspace/{workspace_dir_name}\n'
+            '</uploaded_files>\n'
+            f"I've uploaded a C++ code repository in the directory {workspace_dir_name}. Consider the following issue description:\n\n"
+            f'<issue_description>\n'
+            f'{instance.problem_statement}\n'
+            '</issue_description>\n\n'
+            'Can you help me implement the necessary changes to the repository so that the requirements specified in the <issue_description> are met?\n'
+            "I've already taken care of all changes to any of the test files described in the <issue_description>. This means you DON'T have to modify the testing logic or any of the tests in any way!\n"
+            "Also the development C++ environment is already set up for you (i.e., all dependencies already installed), so you don't need to install other packages.\n"
+            'Your task is to make the minimal changes to non-test files in the /workspace directory to ensure the <issue_description> is satisfied.\n'
+            'Follow these steps to resolve the issue:\n'
+            '1. As a first step, it might be a good idea to explore the repo to familiarize yourself with its structure.\n'
+            '2. Create or adapt a small executable (e.g., a main file or a test driver) to reproduce the issue. Build and run it (for example, by using `g++ -o reproduce reproduce.cpp && ./reproduce` via the BashTool) to confirm the error.\n'
+            '3. Edit the sourcecode of the repo to resolve the issue.\n'
+            '4. Rerun your reproduce script and confirm that the error is fixed!\n'
+            '5. Think about edgecases, add comprehensive tests for them in your reproduce script, and run them to make sure your fix handles them as well.\n'
+            f'6. Once you are done with the initial implementation, please carefully re-read the problem description and check the difference between the current code and the base commit {instance["base_commit"]}. Do you think that the issue has been completely and comprehensively solved? Write tests to check the correctness of the solution, specifically focusing on tests that may point out any remaining problems that are not yet solved. Run all of the tests in the repo and check if any of them fail, and if they do fix the code. Repeat this process of carefully reading the problem description and current implementation, testing, and fixing any problems until you are confident that the current implementation is correct. Find and run any tests in the repo that are related to:\n'
+            '   - The issue you are fixing\n'
+            '   - The files you modified\n'
+            '   - The functions you changed\n'
+            '   Make sure all these tests pass with your changes.\n'
+            "Your thinking should be thorough and so it's fine if it's very long.\n"
+        ),
+        'javascript': (
+            '<uploaded_files>\n'
+            f'/workspace/{workspace_dir_name}\n'
+            '</uploaded_files>\n'
+            f"I've uploaded a Javascript code repository in the directory {workspace_dir_name}. Consider the following issue description:\n\n"
+            f'<issue_description>\n'
+            f'{instance.problem_statement}\n'
+            '</issue_description>\n\n'
+            'Can you help me implement the necessary changes to the repository so that the requirements specified in the <issue_description> are met?\n'
+            "I've already taken care of all changes to any of the test files described in the <issue_description>. This means you DON'T have to modify the testing logic or any of the tests in any way!\n"
+            "Also the development Javascript environment is already set up for you (i.e., all dependencies already installed), so you don't need to install other packages.\n"
+            'Your task is to make the minimal changes to non-test files in the /workspace directory to ensure the <issue_description> is satisfied.\n'
+            'Follow these steps to resolve the issue:\n'
+            '1. As a first step, it might be a good idea to explore the repo to familiarize yourself with its structure.\n'
+            '2. Create a script to reproduce the error and execute it with `node <filename.js>` using the BashTool, to confirm the error.\n'
+            '3. Edit the sourcecode of the repo to resolve the issue.\n'
+            '4. Rerun your reproduce script and confirm that the error is fixed!\n'
+            '5. Think about edgecases, add comprehensive tests for them in your reproduce script, and run them to make sure your fix handles them as well.\n'
+            f'6. Once you are done with the initial implementation, please carefully re-read the problem description and check the difference between the current code and the base commit {instance["base_commit"]}. Do you think that the issue has been completely and comprehensively solved? Write tests to check the correctness of the solution, specifically focusing on tests that may point out any remaining problems that are not yet solved. Run all of the tests in the repo and check if any of them fail, and if they do fix the code. Repeat this process of carefully reading the problem description and current implementation, testing, and fixing any problems until you are confident that the current implementation is correct. Find and run any tests in the repo that are related to:\n'
+            '   - The issue you are fixing\n'
+            '   - The files you modified\n'
+            '   - The functions you changed\n'
+            '   Make sure all these tests pass with your changes.\n'
+            "Your thinking should be thorough and so it's fine if it's very long.\n"
+        ),
+        'typescript': (
+            '<uploaded_files>\n'
+            f'/workspace/{workspace_dir_name}\n'
+            '</uploaded_files>\n'
+            f"I've uploaded a Typescript code repository in the directory {workspace_dir_name}. Consider the following issue description:\n\n"
+            f'<issue_description>\n'
+            f'{instance.problem_statement}\n'
+            '</issue_description>\n\n'
+            'Can you help me implement the necessary changes to the repository so that the requirements specified in the <issue_description> are met?\n'
+            "I've already taken care of all changes to any of the test files described in the <issue_description>. This means you DON'T have to modify the testing logic or any of the tests in any way!\n"
+            "Also the development Typescript environment is already set up for you (i.e., all dependencies already installed), so you don't need to install other packages.\n"
+            'Your task is to make the minimal changes to non-test files in the /workspace directory to ensure the <issue_description> is satisfied.\n'
+            'Follow these steps to resolve the issue:\n'
+            '1. As a first step, it might be a good idea to explore the repo to familiarize yourself with its structure.\n'
+            '2. Create a script to reproduce the error and execute it with `ts-node <filename.ts>` using the BashTool, to confirm the error.\n'
+            '3. Edit the sourcecode of the repo to resolve the issue.\n'
+            '4. Rerun your reproduce script and confirm that the error is fixed!\n'
+            '5. Think about edgecases, add comprehensive tests for them in your reproduce script, and run them to make sure your fix handles them as well.\n'
+            f'6. Once you are done with the initial implementation, please carefully re-read the problem description and check the difference between the current code and the base commit {instance["base_commit"]}. Do you think that the issue has been completely and comprehensively solved? Write tests to check the correctness of the solution, specifically focusing on tests that may point out any remaining problems that are not yet solved. Run all of the tests in the repo and check if any of them fail, and if they do fix the code. Repeat this process of carefully reading the problem description and current implementation, testing, and fixing any problems until you are confident that the current implementation is correct. Find and run any tests in the repo that are related to:\n'
+            '   - The issue you are fixing\n'
+            '   - The files you modified\n'
+            '   - The functions you changed\n'
+            '   Make sure all these tests pass with your changes.\n'
+            "Your thinking should be thorough and so it's fine if it's very long.\n"
+        ),
+        'rust': (
+            '<uploaded_files>\n'
+            f'/workspace/{workspace_dir_name}\n'
+            '</uploaded_files>\n'
+            f"I've uploaded a Rust code repository in the directory {workspace_dir_name}. Consider the following issue description:\n\n"
+            f'<issue_description>\n'
+            f'{instance.problem_statement}\n'
+            '</issue_description>\n\n'
+            'Can you help me implement the necessary changes to the repository so that the requirements specified in the <issue_description> are met?\n'
+            "I've already taken care of all changes to any of the test files described in the <issue_description>. This means you DON'T have to modify the testing logic or any of the tests in any way!\n"
+            "Also the development Rust environment is already set up for you (i.e., all dependencies already installed), so you don't need to install other packages.\n"
+            'Your task is to make the minimal changes to non-test files in the /workspace directory to ensure the <issue_description> is satisfied.\n'
+            'Follow these steps to resolve the issue:\n'
+            '1. As a first step, it might be a good idea to explore the repo to familiarize yourself with its structure.\n'
+            '2. Create a reproduction script (or binary) that triggers the error and execute it with `cargo run --bin <filename>` using the BashTool, to confirm the error.\n'
+            '3. Edit the sourcecode of the repo to resolve the issue.\n'
+            '4. Rerun your reproduce script and confirm that the error is fixed!\n'
+            '5. Think about edgecases, add comprehensive tests for them in your reproduce script, and run them to make sure your fix handles them as well.\n'
+            f'6. Once you are done with the initial implementation, please carefully re-read the problem description and check the difference between the current code and the base commit {instance["base_commit"]}. Do you think that the issue has been completely and comprehensively solved? Write tests to check the correctness of the solution, specifically focusing on tests that may point out any remaining problems that are not yet solved. Run all of the tests in the repo and check if any of them fail, and if they do fix the code. Repeat this process of carefully reading the problem description and current implementation, testing, and fixing any problems until you are confident that the current implementation is correct. Find and run any tests in the repo that are related to:\n'
+            '   - The issue you are fixing\n'
+            '   - The files you modified\n'
+            '   - The functions you changed\n'
+            '   Make sure all these tests pass with your changes.\n'
+            "Your thinking should be thorough and so it's fine if it's very long.\n"
+        ),
+    }
+    instruction = instructions.get(LANGUAGE.lower())
+
+    if instruction and RUN_WITH_BROWSING:
+        instruction += (
+            '<IMPORTANT!>\n'
+            'You SHOULD NEVER attempt to browse the web. '
+            '</IMPORTANT!>\n'
+        )
+    return instruction
+
+
+# TODO: 适应所有的语言
+# def get_instance_docker_image(instance_id: str) -> str:
+#     image_name = 'sweb.eval.x86_64.' + instance_id
+#     if LANGUAGE == 'python':
+#         image_name = image_name.replace(
+#             '__', '_s_'
+#         )  # to comply with docker image naming convention
+#         return (DOCKER_IMAGE_PREFIX.rstrip('/') + '/' + image_name).lower()
+#     else:
+#         return image_name.lower() ##加载本地的
+def get_instance_docker_image(instance: pd.Series):
+    if LANGUAGE == 'python':
+        image_name = 'sweb.eval.x86_64.' + instance['instance_id']
+        image_name = image_name.replace(
+            '__', '_s_'
+        )  # to comply with docker image naming convention
+        return (DOCKER_IMAGE_PREFIX.rstrip('/') + '/' + image_name).lower()
+    else:
+        container_name = instance.get('repo', '').lower()
+        container_name = container_name.replace('/', '_m_')
+        instance_id = instance.get('instance_id', '')
+        tag_suffix = instance_id.split('-')[-1] if instance_id else ''
+        container_tag = f'pr-{tag_suffix}'
+        # pdb.set_trace()
+        return f'mswebench/{container_name}:{container_tag}'
+        # return "kong/insomnia:pr-8284"
+        # return "'sweb.eval.x86_64.local_insomnia"
+        # return "local_insomnia_why"
+        # return "local/kong-insomnia:pr-8117"
+
+
+def get_config(
+    instance: pd.Series,
+    metadata: EvalMetadata,
+) -> AppConfig:
+    SWE_BENCH_CONTAINER_IMAGE = 'ghcr.io/opendevin/eval-swe-bench:full-v1.2.1'
+    if USE_INSTANCE_IMAGE:
+        # We use a different instance image for the each instance of swe-bench eval
+        # base_container_image = get_instance_docker_image(instance['instance_id'])
+        base_container_image = get_instance_docker_image(instance)
+        logger.info(
+            f'Using instance container image: {base_container_image}. '
+            f'Please make sure this image exists. '
+            f'Submit an issue on https://github.com/All-Hands-AI/OpenHands if you run into any issues.'
+        )
+    else:
+        base_container_image = SWE_BENCH_CONTAINER_IMAGE
+        logger.info(f'Using swe-bench container image: {base_container_image}')
+
+    sandbox_config = get_default_sandbox_config_for_eval()
+    sandbox_config.base_container_image = base_container_image
+    sandbox_config.enable_auto_lint = True
+    sandbox_config.use_host_network = False
+    # Add platform to the sandbox config to solve issue 4401
+    sandbox_config.platform = 'linux/amd64'
+    sandbox_config.remote_runtime_resource_factor = get_instance_resource_factor(
+        dataset_name=metadata.dataset,
+        instance_id=instance['instance_id'],
+    )
+
+    config = AppConfig(
+        default_agent=metadata.agent_class,
+        run_as_openhands=False,
+        max_iterations=metadata.max_iterations,
+        runtime=os.environ.get('RUNTIME', 'docker'),
+        sandbox=sandbox_config,
+        # do not mount workspace
+        workspace_base=None,
+        workspace_mount_path=None,
+    )
+    config.set_llm_config(
+        update_llm_config_for_completions_logging(
+            metadata.llm_config, metadata.eval_output_dir, instance['instance_id']
+        )
+    )
+    agent_config = AgentConfig(
+        enable_jupyter=False,
+        enable_browsing=RUN_WITH_BROWSING,
+        enable_llm_editor=False,
+        condenser=metadata.condenser_config,
+        enable_prompt_extensions=False,
+    )
+    config.set_agent_config(agent_config)
+    return config
+
+
+def initialize_runtime(
+    runtime: Runtime,
+    instance: pd.Series,  # this argument is not required
+):
+    """Initialize the runtime for the agent.
+
+    This function is called before the runtime is used to run the agent.
+    """
+    logger.info('-' * 30)
+    logger.info('BEGIN Runtime Initialization Fn')
+    logger.info('-' * 30)
+    workspace_dir_name = _get_swebench_workspace_dir_name(instance)
+    obs: CmdOutputObservation
+
+    REPO_NAME = instance['repo'].split('/')[-1]
+    # Set instance id
+    action = CmdRunAction(
+        command=f"""echo 'export SWE_INSTANCE_ID={instance['instance_id']}' >> ~/.bashrc && echo 'export PIP_CACHE_DIR=~/.cache/pip' >> ~/.bashrc && echo "alias git='git --no-pager'" >> ~/.bashrc && echo 'export REPO_NAME={REPO_NAME}' >> ~/.bashrc"""
+    )
+    action.set_hard_timeout(600)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+    assert_and_raise(
+        obs.exit_code == 0, f'Failed to export SWE_INSTANCE_ID: {str(obs)}'
+    )
+    # pdb.set_trace()
+    action = CmdRunAction(command="""export USER=$(whoami); echo USER=${USER} """)
+    action.set_hard_timeout(600)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+    assert_and_raise(obs.exit_code == 0, f'Failed to export USER: {str(obs)}')
+
+    if USE_INSTANCE_IMAGE:
+        # inject the init script
+        script_dir = os.path.dirname(__file__)
+
+        # inject the instance info
+        action = CmdRunAction(command='mkdir -p /swe_util/eval_data/instances')
+        action.set_hard_timeout(600)
+        logger.info(action, extra={'msg_type': 'ACTION'})
+        obs = runtime.run_action(action)
+        logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+        assert_and_raise(
+            obs.exit_code == 0,
+            f'Failed to create /swe_util/eval_data/instances: {str(obs)}',
+        )
+
+        swe_instance_json_name = 'swe-bench-instance.json'
+        with tempfile.TemporaryDirectory() as temp_dir:
+            # Construct the full path for the desired file name within the temporary directory
+            temp_file_path = os.path.join(temp_dir, swe_instance_json_name)
+            # Write to the file with the desired name within the temporary directory
+            with open(temp_file_path, 'w') as f:
+                if not isinstance(instance, dict):
+                    json.dump([instance.to_dict()], f)
+                else:
+                    json.dump([instance], f)
+
+            # Copy the file to the desired location
+            runtime.copy_to(temp_file_path, '/swe_util/eval_data/instances/')
+
+        # inject the instance swe entry
+        runtime.copy_to(
+            str(os.path.join(script_dir, 'scripts/setup/instance_swe_entry.sh')),
+            '/swe_util/',
+        )
+        action = CmdRunAction(command='cat ~/.bashrc')
+        action.set_hard_timeout(600)
+        logger.info(action, extra={'msg_type': 'ACTION'})
+        obs = runtime.run_action(action)
+        logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+        assert_and_raise(obs.exit_code == 0, f'Failed to cat ~/.bashrc: {str(obs)}')
+
+        action = CmdRunAction(command='source ~/.bashrc')
+        action.set_hard_timeout(600)
+        logger.info(action, extra={'msg_type': 'ACTION'})
+        obs = runtime.run_action(action)
+        logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+        if isinstance(obs, ErrorObservation):
+            logger.error(f'Failed to source ~/.bashrc: {str(obs)}')
+        assert_and_raise(obs.exit_code == 0, f'Failed to source ~/.bashrc: {str(obs)}')
+
+        action = CmdRunAction(command='source /swe_util/instance_swe_entry.sh')
+        action.set_hard_timeout(600)
+        logger.info(action, extra={'msg_type': 'ACTION'})
+        obs = runtime.run_action(action)
+        logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+        assert_and_raise(
+            obs.exit_code == 0,
+            f'Failed to source /swe_util/instance_swe_entry.sh: {str(obs)}',
+        )
+    else:
+        action = CmdRunAction(command='source /swe_util/swe_entry.sh')
+        action.set_hard_timeout(1800)
+        logger.info(action, extra={'msg_type': 'ACTION'})
+        obs = runtime.run_action(action)
+        logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+        assert_and_raise(
+            obs.exit_code == 0,
+            f'Failed to source /swe_util/swe_entry.sh: {str(obs)}',
+        )
+
+    action = CmdRunAction(command=f'cd /workspace/{workspace_dir_name}')
+    action.set_hard_timeout(600)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+    assert_and_raise(
+        obs.exit_code == 0,
+        f'Failed to cd to /workspace/{workspace_dir_name}: {str(obs)}',
+    )
+
+    action = CmdRunAction(command='git reset --hard')
+    action.set_hard_timeout(600)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+    assert_and_raise(obs.exit_code == 0, f'Failed to git reset --hard: {str(obs)}')
+
+    action = CmdRunAction(
+        command='for remote_name in $(git remote); do git remote remove "${remote_name}"; done'
+    )
+    action.set_hard_timeout(600)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+    assert_and_raise(obs.exit_code == 0, f'Failed to remove git remotes: {str(obs)}')
+    ##TODO:这里看看需不需要判断其他语言的环境
+    # action = CmdRunAction(command='which python')
+    # action.set_hard_timeout(600)
+    # logger.info(action, extra={'msg_type': 'ACTION'})
+    # obs = runtime.run_action(action)
+    # logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+    # assert_and_raise(
+    #     obs.exit_code == 0 and 'testbed' in obs.content,
+    #     f'Expected to find python interpreter from testbed, but got: {str(obs)}',
+    # )
+
+    logger.info('-' * 30)
+    logger.info('END Runtime Initialization Fn')
+    logger.info('-' * 30)
+
+
+def complete_runtime(
+    runtime: Runtime,
+    instance: pd.Series,  # this argument is not required, but it is used to get the workspace_dir_name
+) -> dict[str, Any]:
+    """Complete the runtime for the agent.
+
+    This function is called before the runtime is used to run the agent.
+    If you need to do something in the sandbox to get the correctness metric after
+    the agent has run, modify this function.
+    """
+    logger.info('-' * 30)
+    logger.info('BEGIN Runtime Completion Fn')
+    logger.info('-' * 30)
+    obs: CmdOutputObservation
+    workspace_dir_name = _get_swebench_workspace_dir_name(instance)
+
+    action = CmdRunAction(command=f'cd /workspace/{workspace_dir_name}')
+    action.set_hard_timeout(600)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+
+    if obs.exit_code == -1:
+        # The previous command is still running
+        # We need to kill previous command
+        logger.info('The previous command is still running, trying to kill it...')
+        action = CmdRunAction(command='C-c')
+        obs = runtime.run_action(action)
+        logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+
+        # Then run the command again
+        action = CmdRunAction(command=f'cd /workspace/{workspace_dir_name}')
+        action.set_hard_timeout(600)
+        logger.info(action, extra={'msg_type': 'ACTION'})
+        obs = runtime.run_action(action)
+        logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+
+    assert_and_raise(
+        isinstance(obs, CmdOutputObservation) and obs.exit_code == 0,
+        f'Failed to cd to /workspace/{workspace_dir_name}: {str(obs)}',
+    )
+
+    action = CmdRunAction(command='git config --global core.pager ""')
+    action.set_hard_timeout(600)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+    assert_and_raise(
+        isinstance(obs, CmdOutputObservation) and obs.exit_code == 0,
+        f'Failed to git config --global core.pager "": {str(obs)}',
+    )
+
+    action = CmdRunAction(command='git add -A')
+    action.set_hard_timeout(600)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+    assert_and_raise(
+        isinstance(obs, CmdOutputObservation) and obs.exit_code == 0,
+        f'Failed to git add -A: {str(obs)}',
+    )
+
+    ##删除二进制文件
+    action = CmdRunAction(
+        command="""
+        for file in $(git status --porcelain | grep -E "^(M| M|\\?\\?|A| A)" | cut -c4-); do
+            if [ -f "$file" ] && (file "$file" | grep -q "executable" || git check-attr binary "$file" | grep -q "binary: set"); then
+                git rm -f "$file" 2>/dev/null || rm -f "$file"
+                echo "Removed: $file"
+            fi
+        done
+        """
+    )
+    action.set_hard_timeout(600)
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+    assert_and_raise(
+        isinstance(obs, CmdOutputObservation) and obs.exit_code == 0,
+        f'Failed to remove binary files: {str(obs)}',
+    )
+
+    # pdb.set_trace()
+
+    n_retries = 0
+    git_patch = None
+    while n_retries < 5:
+        action = CmdRunAction(
+            command=f'git diff --no-color --cached {instance["base_commit"]} > patch.diff'
+        )
+        action.set_hard_timeout(max(300 + 100 * n_retries, 600))
+        logger.info(action, extra={'msg_type': 'ACTION'})
+        obs = runtime.run_action(action)
+        logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+        n_retries += 1
+        if isinstance(obs, CmdOutputObservation):
+            if obs.exit_code == 0:
+                # git_patch = obs.content.strip()
+                break
+            else:
+                logger.info('Failed to get git diff, retrying...')
+                sleep_if_should_continue(10)
+        elif isinstance(obs, ErrorObservation):
+            logger.error(f'Error occurred: {obs.content}. Retrying...')
+            sleep_if_should_continue(10)
+        else:
+            assert_and_raise(False, f'Unexpected observation type: {str(obs)}')
+
+    action = FileReadAction(path='patch.diff')
+    action.set_hard_timeout(max(300 + 100 * n_retries, 600))
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    git_patch = obs.content
+    # pdb.set_trace()
+
+    assert_and_raise(git_patch is not None, 'Failed to get git diff (None)')
+
+    logger.info('-' * 30)
+    logger.info('END Runtime Completion Fn')
+    logger.info('-' * 30)
+    return {'git_patch': git_patch}
+
+
+def process_instance(
+    instance: pd.Series,
+    metadata: EvalMetadata,
+    reset_logger: bool = True,
+    runtime_failure_count: int = 0,
+) -> EvalOutput:
+    config = get_config(instance, metadata)
+
+    # Setup the logger properly, so you can run multi-processing to parallelize the evaluation
+    if reset_logger:
+        log_dir = os.path.join(metadata.eval_output_dir, 'infer_logs')
+        reset_logger_for_multiprocessing(logger, instance.instance_id, log_dir)
+    else:
+        logger.info(f'Starting evaluation for instance {instance.instance_id}.')
+
+    # Increase resource_factor with increasing attempt_id
+    if runtime_failure_count > 0:
+        config.sandbox.remote_runtime_resource_factor = min(
+            config.sandbox.remote_runtime_resource_factor * (2**runtime_failure_count),
+            8,
+        )
+        logger.warning(
+            f'This is the {runtime_failure_count + 1}th attempt for instance {instance.instance_id}, setting resource factor to {config.sandbox.remote_runtime_resource_factor}'
+        )
+    # pdb.set_trace()
+    runtime = create_runtime(config)
+    call_async_from_sync(runtime.connect)
+
+    try:
+        initialize_runtime(runtime, instance)
+
+        instruction = get_instruction(instance, metadata)
+
+        # Here's how you can run the agent (similar to the `main` function) and get the final task state
+        state: State | None = asyncio.run(
+            run_controller(
+                config=config,
+                initial_user_action=MessageAction(content=instruction),
+                runtime=runtime,
+                fake_user_response_fn=AGENT_CLS_TO_FAKE_USER_RESPONSE_FN[
+                    metadata.agent_class
+                ],
+            )
+        )
+
+        # if fatal error, throw EvalError to trigger re-run
+        if is_fatal_evaluation_error(state.last_error):
+            raise EvalException('Fatal error detected: ' + state.last_error)
+
+        # ======= THIS IS SWE-Bench specific =======
+        # Get git patch
+        return_val = complete_runtime(runtime, instance)
+        git_patch = return_val['git_patch']
+        logger.info(
+            f'Got git diff for instance {instance.instance_id}:\n--------\n{git_patch}\n--------'
+        )
+    finally:
+        runtime.close()
+    # ==========================================
+
+    # ======= Attempt to evaluate the agent's edits =======
+    # we use eval_infer.sh to evaluate the agent's edits, not here
+    # because the agent may alter the environment / testcases
+    ###remove binary diffs
+    def remove_binary_diffs(patch_text):
+        lines = patch_text.splitlines()
+        cleaned_lines = []
+        block = []
+        is_binary_block = False
+
+        for line in lines:
+            if line.startswith('diff --git '):
+                if block and not is_binary_block:
+                    cleaned_lines.extend(block)
+                block = [line]
+                is_binary_block = False
+            elif 'Binary files' in line:
+                is_binary_block = True
+                block.append(line)
+            else:
+                block.append(line)
+
+        if block and not is_binary_block:
+            cleaned_lines.extend(block)
+        return '\n'.join(cleaned_lines)
+
+    git_patch = remove_binary_diffs(git_patch)
+    test_result = {
+        'git_patch': git_patch,
+    }
+
+    # If you are working on some simpler benchmark that only evaluates the final model output (e.g., in a MessageAction)
+    # You can simply get the LAST `MessageAction` from the returned `state.history` and parse it for evaluation.
+    if state is None:
+        raise ValueError('State should not be None.')
+
+    # NOTE: this is NO LONGER the event stream, but an agent history that includes delegate agent's events
+    histories = [event_to_dict(event) for event in state.history]
+    metrics = get_metrics(state)
+
+    # Save the output
+    output = EvalOutput(
+        instance_id=instance.instance_id,
+        instruction=instruction,
+        instance=instance.to_dict(),  # SWE Bench specific
+        test_result=test_result,
+        metadata=metadata,
+        history=histories,
+        metrics=metrics,
+        error=state.last_error if state and state.last_error else None,
+    )
+    return output
+
+
+def filter_dataset(dataset: pd.DataFrame, filter_column: str) -> pd.DataFrame:
+    file_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'config.toml')
+    if os.path.exists(file_path):
+        with open(file_path, 'r') as file:
+            data = toml.load(file)
+            if 'selected_ids' in data:
+                selected_ids = data['selected_ids']
+                logger.info(
+                    f'Filtering {len(selected_ids)} tasks from "selected_ids"...'
+                )
+                subset = dataset[dataset[filter_column].isin(selected_ids)]
+                logger.info(f'Retained {subset.shape[0]} tasks after filtering')
+                return subset
+    skip_ids = os.environ.get('SKIP_IDS', '').split(',')
+    if len(skip_ids) > 0:
+        logger.info(f'Filtering {len(skip_ids)} tasks from "SKIP_IDS"...')
+        return dataset[~dataset[filter_column].isin(skip_ids)]
+    return dataset
+
+
+if __name__ == '__main__':
+    # pdb.set_trace()
+    parser = get_parser()
+    parser.add_argument(
+        '--dataset',
+        type=str,
+        default='princeton-nlp/SWE-bench',
+        help='data set to evaluate on, either full-test or lite-test',
+    )
+    parser.add_argument(
+        '--split',
+        type=str,
+        default='test',
+        help='split to evaluate on',
+    )
+    args, _ = parser.parse_known_args()
+
+    # NOTE: It is preferable to load datasets from huggingface datasets and perform post-processing
+    # so we don't need to manage file uploading to OpenHands's repo
+    # dataset = load_dataset(args.dataset, split=args.split)
+    # dataset = load_dataset(args.dataset)
+    dataset = load_dataset('json', data_files=args.dataset)
+    dataset = dataset[args.split]
+    swe_bench_tests = filter_dataset(dataset.to_pandas(), 'instance_id')
+    logger.info(
+        f'Loaded dataset {args.dataset} with split {args.split}: {len(swe_bench_tests)} tasks'
+    )
+
+    llm_config = None
+    if args.llm_config:
+        llm_config = get_llm_config_arg(args.llm_config)
+        llm_config.log_completions = True
+        # modify_params must be False for evaluation purpose, for reproducibility and accurancy of results
+        llm_config.modify_params = False
+
+    if llm_config is None:
+        raise ValueError(f'Could not find LLM config: --llm_config {args.llm_config}')
+
+    details = {}
+    _agent_cls = openhands.agenthub.Agent.get_cls(args.agent_cls)
+
+    dataset_descrption = (
+        args.dataset.replace('/', '__') + '-' + args.split.replace('/', '__')
+    )
+    metadata = make_metadata(
+        llm_config,
+        dataset_descrption,
+        args.agent_cls,
+        args.max_iterations,
+        args.eval_note,
+        args.eval_output_dir,
+        details=details,
+    )
+
+    output_file = os.path.join(metadata.eval_output_dir, 'output.jsonl')
+    print(f'### OUTPUT FILE: {output_file} ###')
+    instances = prepare_dataset(swe_bench_tests, output_file, args.eval_n_limit)
+
+    if len(instances) > 0 and not isinstance(
+        instances['FAIL_TO_PASS'][instances['FAIL_TO_PASS'].index[0]], str
+    ):
+        for col in ['PASS_TO_PASS', 'FAIL_TO_PASS']:
+            instances[col] = instances[col].apply(lambda x: str(x))
+    # if LANGUAGE == "java": ##TODO:适配多语言的版本
+    #     for col in ['issue_numbers', 'created_at']:
+    #         instances[col] = instances[col].apply(lambda x: str(x))
+    run_evaluation(
+        instances,
+        metadata,
+        output_file,
+        args.eval_num_workers,
+        process_instance,
+        timeout_seconds=120 * 60,  # 2 hour PER instance should be more than enough
+        max_retries=5,
+    )

evaluation/benchmarks/multi_swe_bench/scripts/data/data_change.py

@@ -0,0 +1,36 @@
+import json
+
+input_file = 'XXX.jsonl'
+output_file = 'YYY.jsonl'
+
+with open(input_file, 'r', encoding='utf-8') as fin, open(
+    output_file, 'w', encoding='utf-8'
+) as fout:
+    for line in fin:
+        line = line.strip()
+        if not line:
+            continue
+
+        data = json.loads(line)
+        item = data
+
+        # 提取原始数据
+        org = item.get('org', '')
+        repo = item.get('repo', '')
+        number = str(item.get('number', ''))
+
+        new_item = {}
+        new_item['repo'] = f'{org}/{repo}'
+        new_item['instance_id'] = f'{org}__{repo}-{number}'
+        new_item['problem_statement'] = (
+            item['resolved_issues'][0].get('title', '')
+            + '\n'
+            + item['resolved_issues'][0].get('body', '')
+        )
+        new_item['FAIL_TO_PASS'] = []
+        new_item['PASS_TO_PASS'] = []
+        new_item['base_commit'] = item['base'].get('sha', '')
+        new_item['version'] = '0.1'  # depends
+
+        output_data = new_item
+        fout.write(json.dumps(output_data, ensure_ascii=False) + '\n')

evaluation/benchmarks/multi_swe_bench/scripts/eval/convert.py

@@ -0,0 +1,24 @@
+import json
+import re
+
+IN_FILE = 'output.jsonl'
+OUT_FILE = 'patch.jsonl'
+
+
+def main():
+    with open(IN_FILE, 'r') as fin:
+        with open(OUT_FILE, 'w') as fout:
+            for line in fin:
+                data = json.loads(line)
+                groups = re.match(r'(.*)__(.*)-(.*)', data['instance_id'])
+                patch = {
+                    'org': groups.group(1),
+                    'repo': groups.group(2),
+                    'number': groups.group(3),
+                    'fix_patch': data['test_result']['git_patch'],
+                }
+                fout.write(json.dumps(patch) + '\n')
+
+
+if __name__ == '__main__':
+    main()

evaluation/benchmarks/multi_swe_bench/scripts/run_infer.sh

@@ -0,0 +1,155 @@
+#!/bin/bash
+set -eo pipefail
+
+source "evaluation/utils/version_control.sh"
+
+MODEL_CONFIG=$1
+COMMIT_HASH=$2
+AGENT=$3
+EVAL_LIMIT=$4
+MAX_ITER=$5
+NUM_WORKERS=$6
+DATASET=$7
+# SPLIT=$8
+LANGUAGE=$8
+# N_RUNS=$10
+
+if [ -z "$NUM_WORKERS" ]; then
+  NUM_WORKERS=1
+  echo "Number of workers not specified, use default $NUM_WORKERS"
+fi
+checkout_eval_branch
+
+if [ -z "$AGENT" ]; then
+  echo "Agent not specified, use default CodeActAgent"
+  AGENT="CodeActAgent"
+fi
+
+if [ -z "$MAX_ITER" ]; then
+  echo "MAX_ITER not specified, use default 100"
+  MAX_ITER=100
+fi
+
+if [ -z "$USE_INSTANCE_IMAGE" ]; then
+  echo "USE_INSTANCE_IMAGE not specified, use default true"
+  USE_INSTANCE_IMAGE=true
+fi
+
+if [ -z "$RUN_WITH_BROWSING" ]; then
+  echo "RUN_WITH_BROWSING not specified, use default false"
+  RUN_WITH_BROWSING=false
+fi
+
+
+if [ -z "$DATASET" ]; then
+  echo "DATASET not specified, use default princeton-nlp/SWE-bench_Lite"
+  DATASET="princeton-nlp/SWE-bench_Lite"
+fi
+
+if [ -z "$LANGUAGE" ]; then
+  echo "LANUGUAGE not specified, use default python"
+  LANGUAGE="python"
+fi
+
+if [ -z "$SPLIT" ]; then
+  echo "LANUGUAGE not specified, use default python"
+  SPLIT="train"
+fi
+
+##TODO:适配多语言的版本
+# if [ -z "$SPLIT" ]; then
+#   if [ "$LANGUAGE" = "python" ]; then
+#   echo "SPLIT is test as LANUGUAGE is python"
+#     SPLIT="test"
+#   elif [ "$LANGUAGE" = "java" ]; then
+#   echo "SPLIT is java_verified as LANUGUAGE is java"
+#     SPLIT="java_verified"
+#   fi
+# fi
+
+if [ -z "$EVAL_DOCKER_IMAGE_PREFIX" ]; then
+  if [ "$LANGUAGE" = "python" ]; then
+  echo "EVAL_DOCKER_IMAGE_PREFIX is docker.io/xingyaoww/ as default as LANUGUAGE is python"
+    EVAL_DOCKER_IMAGE_PREFIX="docker.io/xingyaoww/"
+  elif [ "$LANGUAGE" = "java" ]; then
+  echo "EVAL_DOCKER_IMAGE_PREFIX is java_verified as LANUGUAGE is java"
+    EVAL_DOCKER_IMAGE_PREFIX=""
+  fi
+fi
+
+export EVAL_DOCKER_IMAGE_PREFIX=$EVAL_DOCKER_IMAGE_PREFIX
+echo "EVAL_DOCKER_IMAGE_PREFIX: $EVAL_DOCKER_IMAGE_PREFIX"
+export USE_INSTANCE_IMAGE=$USE_INSTANCE_IMAGE
+echo "USE_INSTANCE_IMAGE: $USE_INSTANCE_IMAGE"
+export RUN_WITH_BROWSING=$RUN_WITH_BROWSING
+echo "RUN_WITH_BROWSING: $RUN_WITH_BROWSING"
+export LANGUAGE=$LANGUAGE
+echo "LANGUAGE: $LANGUAGE"
+
+get_openhands_version
+
+echo "AGENT: $AGENT"
+echo "OPENHANDS_VERSION: $OPENHANDS_VERSION"
+echo "MODEL_CONFIG: $MODEL_CONFIG"
+echo "DATASET: $DATASET"
+echo "SPLIT: $SPLIT"
+
+# Default to NOT use Hint
+if [ -z "$USE_HINT_TEXT" ]; then
+  export USE_HINT_TEXT=false
+fi
+echo "USE_HINT_TEXT: $USE_HINT_TEXT"
+EVAL_NOTE="$OPENHANDS_VERSION"
+# if not using Hint, add -no-hint to the eval note
+if [ "$USE_HINT_TEXT" = false ]; then
+  EVAL_NOTE="$EVAL_NOTE-no-hint"
+fi
+
+if [ "$RUN_WITH_BROWSING" = true ]; then
+  EVAL_NOTE="$EVAL_NOTE-with-browsing"
+fi
+
+if [ -n "$EXP_NAME" ]; then
+  EVAL_NOTE="$EVAL_NOTE-$EXP_NAME"
+fi
+
+function run_eval() {
+  local eval_note=$1
+  COMMAND="poetry run python evaluation/benchmarks/multi_swe_bench/run_infer.py \
+    --agent-cls $AGENT \
+    --llm-config $MODEL_CONFIG \
+    --max-iterations $MAX_ITER \
+    --eval-num-workers $NUM_WORKERS \
+    --eval-note $eval_note \
+    --dataset $DATASET \
+    --split $SPLIT"
+
+  if [ -n "$EVAL_LIMIT" ]; then
+    echo "EVAL_LIMIT: $EVAL_LIMIT"
+    COMMAND="$COMMAND --eval-n-limit $EVAL_LIMIT"
+  fi
+
+  # Run the command
+  eval $COMMAND
+}
+
+unset SANDBOX_ENV_GITHUB_TOKEN # prevent the agent from using the github token to push
+if [ -z "$N_RUNS" ]; then
+  N_RUNS=1
+  echo "N_RUNS not specified, use default $N_RUNS"
+fi
+
+# Skip runs if the run number is in the SKIP_RUNS list
+# read from env variable SKIP_RUNS as a comma separated list of run numbers
+SKIP_RUNS=(${SKIP_RUNS//,/ })
+for i in $(seq 1 $N_RUNS); do
+  if [[ " ${SKIP_RUNS[@]} " =~ " $i " ]]; then
+    echo "Skipping run $i"
+    continue
+  fi
+  current_eval_note="$EVAL_NOTE-run_$i"
+  echo "EVAL_NOTE: $current_eval_note"
+  run_eval $current_eval_note
+done
+
+checkout_original_branch

evaluation/benchmarks/multi_swe_bench/scripts/setup/compare_patch_filename.py

@@ -0,0 +1,54 @@
+"""This script compares gold patches with OpenHands-generated patches and check whether
+OpenHands found the right (set of) files to modify.
+"""
+
+import argparse
+import json
+import re
+
+
+def extract_modified_files(patch):
+    modified_files = set()
+    file_pattern = re.compile(r'^diff --git a/(.*?) b/')
+
+    for line in patch.split('\n'):
+        match = file_pattern.match(line)
+        if match:
+            modified_files.add(match.group(1))
+
+    return modified_files
+
+
+def process_report(oh_output_file):
+    succ = 0
+    fail = 0
+    for line in open(oh_output_file):
+        line = json.loads(line)
+        instance_id = line['instance_id']
+        gold_patch = line['swe_instance']['patch']
+        generated_patch = line['git_patch']
+        gold_modified_files = extract_modified_files(gold_patch)
+        # swe-bench lite only: a gold patch always contains exactly one file
+        assert len(gold_modified_files) == 1
+        generated_modified_files = extract_modified_files(generated_patch)
+
+        # Check if all files in gold_patch are also in generated_patch
+        all_files_in_generated = gold_modified_files.issubset(generated_modified_files)
+        if all_files_in_generated:
+            succ += 1
+        else:
+            fail += 1
+            print(
+                f'{instance_id}: file mismatch, gold = {gold_modified_files}, generated = {generated_modified_files}'
+            )
+    print(
+        f'\nSUMMARY: {succ} out of {succ + fail} instances found correct files to edit, success rate = {succ / float(succ + fail)}'
+    )
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--oh_output_file', help='Path to the OH output file')
+    args = parser.parse_args()
+
+    process_report(args.oh_output_file)

evaluation/benchmarks/multi_swe_bench/scripts/setup/instance_swe_entry.sh

@@ -0,0 +1,45 @@
+#!/bin/bash
+
+source ~/.bashrc
+SWEUTIL_DIR=/swe_util
+
+# FIXME: Cannot read SWE_INSTANCE_ID from the environment variable
+# SWE_INSTANCE_ID=django__django-11099
+if [ -z "$SWE_INSTANCE_ID" ]; then
+    echo "Error: SWE_INSTANCE_ID is not set." >&2
+    exit 1
+fi
+
+if [ -z "$REPO_NAME" ]; then
+    echo "Error: REPO_NAME is not set." >&2
+    exit 1
+fi
+
+# Read the swe-bench-test-lite.json file and extract the required item based on instance_id
+item=$(jq --arg INSTANCE_ID "$SWE_INSTANCE_ID" '.[] | select(.instance_id == $INSTANCE_ID)' $SWEUTIL_DIR/eval_data/instances/swe-bench-instance.json)
+
+if [[ -z "$item" ]]; then
+  echo "No item found for the provided instance ID."
+  exit 1
+fi
+
+WORKSPACE_NAME=$(echo "$item" | jq -r '(.repo | tostring) + "__" + (.version | tostring) | gsub("/"; "__")')
+
+echo "WORKSPACE_NAME: $WORKSPACE_NAME"
+
+# Clear the workspace
+if [ -d /workspace ]; then
+    rm -rf /workspace/*
+else
+    mkdir /workspace
+fi
+# Copy repo to workspace
+if [ -d /workspace/$WORKSPACE_NAME ]; then
+    rm -rf /workspace/$WORKSPACE_NAME
+fi
+mkdir -p /workspace
+cp -r /home/$REPO_NAME /workspace/$WORKSPACE_NAME
+
+# Activate instance-specific environment
+# . /opt/miniconda3/etc/profile.d/conda.sh
+# conda activate testbed

evaluation/benchmarks/multi_swe_bench/scripts/setup/prepare_swe_utils.sh

@@ -0,0 +1,27 @@
+#!/bin/bash
+
+set -e
+EVAL_WORKSPACE="evaluation/benchmarks/swe_bench/eval_workspace"
+mkdir -p $EVAL_WORKSPACE
+
+# 1. Prepare REPO
+echo "==== Prepare SWE-bench repo ===="
+OH_SWE_BENCH_REPO_PATH="https://github.com/All-Hands-AI/SWE-bench.git"
+OH_SWE_BENCH_REPO_BRANCH="eval"
+git clone -b $OH_SWE_BENCH_REPO_BRANCH $OH_SWE_BENCH_REPO_PATH $EVAL_WORKSPACE/OH-SWE-bench
+
+# 2. Prepare DATA
+echo "==== Prepare SWE-bench data ===="
+EVAL_IMAGE=ghcr.io/all-hands-ai/eval-swe-bench:builder_with_conda
+EVAL_WORKSPACE=$(realpath $EVAL_WORKSPACE)
+chmod +x $EVAL_WORKSPACE/OH-SWE-bench/swebench/harness/prepare_data.sh
+if [ -d $EVAL_WORKSPACE/eval_data ]; then
+    rm -r $EVAL_WORKSPACE/eval_data
+fi
+docker run \
+    -v $EVAL_WORKSPACE:/workspace \
+    -w /workspace \
+    -u $(id -u):$(id -g) \
+    -e HF_DATASETS_CACHE="/tmp" \
+    --rm -it $EVAL_IMAGE \
+    bash -c "cd OH-SWE-bench/swebench/harness && /swe_util/miniforge3/bin/conda run -n swe-bench-eval ./prepare_data.sh && mv eval_data /workspace/"

evaluation/benchmarks/multi_swe_bench/scripts/setup/swe_entry.sh

@@ -0,0 +1,96 @@
+#!/bin/bash
+
+set -e
+
+# assert user name is `root`
+if [ "$USER" != "root" ]; then
+    echo "Error: This script is intended to be run by the 'root' user only." >&2
+    exit 1
+fi
+
+source ~/.bashrc
+
+SWEUTIL_DIR=/swe_util
+
+# Create logs directory
+LOG_DIR=/openhands/logs
+mkdir -p $LOG_DIR && chmod 777 $LOG_DIR
+
+# FIXME: Cannot read SWE_INSTANCE_ID from the environment variable
+# SWE_INSTANCE_ID=django__django-11099
+if [ -z "$SWE_INSTANCE_ID" ]; then
+    echo "Error: SWE_INSTANCE_ID is not set." >&2
+    exit 1
+fi
+
+# Read the swe-bench-test-lite.json file and extract the required item based on instance_id
+item=$(jq --arg INSTANCE_ID "$SWE_INSTANCE_ID" '.[] | select(.instance_id == $INSTANCE_ID)' $SWEUTIL_DIR/eval_data/instances/swe-bench-test-lite.json)
+
+if [[ -z "$item" ]]; then
+  echo "No item found for the provided instance ID."
+  exit 1
+fi
+
+CONDA_ENV_NAME=$(echo "$item" | jq -r '.repo + "__" + .version | gsub("/"; "__")')
+
+echo "CONDA_ENV_NAME: $CONDA_ENV_NAME"
+
+SWE_TASK_DIR=/openhands/swe_tasks
+mkdir -p $SWE_TASK_DIR
+# Dump test_patch to /workspace/test.patch
+echo "$item" | jq -r '.test_patch' > $SWE_TASK_DIR/test.patch
+# Dump patch to /workspace/gold.patch
+echo "$item" | jq -r '.patch' > $SWE_TASK_DIR/gold.patch
+# Dump the item to /workspace/instance.json except for the "test_patch" and "patch" fields
+echo "$item" | jq 'del(.test_patch, .patch)' > $SWE_TASK_DIR/instance.json
+
+# Clear the workspace
+rm -rf /workspace/*
+# Copy repo to workspace
+if [ -d /workspace/$CONDA_ENV_NAME ]; then
+    rm -rf /workspace/$CONDA_ENV_NAME
+fi
+cp -r $SWEUTIL_DIR/eval_data/testbeds/$CONDA_ENV_NAME /workspace
+
+# Reset swe-bench testbed and install the repo
+. $SWEUTIL_DIR/miniforge3/etc/profile.d/conda.sh
+conda config --set changeps1 False
+conda config --append channels conda-forge
+conda activate swe-bench-eval
+
+mkdir -p $SWE_TASK_DIR/reset_testbed_temp
+mkdir -p $SWE_TASK_DIR/reset_testbed_log_dir
+SWE_BENCH_DIR=/swe_util/OH-SWE-bench
+output=$(
+    export PYTHONPATH=$SWE_BENCH_DIR && \
+    cd $SWE_BENCH_DIR && \
+    python swebench/harness/reset_swe_env.py \
+    --swe_bench_tasks $SWEUTIL_DIR/eval_data/instances/swe-bench-test.json \
+    --temp_dir $SWE_TASK_DIR/reset_testbed_temp \
+    --testbed /workspace \
+    --conda_path $SWEUTIL_DIR/miniforge3 \
+    --instance_id $SWE_INSTANCE_ID \
+    --log_dir $SWE_TASK_DIR/reset_testbed_log_dir \
+    --timeout 900 \
+    --verbose
+)
+
+REPO_PATH=$(echo "$output" | awk -F': ' '/repo_path:/ {print $2}')
+TEST_CMD=$(echo "$output" | awk -F': ' '/test_cmd:/ {print $2}')
+echo "Repo Path: $REPO_PATH"
+echo "Test Command: $TEST_CMD"
+
+echo "export SWE_BENCH_DIR=\"$SWE_BENCH_DIR\"" >> ~/.bashrc
+echo "export REPO_PATH=\"$REPO_PATH\"" >> ~/.bashrc
+echo "export TEST_CMD=\"$TEST_CMD\"" >> ~/.bashrc
+
+if [[ "$REPO_PATH" == "None" ]]; then
+    echo "Error: Failed to retrieve repository path. Tests may not have passed or output was not as expected." >&2
+    exit 1
+fi
+
+# Activate instance-specific environment
+. $SWEUTIL_DIR/miniforge3/etc/profile.d/conda.sh
+conda activate $CONDA_ENV_NAME
+
+set +e

evaluation/benchmarks/swe_bench/README.md

@@ -2,6 +2,8 @@
 
 This folder contains the evaluation harness that we built on top of the original [SWE-Bench benchmark](https://www.swebench.com/) ([paper](https://arxiv.org/abs/2310.06770)).
 
+**UPDATE (4/8/2025): We now support running SWT-Bench evaluation! For more details, checkout [the corresponding section](#SWT-Bench-Evaluation).**
+
 **UPDATE (03/27/2025): We now support SWE-Bench multimodal evaluation! Simply use "princeton-nlp/SWE-bench_Multimodal" as the dataset name in the `run_infer.sh` script to evaluate on multimodal instances.**
 
 **UPDATE (2/18/2025): We now support running SWE-Gym using the same evaluation harness here. For more details, checkout [this README](./SWE-Gym.md).**
@@ -141,7 +143,7 @@ With `output.jsonl` file, you can run `eval_infer.sh` to evaluate generated patc
 ./evaluation/benchmarks/swe_bench/scripts/eval_infer.sh $YOUR_OUTPUT_JSONL [instance_id] [dataset_name] [split]
 
 # Example
-./evaluation/benchmarks/swe_bench/scripts/eval_infer.sh evaluation/evaluation_outputs/outputs/swe_bench/CodeActAgent/gpt-4-1106-preview_maxiter_50_N_v1.0/output.jsonl
+./evaluation/benchmarks/swe_bench/scripts/eval_infer.sh evaluation/evaluation_outputs/outputs/princeton-nlp__SWE-bench_Lite/CodeActAgent/gpt-4-1106-preview_maxiter_50_N_v1.0/output.jsonl
 ```
 
 The script now accepts optional arguments:
@@ -182,3 +184,58 @@ To clean-up all existing runtimes that you've already started, run:
 ```bash
 ALLHANDS_API_KEY="YOUR-API-KEY" ./evaluation/utils/scripts/cleanup_remote_runtime.sh
 ```
+
+## SWT-Bench Evaluation
+
+[SWT-Bench](https://swtbench.com/) ([paper](https://arxiv.org/abs/2406.12952)) is a benchmark for evaluating the capability of LLMs at creating unit tests. It is performed on the same instances as SWE-Bench, but requires a separate evaluation harness to capture coverage and issue reproduction. We therefore detail below how to leverage the inference script in this folder to run inference on SWT-Bench and how to use the SWT-Bench evaluation harness to evaluate them.
+
+### Run inference on SWT-Bench
+
+To run inference on SWT-Bench, you can use the same `run_infer.sh` script as described for evaluation on plain SWE-Bench. The only differences is that you need to specify the `mode` parameter to `swt` or `swt-ci` when running the script. For example, to run inference on SWT-Bench Verified, run the following command:
+
+```bash
+./evaluation/benchmarks/swe_bench/scripts/run_infer.sh [model_config] [git-version] [agent] [eval_limit] [max_iter] [num_workers] [swe-dataset] test 1 swt
+
+# Example - This runs evaluation on CodeActAgent for 500 instances on "SWT-bench_Verified"'s test set (corresponding to SWE-bench_Verified), with max 100 iteration per instances, with 1 number of workers running in parallel
+./evaluation/benchmarks/swe_bench/scripts/run_infer.sh llm.eval_gpt4o-2024-11-20 HEAD CodeActAgent 500 100 1 princeton-nlp/SWE-bench_Verified test 1 swt
+```
+
+The two modes `swt` and `swt-ci` have the following effect:
+- `swt`: This mode will change the prompt to instruct the agent to generate reproducing test cases instead of resolving the issue.
+- `swt-ci`: In addition to the changes by `swt`, this mode sets up the CI environment by i) pre-installing the environment in the docker image, such that the test framework can be executed without errors and ii) telling the model the exact command to run the test framework.
+
+### Run evaluation for SWT-bench
+
+The evaluation of these results is done leveraging [the SWT-Bench evaluation harness](https://github.com/logic-star-ai/swt-bench/tree/master).
+
+#### Extracting results into SWT-Bench harness format
+In order to run evaluation of the obtained inference results in the SWT-Bench harness, we transform the results to a format that the SWT-Bench evaluation harness expects.
+
+```bash
+python3 evaluation/benchmarks/swe_bench/scripts/swtbench/convert.py --prediction_file [output.jsonl] > [output_swt.jsonl]
+
+# Example
+python3 evaluation/benchmarks/swe_bench/scripts/swtbench/convert.py --prediction_file "evaluation/evaluation_outputs/outputs/princeton-nlp__SWE-bench_Verified-test/CodeActAgent/gpt-4o-2024-11-20_maxiter_100_N_v0.31.0-no-hint-swt-run_1/output.jsonl" > OpenHands-gpt-4o-2024-11-20.jsonl
+```
+
+#### Running the results in SWT-Bench
+
+Next, we run the [SWT-Bench evaluation harness](https://github.com/logic-star-ai/swt-bench/tree/master) with these results.
+First set-up and validate the setup as described in the harness [here](https://github.com/logic-star-ai/swt-bench/tree/master?tab=readme-ov-file#-set-up).
+Then, run the evaluation with the following command:
+
+```bash
+# Example
+python3 -m src.main \
+    --dataset_name princeton-nlp/SWE-bench_Verified \
+    --predictions_path <pathTo>/OpenHands-gpt-4o-2024-11-20.jsonl \
+    --max_workers 12 \
+    --run_id OpenHands-CodeAct-gpt-4o-2024-11-20  --patch_types vanilla  --build_mode api
+```
+
+The results of the evaluation can be obtained by running the reporting script of the harness.
+
+```bash
+# Example
+python -m src.report run_instance_swt_logs/OpenHands-CodeAct-gpt-4o-2024-11-20/OpenHands__CodeActAgent__gpt-4o-2024-11-20 --dataset verified
+```