Fix non-function calls messages (#5026)

Co-authored-by: Xingyao Wang <xingyao@all-hands.dev>

.github/ISSUE_TEMPLATE/bug_template.yml

@@ -31,6 +31,8 @@ body:
       options:
         - Docker command in README
         - Development workflow
+        - app.all-hands.dev
+        - Other
       default: 0
 
   - type: input

.github/workflows/ghcr-build.yml

@@ -286,7 +286,6 @@ jobs:
           image_name=ghcr.io/${{ github.repository_owner }}/runtime:${{ env.RELEVANT_SHA }}-${{ matrix.base_image }}
           image_name=$(echo $image_name | tr '[:upper:]' '[:lower:]')
 
-          SKIP_CONTAINER_LOGS=true \
           TEST_RUNTIME=eventstream \
           SANDBOX_USER_ID=$(id -u) \
           SANDBOX_RUNTIME_CONTAINER_IMAGE=$image_name \
@@ -364,7 +363,6 @@ jobs:
           image_name=ghcr.io/${{ github.repository_owner }}/runtime:${{ env.RELEVANT_SHA }}-${{ matrix.base_image }}
           image_name=$(echo $image_name | tr '[:upper:]' '[:lower:]')
 
-          SKIP_CONTAINER_LOGS=true \
           TEST_RUNTIME=eventstream \
           SANDBOX_USER_ID=$(id -u) \
           SANDBOX_RUNTIME_CONTAINER_IMAGE=$image_name \
@@ -424,9 +422,9 @@ jobs:
             -p 3000:3000 \
             -v /var/run/docker.sock:/var/run/docker.sock \
             --add-host host.docker.internal:host-gateway \
-            -e SANDBOX_RUNTIME_CONTAINER_IMAGE=ghcr.io/all-hands-ai/runtime:$SHORT_SHA-nikolaik \
+            -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:$SHORT_SHA-nikolaik \
             --name openhands-app-$SHORT_SHA \
-            ghcr.io/all-hands-ai/runtime:$SHORT_SHA"
+            docker.all-hands.dev/all-hands-ai/openhands:$SHORT_SHA"
 
           PR_BODY=$(gh pr view $PR_NUMBER --json body --jq .body)
 

.github/workflows/lint-fix.yml

@@ -0,0 +1,62 @@
+name: Lint Fix
+
+on:
+  pull_request:
+    types: [labeled]
+
+jobs:
+  lint-fix:
+    if: github.event.label.name == 'lint-fix'
+    name: Fix linting issues
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write
+      pull-requests: write
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          ref: ${{ github.head_ref }}
+          repository: ${{ github.event.pull_request.head.repo.full_name }}
+          fetch-depth: 0
+          token: ${{ secrets.GITHUB_TOKEN }}
+
+      # Frontend lint fixes
+      - name: Install Node.js 20
+        uses: actions/setup-node@v4
+        with:
+          node-version: 20
+      - name: Install frontend dependencies
+        run: |
+          cd frontend
+          npm install --frozen-lockfile
+      - name: Fix frontend lint issues
+        run: |
+          cd frontend
+          npm run lint:fix
+
+      # Python lint fixes
+      - name: Set up python
+        uses: actions/setup-python@v5
+        with:
+          python-version: 3.12
+          cache: 'pip'
+      - name: Install pre-commit
+        run: pip install pre-commit==3.7.0
+      - name: Fix python lint issues
+        run: |
+          # Run all pre-commit hooks and continue even if they modify files (exit code 1)
+          pre-commit run --config ./dev_config/python/.pre-commit-config.yaml --files openhands/**/* evaluation/**/* tests/**/* || true
+
+      # Commit and push changes if any
+      - name: Check for changes
+        id: git-check
+        run: |
+          git diff --quiet || echo "changes=true" >> $GITHUB_OUTPUT
+      - name: Commit and push if there are changes
+        if: steps.git-check.outputs.changes == 'true'
+        run: |
+          git config --local user.email "openhands@all-hands.dev"
+          git config --local user.name "OpenHands Bot"
+          git add -A
+          git commit -m "🤖 Auto-fix linting issues"
+          git push

.github/workflows/openhands-resolver.yml

@@ -1,15 +1,283 @@
-name: Resolve Issues with OpenHands
+name: Auto-Fix Tagged Issue with OpenHands
 
 on:
+  workflow_call:
+    inputs:
+      max_iterations:
+        required: false
+        type: number
+        default: 50
+      macro:
+        required: false
+        type: string
+        default: "@openhands-agent"
+    secrets:
+      LLM_MODEL:
+        required: true
+      LLM_API_KEY:
+        required: true
+      LLM_BASE_URL:
+        required: false
+      PAT_TOKEN:
+        required: true
+      PAT_USERNAME:
+        required: true
+
   issues:
     types: [labeled]
   pull_request:
     types: [labeled]
+  issue_comment:
+    types: [created]
+  pull_request_review_comment:
+    types: [created]
+  pull_request_review:
+    types: [submitted]
+
+permissions:
+  contents: write
+  pull-requests: write
+  issues: write
 
 jobs:
-  call-openhands-resolver:
-    uses: All-Hands-AI/openhands-resolver/.github/workflows/openhands-resolver.yml@main
-    if: github.event.label.name == 'fix-me'
-    with:
-      issue_number: ${{ github.event.issue.number }}
-    secrets: inherit
+  auto-fix:
+    if: |
+      github.event_name == 'workflow_call' ||
+      github.event.label.name == 'fix-me' ||
+      github.event.label.name == 'fix-me-experimental' ||
+
+      (
+        ((github.event_name == 'issue_comment' || github.event_name == 'pull_request_review_comment') &&
+        startsWith(github.event.comment.body, inputs.macro || '@openhands-agent') &&
+        (github.event.comment.author_association == 'OWNER' || github.event.comment.author_association == 'COLLABORATOR' || github.event.comment.author_association == 'MEMBER')
+        ) ||
+
+        (github.event_name == 'pull_request_review' &&
+        startsWith(github.event.review.body, inputs.macro || '@openhands-agent') &&
+        (github.event.review.author_association == 'OWNER' || github.event.review.author_association == 'COLLABORATOR' || github.event.review.author_association == 'MEMBER')
+        )
+      )
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Get latest versions and create requirements.txt
+        run: |
+          python -m pip index versions openhands-ai > openhands_versions.txt
+          OPENHANDS_VERSION=$(head -n 1 openhands_versions.txt | awk '{print $2}' | tr -d '()')
+          echo "openhands-ai==${OPENHANDS_VERSION}" >> requirements.txt
+          cat requirements.txt
+
+      - name: Cache pip dependencies
+        if: |
+          !(
+            github.event.label.name == 'fix-me-experimental' ||
+            (
+              (github.event_name == 'issue_comment' || github.event_name == 'pull_request_review_comment') &&
+              startsWith(github.event.comment.body, '@openhands-agent-exp')
+            ) ||
+            (
+              github.event_name == 'pull_request_review' &&
+              startsWith(github.event.review.body, '@openhands-agent-exp')
+            )
+          )
+        uses: actions/cache@v3
+        with:
+          path: ${{ env.pythonLocation }}/lib/python3.12/site-packages/*
+          key: ${{ runner.os }}-pip-openhands-resolver-${{ hashFiles('requirements.txt') }}
+          restore-keys: |
+            ${{ runner.os }}-pip-openhands-resolver-${{ hashFiles('requirements.txt') }}
+
+      - name: Check required environment variables
+        env:
+          LLM_MODEL: ${{ secrets.LLM_MODEL }}
+          LLM_API_KEY: ${{ secrets.LLM_API_KEY }}
+          LLM_BASE_URL: ${{ secrets.LLM_BASE_URL }}
+          PAT_TOKEN: ${{ secrets.PAT_TOKEN }}
+          PAT_USERNAME: ${{ secrets.PAT_USERNAME }}
+        run: |
+          required_vars=("LLM_MODEL" "LLM_API_KEY" "PAT_TOKEN" "PAT_USERNAME")
+          for var in "${required_vars[@]}"; do
+            if [ -z "${!var}" ]; then
+              echo "Error: Required environment variable $var is not set."
+              exit 1
+            fi
+          done
+
+      - name: Set environment variables
+        run: |
+          if [ -n "${{ github.event.review.body }}" ]; then
+            echo "ISSUE_NUMBER=${{ github.event.pull_request.number }}" >> $GITHUB_ENV
+            echo "ISSUE_TYPE=pr" >> $GITHUB_ENV
+          elif [ -n "${{ github.event.issue.pull_request }}" ]; then
+            echo "ISSUE_NUMBER=${{ github.event.issue.number }}" >> $GITHUB_ENV
+            echo "ISSUE_TYPE=pr" >> $GITHUB_ENV
+          elif [ -n "${{ github.event.pull_request.number }}" ]; then
+            echo "ISSUE_NUMBER=${{ github.event.pull_request.number }}" >> $GITHUB_ENV
+            echo "ISSUE_TYPE=pr" >> $GITHUB_ENV
+          else
+            echo "ISSUE_NUMBER=${{ github.event.issue.number }}" >> $GITHUB_ENV
+            echo "ISSUE_TYPE=issue" >> $GITHUB_ENV
+          fi
+
+          if [ -n "${{ github.event.review.body }}" ]; then
+            echo "COMMENT_ID=${{ github.event.review.id || 'None' }}" >> $GITHUB_ENV
+          else
+            echo "COMMENT_ID=${{ github.event.comment.id || 'None' }}" >> $GITHUB_ENV
+          fi
+
+          echo "MAX_ITERATIONS=${{ inputs.max_iterations || 50 }}" >> $GITHUB_ENV
+          echo "SANDBOX_ENV_GITHUB_TOKEN=${{ secrets.GITHUB_TOKEN }}" >> $GITHUB_ENV
+
+      - name: Comment on issue with start message
+        uses: actions/github-script@v7
+        with:
+          github-token: ${{secrets.GITHUB_TOKEN}}
+          script: |
+            const issueType = process.env.ISSUE_TYPE;
+            github.rest.issues.createComment({
+              issue_number: ${{ env.ISSUE_NUMBER }},
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              body: `[OpenHands](https://github.com/All-Hands-AI/OpenHands) started fixing the ${issueType}! You can monitor the progress [here](https://github.com/${context.repo.owner}/${context.repo.repo}/actions/runs/${context.runId}).`
+            });
+
+      - name: Install OpenHands
+        run: |
+          if [[ "${{ github.event.label.name }}" == "fix-me-experimental" ]] ||
+             ([[ "${{ github.event_name }}" == "issue_comment" || "${{ github.event_name }}" == "pull_request_review_comment" ]] &&
+              [[ "${{ github.event.comment.body }}" == "@openhands-agent-exp"* ]]) ||
+             ([[ "${{ github.event_name }}" == "pull_request_review" ]] &&
+              [[ "${{ github.event.review.body }}" == "@openhands-agent-exp"* ]]); then
+            python -m pip install --upgrade pip
+            pip install git+https://github.com/all-hands-ai/openhands.git
+          else
+            python -m pip install --upgrade -r requirements.txt
+          fi
+
+      - name: Attempt to resolve issue
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          GITHUB_USERNAME: ${{ secrets.PAT_USERNAME }}
+          LLM_MODEL: ${{ secrets.LLM_MODEL }}
+          LLM_API_KEY: ${{ secrets.LLM_API_KEY }}
+          LLM_BASE_URL: ${{ secrets.LLM_BASE_URL }}
+          PYTHONPATH: ""
+        run: |
+          cd /tmp && python -m openhands.resolver.resolve_issue \
+            --repo ${{ github.repository }} \
+            --issue-number ${{ env.ISSUE_NUMBER }} \
+            --issue-type ${{ env.ISSUE_TYPE }} \
+            --max-iterations ${{ env.MAX_ITERATIONS }} \
+            --comment-id ${{ env.COMMENT_ID }}
+
+      - name: Check resolution result
+        id: check_result
+        run: |
+          if cd /tmp && grep -q '"success":true' output/output.jsonl; then
+            echo "RESOLUTION_SUCCESS=true" >> $GITHUB_OUTPUT
+          else
+            echo "RESOLUTION_SUCCESS=false" >> $GITHUB_OUTPUT
+          fi
+
+      - name: Upload output.jsonl as artifact
+        uses: actions/upload-artifact@v4
+        if: always() # Upload even if the previous steps fail
+        with:
+          name: resolver-output
+          path: /tmp/output/output.jsonl
+          retention-days: 30 # Keep the artifact for 30 days
+
+      - name: Create draft PR or push branch
+        if: always() # Create PR or branch even if the previous steps fail
+        env:
+          GITHUB_TOKEN: ${{ secrets.PAT_TOKEN }}
+          GITHUB_USERNAME: ${{ secrets.PAT_USERNAME }}
+          LLM_MODEL: ${{ secrets.LLM_MODEL }}
+          LLM_API_KEY: ${{ secrets.LLM_API_KEY }}
+          LLM_BASE_URL: ${{ secrets.LLM_BASE_URL }}
+          PYTHONPATH: ""
+        run: |
+          if [ "${{ steps.check_result.outputs.RESOLUTION_SUCCESS }}" == "true" ]; then
+            cd /tmp && python -m openhands.resolver.send_pull_request \
+              --issue-number ${{ env.ISSUE_NUMBER }} \
+              --pr-type draft | tee pr_result.txt && \
+              grep "draft 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 }} \
+              --pr-type branch \
+              --send-on-failure | tee branch_result.txt && \
+              grep "branch created" branch_result.txt | sed 's/.*\///g; s/.expand=1//g' > branch_name.txt
+          fi
+
+      - name: Comment on issue
+        uses: actions/github-script@v7
+        if: always() # Comment on issue even if the previous steps fail
+        with:
+          github-token: ${{secrets.GITHUB_TOKEN}}
+          script: |
+            const fs = require('fs');
+            const issueNumber = ${{ env.ISSUE_NUMBER }};
+            const success = ${{ steps.check_result.outputs.RESOLUTION_SUCCESS }};
+
+            let prNumber = '';
+            let branchName = '';
+            let logContent = '';
+            const noChangesMessage = `No changes to commit for issue #${issueNumber}. Skipping commit.`;
+
+            try {
+              if (success){
+                logContent = fs.readFileSync('/tmp/pr_result.txt', 'utf8').trim();
+              } else {
+                logContent = fs.readFileSync('/tmp/branch_result.txt', 'utf8').trim();
+              }
+            } catch (error) {
+              console.error('Error reading results file:', error);
+            }
+
+            try {
+              if (success) {
+                prNumber = fs.readFileSync('/tmp/pr_number.txt', 'utf8').trim();
+              } else {
+                branchName = fs.readFileSync('/tmp/branch_name.txt', 'utf8').trim();
+              }
+            } catch (error) {
+              console.error('Error reading file:', error);
+            }
+
+            if (logContent.includes(noChangesMessage)) {
+              github.rest.issues.createComment({
+                issue_number: issueNumber,
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                body: `The workflow to fix this issue encountered an error. Openhands failed to create any code changes.`
+              });
+            } else if (success && prNumber) {
+              github.rest.issues.createComment({
+                issue_number: issueNumber,
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                body: `A potential fix has been generated and a draft PR #${prNumber} has been created. Please review the changes.`
+              });
+            } else if (!success && branchName) {
+              github.rest.issues.createComment({
+                issue_number: issueNumber,
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                body: `An attempt was made to automatically fix this issue, but it was unsuccessful. A branch named '${branchName}' has been created with the attempted changes. You can view the branch [here](https://github.com/${context.repo.owner}/${context.repo.repo}/tree/${branchName}). Manual intervention may be required.`
+              });
+            } else {
+              github.rest.issues.createComment({
+                issue_number: issueNumber,
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                body: `The workflow to fix this issue encountered an error. Please check the [workflow logs](https://github.com/${context.repo.owner}/${context.repo.repo}/actions/runs/${context.runId}) for more information.`
+              });
+            }

.gitignore

@@ -176,6 +176,9 @@ evaluation/gorilla/data
 evaluation/toolqa/data
 evaluation/scienceagentbench/benchmark
 
+# openhands resolver
+output/
+
 # frontend
 
 # dependencies

COMMUNITY.md

@@ -0,0 +1,43 @@
+# 🙌 The OpenHands Community
+
+The OpenHands community is built around the belief that (1) AI and AI agents are going to fundamentally change the way
+we build software, and (2) if this is true, we should do everything we can to make sure that the benefits provided by
+such powerful technology are accessible to everyone.
+
+If this resonates with you, we'd love to have you join us in our quest!
+
+## 🤝 How to Join
+
+Check out our [How to Join the Community section.](https://github.com/All-Hands-AI/OpenHands?tab=readme-ov-file#-how-to-join-the-community)
+
+## 💪 Becoming a Contributor
+
+We welcome contributions from everyone! Whether you're a developer, a researcher, or simply enthusiastic about advancing
+the field of software engineering with AI, there are many ways to get involved:
+
+- **Code Contributions:** Help us develop new core functionality, improve our agents, improve the frontend and other
+interfaces, or anything else that would help make OpenHands better.
+- **Research and Evaluation:** Contribute to our understanding of LLMs in software engineering, participate in
+evaluating the models, or suggest improvements.
+- **Feedback and Testing:** Use the OpenHands toolset, report bugs, suggest features, or provide feedback on usability.
+
+For details, please check [CONTRIBUTING.md](./CONTRIBUTING.md).
+
+## Code of Conduct
+
+We have a [Code of Conduct](./CODE_OF_CONDUCT.md) that we expect all contributors to adhere to.
+Long story short, we are aiming for an open, welcoming, diverse, inclusive, and healthy community.
+All contributors are expected to contribute to building this sort of community.
+
+## 🛠️ Becoming a Maintainer
+
+For contributors who have made significant and sustained contributions to the project, there is a possibility of joining
+the maintainer team. The process for this is as follows:
+
+1. Any contributor who has made sustained and high-quality contributions to the codebase can be nominated by any
+maintainer. If you feel that you may qualify you can reach out to any of the maintainers that have reviewed your PRs and ask if you can be nominated.
+2. Once a maintainer nominates a new maintainer, there will be a discussion period among the maintainers for at least 3 days.
+3. If no concerns are raised the nomination will be accepted by acclamation, and if concerns are raised there will be a discussion and possible vote.
+
+Note that just making many PRs does not immediately imply that you will become a maintainer. We will be looking
+at sustained high-quality contributions over a period of time, as well as good teamwork and adherence to our [Code of Conduct](./CODE_OF_CONDUCT.md).

CONTRIBUTING.md

@@ -54,7 +54,7 @@ The agent needs a place to run code and commands. When you run OpenHands on your
 to do this by default. But there are other ways of creating a sandbox for the agent.
 
 If you work for a company that provides a cloud-based runtime, you could help us add support for that runtime
-by implementing the [interface specified here](https://github.com/All-Hands-AI/OpenHands/blob/main/openhands/runtime/runtime.py).
+by implementing the [interface specified here](https://github.com/All-Hands-AI/OpenHands/blob/main/openhands/runtime/base.py).
 
 #### Testing
 When you write code, it is also good to write tests. Please navigate to the `tests` folder to see existing test suites.
@@ -92,3 +92,32 @@ You may also check out previous PRs in the [PR list](https://github.com/All-Hand
 
 If your changes are user-facing (e.g. a new feature in the UI, a change in behavior, or a bugfix)
 please include a short message that we can add to our changelog.
+
+## How to Make Effective Contributions
+
+### Opening Issues
+
+If you notice any bugs or have any feature requests please open them via the [issues page](https://github.com/All-Hands-AI/OpenHands/issues). We will triage based on how critical the bug is or how potentially useful the improvement is, discuss, and implement the ones that the community has interest/effort for.
+
+Further, if you see an issue you like, please leave a "thumbs-up" or a comment, which will help us prioritize.
+
+### Making Pull Requests
+
+We're generally happy to consider all PRs, with the evaluation process varying based on the type of change:
+
+#### For Small Improvements
+
+Small improvements with few downsides are typically reviewed and approved quickly.
+One thing to check when making changes is to ensure that all continuous integration tests pass, which you can check before getting a review.
+
+#### For Core Agent Changes
+
+We need to be more careful with changes to the core agent, as it is imperative to maintain high quality. These PRs are evaluated based on three key metrics:
+
+1. **Accuracy**
+2. **Efficiency**
+3. **Code Complexity**
+
+If it improves accuracy, efficiency, or both with only a minimal change to code quality, that's great we're happy to merge it in!
+If there are bigger tradeoffs (e.g. helping efficiency a lot and hurting accuracy a little) we might want to put it behind a feature flag.
+Either way, please feel free to discuss on github issues or slack, and we will give guidance and preliminary feedback.

Development.md

@@ -38,7 +38,9 @@ make build
 ```
 
 ### 3. Configuring the Language Model
-OpenHands supports a diverse array of Language Models (LMs) through the powerful [litellm](https://docs.litellm.ai) library. By default, we've chosen the mighty GPT-4 from OpenAI as our go-to model, but the world is your oyster! You can unleash the potential of Anthropic's suave Claude, the enigmatic Llama, or any other LM that piques your interest.
+OpenHands supports a diverse array of Language Models (LMs) through the powerful [litellm](https://docs.litellm.ai) library.
+By default, we've chosen Claude Sonnet 3.5 as our go-to model, but the world is your oyster! You can unleash the
+potential of any other LM that piques your interest.
 
 To configure the LM of your choice, run:
 
@@ -52,10 +54,7 @@ To configure the LM of your choice, run:
    Environment variables > config.toml variables > default variables
 
 **Note on Alternative Models:**
-Some alternative models may prove more challenging to tame than others. Fear not, brave adventurer! We shall soon unveil LLM-specific documentation to guide you on your quest.
-And if you've already mastered the art of wielding a model other than OpenAI's GPT, we encourage you to share your setup instructions with us by creating instructions and adding it [to our documentation](https://github.com/All-Hands-AI/OpenHands/tree/main/docs/modules/usage/llms).
-
-For a full list of the LM providers and models available, please consult the [litellm documentation](https://docs.litellm.ai/docs/providers).
+See [our documentation](https://docs.all-hands.dev/modules/usage/llms) for recommended models.
 
 ### 4. Running the application
 #### Option A: Run the Full Application
@@ -98,9 +97,10 @@ poetry run pytest ./tests/unit/test_*.py
 2. Update the poetry.lock file via `poetry lock --no-update`
 
 ### 9. Use existing Docker image
-To reduce build time (e.g., if no changes were made to the client-runtime component), you can use an existing Docker container image. Follow these steps:
-1. Set the SANDBOX_RUNTIME_CONTAINER_IMAGE environment variable to the desired Docker image.
-2. Example: export SANDBOX_RUNTIME_CONTAINER_IMAGE=ghcr.io/all-hands-ai/runtime:0.9-nikolaik
+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:0.14-nikolaik`
 
 ## Develop inside Docker container
 

ISSUE_TRIAGE.md

@@ -6,9 +6,9 @@ These are the procedures and guidelines on how issues are triaged in this repo b
 * Issues may be tagged with what it relates to (**backend**, **frontend**, **agent quality**, etc.)
 
 ## Severity
-* **Low**: Minor issues, single user report
-* **Medium**: Affecting multiple users
-* **Critical**: Affecting all users or potential security issues
+* **Low**: Minor issues or affecting single user.
+* **Medium**: Affecting multiple users.
+* **Critical**: Affecting all users or potential security issues.
 
 ## Effort
 * Issues may be estimated with effort required (**small effort**, **medium effort**, **large effort**)
@@ -17,9 +17,9 @@ These are the procedures and guidelines on how issues are triaged in this repo b
 * Issues with low implementation difficulty may be tagged with **good first issue**
 
 ## Not Enough Information
-* User is asked to provide more information (logs, how to reproduce, etc.) when the issue is not clear
-* If an issue is unclear and the author does not provide more information or respond to a request, the issue may be closed as **not planned** (Usually after a week)
+* User is asked to provide more information (logs, how to reproduce, etc.) when the issue is not clear.
+* If an issue is unclear and the author does not provide more information or respond to a request, the issue may be closed as **not planned** (Usually after a week).
 
 ## Multiple Requests/Fixes in One Issue
-* These issues will be narrowed down to one request/fix so the issue is more easily tracked and fixed
-* Issues may be broken down into multiple issues if required
+* These issues will be narrowed down to one request/fix so the issue is more easily tracked and fixed.
+* Issues may be broken down into multiple issues if required.

README.md

@@ -38,15 +38,16 @@ See the [Installation](https://docs.all-hands.dev/modules/usage/installation) gu
 system requirements and more information.
 
 ```bash
-docker pull docker.all-hands.dev/all-hands-ai/runtime:0.12-nikolaik
+docker pull docker.all-hands.dev/all-hands-ai/runtime:0.14-nikolaik
 
-docker run -it --rm --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.12-nikolaik \
+docker run -it --pull=always \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.14-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:0.12
+    docker.all-hands.dev/all-hands-ai/openhands:0.14
 ```
 
 You'll find OpenHands running at [http://localhost:3000](http://localhost:3000)!
@@ -60,7 +61,7 @@ works best, but you have [many options](https://docs.all-hands.dev/modules/usage
 You can also [connect OpenHands to your local filesystem](https://docs.all-hands.dev/modules/usage/runtimes),
 run OpenHands in a scriptable [headless mode](https://docs.all-hands.dev/modules/usage/how-to/headless-mode),
 interact with it via a [friendly CLI](https://docs.all-hands.dev/modules/usage/how-to/cli-mode),
-or run it on tagged issues with [a github action](https://github.com/All-Hands-AI/OpenHands-resolver).
+or run it on tagged issues with [a github action](https://github.com/All-Hands-AI/OpenHands/blob/main/openhands/resolver/README.md).
 
 Visit [Installation](https://docs.all-hands.dev/modules/usage/installation) for more information and setup instructions.
 
@@ -76,25 +77,16 @@ To learn more about the project, and for tips on using OpenHands,
 There you'll find resources on how to use different LLM providers,
 troubleshooting resources, and advanced configuration options.
 
-## 🤝 How to Contribute
+## 🤝 How to Join the Community
 
-OpenHands is a community-driven project, and we welcome contributions from everyone.
-Whether you're a developer, a researcher, or simply enthusiastic about advancing the field of
-software engineering with AI, there are many ways to get involved:
+OpenHands is a community-driven project, and we welcome contributions from everyone. We do most of our communication
+through Slack, so this is the best place to start, but we also are happy to have you contact us on Discord or Github:
 
-- **Code Contributions:** Help us develop new agents, core functionality, the frontend and other interfaces, or sandboxing solutions.
-- **Research and Evaluation:** Contribute to our understanding of LLMs in software engineering, participate in evaluating the models, or suggest improvements.
-- **Feedback and Testing:** Use the OpenHands toolset, report bugs, suggest features, or provide feedback on usability.
+- [Join our Slack workspace](https://join.slack.com/t/openhands-ai/shared_invite/zt-2tom0er4l-JeNUGHt_AxpEfIBstbLPiw) - Here we talk about research, architecture, and future development.
+- [Join our Discord server](https://discord.gg/ESHStjSjD4) - This is a community-run server for general discussion, questions, and feedback.
+- [Read or post Github Issues](https://github.com/All-Hands-AI/OpenHands/issues) - Check out the issues we're working on, or add your own ideas.
 
-For details, please check [CONTRIBUTING.md](./CONTRIBUTING.md).
-
-## 🤖 Join Our Community
-
-Whether you're a developer, a researcher, or simply enthusiastic about OpenHands, we'd love to have you in our community.
-Let's make software engineering better together!
-
-- [Slack workspace](https://join.slack.com/t/openhands-ai/shared_invite/zt-2tom0er4l-JeNUGHt_AxpEfIBstbLPiw) - Here we talk about research, architecture, and future development.
-- [Discord server](https://discord.gg/ESHStjSjD4) - This is a community-run server for general discussion, questions, and feedback.
+See more about the community in [COMMUNITY.md](./COMMUNITY.md) or find details on contributing in [CONTRIBUTING.md](./CONTRIBUTING.md).
 
 ## 📈 Progress
 

compose.yml

@@ -7,7 +7,7 @@ services:
     image: openhands:latest
     container_name: openhands-app-${DATE:-}
     environment:
-      - SANDBOX_RUNTIME_CONTAINER_IMAGE=${SANDBOX_RUNTIME_CONTAINER_IMAGE:-ghcr.io/all-hands-ai/runtime:0.9-nikolaik}
+      - SANDBOX_RUNTIME_CONTAINER_IMAGE=${SANDBOX_RUNTIME_CONTAINER_IMAGE:-ghcr.io/all-hands-ai/runtime:0.14-nikolaik}
       - SANDBOX_USER_ID=${SANDBOX_USER_ID:-1234}
       - WORKSPACE_MOUNT_PATH=${WORKSPACE_BASE:-$PWD/workspace}
     ports:

config.template.toml

@@ -32,7 +32,8 @@ workspace_base = "./workspace"
 # Enable saving and restoring the session when run from CLI
 #enable_cli_session = false
 
-# Path to store trajectories
+# Path to store trajectories, can be a folder or a file
+# If it's a folder, the session id will be used as the file name
 #trajectories_path="./trajectories"
 
 # File store path

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:0.9-nikolaik}
+      - SANDBOX_RUNTIME_CONTAINER_IMAGE=${SANDBOX_RUNTIME_CONTAINER_IMAGE:-ghcr.io/all-hands-ai/runtime:0.14-nikolaik}
       - SANDBOX_USER_ID=${SANDBOX_USER_ID:-1234}
       - WORKSPACE_MOUNT_PATH=${WORKSPACE_BASE:-$PWD/workspace}
     ports:

docs/i18n/fr/docusaurus-plugin-content-docs/current/usage/how-to/evaluation-harness.md

@@ -161,7 +161,7 @@ Pour créer un workflow d'évaluation pour votre benchmark, suivez ces étapes :
            instruction=instruction,
            test_result=evaluation_result,
            metadata=metadata,
-           history=state.history.compatibility_for_eval_history_pairs(),
+           history=compatibility_for_eval_history_pairs(state.history),
            metrics=state.metrics.get() if state.metrics else None,
            error=state.last_error if state and state.last_error else None,
        )
@@ -260,7 +260,7 @@ def codeact_user_response(state: State | None) -> str:
         # vérifier si l'agent a essayé de parler à l'utilisateur 3 fois, si oui, faire savoir à l'agent qu'il peut abandonner
         user_msgs = [
             event
-            for event in state.history.get_events()
+            for event in state.history
             if isinstance(event, MessageAction) and event.source == 'user'
         ]
         if len(user_msgs) >= 2:
@@ -279,4 +279,3 @@ Cette fonction fait ce qui suit :
 3. Si l'agent a fait plusieurs tentatives, il lui donne la possibilité d'abandonner
 
 En utilisant cette fonction, vous pouvez garantir un comportement cohérent sur plusieurs exécutions d'évaluation et empêcher l'agent de rester bloqué en attendant une entrée humaine.
-

docs/i18n/fr/docusaurus-plugin-content-docs/current/usage/how-to/github-action.md

@@ -14,4 +14,4 @@ Pour utiliser l'Action GitHub OpenHands dans le dépôt OpenHands, un mainteneur
 
 ## Installation de l'Action dans un nouveau dépôt
 
-Pour installer l'Action GitHub OpenHands dans votre propre dépôt, suivez les [instructions dans le dépôt OpenHands Resolver](https://github.com/All-Hands-AI/OpenHands-resolver?tab=readme-ov-file#using-the-github-actions-workflow).
+Pour installer l'Action GitHub OpenHands dans votre propre dépôt, suivez les [instructions dans le dépôt OpenHands Resolver](https://github.com/All-Hands-AI/OpenHands/blob/main/openhands/resolver/README.md).

docs/i18n/zh-Hans/docusaurus-plugin-content-docs/current/usage/how-to/evaluation-harness.md

@@ -158,7 +158,7 @@ OpenHands 的主要入口点在 `openhands/core/main.py` 中。以下是它工
            instruction=instruction,
            test_result=evaluation_result,
            metadata=metadata,
-           history=state.history.compatibility_for_eval_history_pairs(),
+           history=compatibility_for_eval_history_pairs(state.history),
            metrics=state.metrics.get() if state.metrics else None,
            error=state.last_error if state and state.last_error else None,
        )
@@ -257,7 +257,7 @@ def codeact_user_response(state: State | None) -> str:
         # 检查代理是否已尝试与用户对话 3 次，如果是，让代理知道它可以放弃
         user_msgs = [
             event
-            for event in state.history.get_events()
+            for event in state.history
             if isinstance(event, MessageAction) and event.source == 'user'
         ]
         if len(user_msgs) >= 2:

docs/i18n/zh-Hans/docusaurus-plugin-content-docs/current/usage/how-to/github-action.md

@@ -12,4 +12,4 @@
 
 ## 在新仓库中安装 Action
 
-要在你自己的仓库中安装 OpenHands GitHub Action，请按照 [OpenHands Resolver 仓库中的说明](https://github.com/All-Hands-AI/OpenHands-resolver?tab=readme-ov-file#using-the-github-actions-workflow) 进行操作。
+要在你自己的仓库中安装 OpenHands GitHub Action，请按照 [OpenHands Resolver 仓库中的说明](https://github.com/All-Hands-AI/OpenHands/blob/main/openhands/resolver/README.md) 进行操作。

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

@@ -58,4 +58,3 @@ docker run -it \
     ghcr.io/all-hands-ai/openhands:0.11 \
     python -m openhands.core.main -t "write a bash script that prints hi"
 ```
-

docs/modules/usage/about.md

@@ -1,6 +1,6 @@
-# 📚 Misc
+# About OpenHands
 
-## ⭐️ Research Strategy
+## Research Strategy
 
 Achieving full replication of production-grade applications with LLMs is a complex endeavor. Our strategy involves:
 
@@ -9,34 +9,11 @@ Achieving full replication of production-grade applications with LLMs is a compl
 3. **Task Planning:** Developing capabilities for bug detection, codebase management, and optimization
 4. **Evaluation:** Establishing comprehensive evaluation metrics to better understand and improve our models
 
-## 🚧 Default Agent
+## Default Agent
 
 Our default Agent is currently the [CodeActAgent](agents), which is capable of generating code and handling files.
 
-## 🤝 How to Contribute
-
-OpenHands is a community-driven project, and we welcome contributions from everyone. Whether you're a developer, a researcher, or simply enthusiastic about advancing the field of software engineering with AI, there are many ways to get involved:
-
-- **Code Contributions:** Help us develop the core functionalities, frontend interface, or sandboxing solutions
-- **Research and Evaluation:** Contribute to our understanding of LLMs in software engineering, participate in evaluating the models, or suggest improvements
-- **Feedback and Testing:** Use the OpenHands toolset, report bugs, suggest features, or provide feedback on usability
-
-For details, please check [this document](https://github.com/All-Hands-AI/OpenHands/blob/main/CONTRIBUTING.md).
-
-## 🤖 Join Our Community
-
-We have both Slack workspace for the collaboration on building OpenHands and Discord server for discussion about anything related, e.g., this project, LLM, agent, etc.
-
-- [Slack workspace](https://join.slack.com/t/opendevin/shared_invite/zt-2oikve2hu-UDxHeo8nsE69y6T7yFX_BA)
-- [Discord server](https://discord.gg/ESHStjSjD4)
-
-If you would love to contribute, feel free to join our community. Let's simplify software engineering together!
-
-🐚 **Code less, make more with OpenHands.**
-
-[![Star History Chart](https://api.star-history.com/svg?repos=All-Hands-AI/OpenHands&type=Date)](https://star-history.com/#All-Hands-AI/OpenHands&Date)
-
-## 🛠️ Built With
+## Built With
 
 OpenHands is built using a combination of powerful frameworks and libraries, providing a robust foundation for its development. Here are the key technologies used in the project:
 
@@ -44,6 +21,6 @@ OpenHands is built using a combination of powerful frameworks and libraries, pro
 
 Please note that the selection of these technologies is in progress, and additional technologies may be added or existing ones may be removed as the project evolves. We strive to adopt the most suitable and efficient tools to enhance the capabilities of OpenHands.
 
-## 📜 License
+## License
 
-Distributed under the MIT License. See [our license](https://github.com/All-Hands-AI/OpenHands/blob/main/LICENSE) for more information.
+Distributed under MIT [License](https://github.com/All-Hands-AI/OpenHands/blob/main/LICENSE).

docs/modules/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:0.12-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.14-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:0.12 \
+    docker.all-hands.dev/all-hands-ai/openhands:0.14 \
     python -m openhands.core.cli
 ```
 

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

@@ -62,25 +62,3 @@ Run OpenHands by running ```make run``` in the top level directory.
 ## Technical Explanation
 
 Please refer to [custom docker image section of the runtime documentation](https://docs.all-hands.dev/modules/usage/architecture/runtime#advanced-how-openhands-builds-and-maintains-od-runtime-images) for more details.
-
-## Troubleshooting / Errors
-
-### Error: ```useradd: UID 1000 is not unique```
-
-If you see this error in the console output it is because OpenHands is trying to create the openhands user in the sandbox with a UID of 1000, however this UID is already being used in the image (for some reason). To fix this change the sandbox_user_id field in the config.toml file to a different value:
-
-```toml
-[core]
-workspace_base="./workspace"
-run_as_openhands=true
-sandbox_base_container_image="custom_image"
-sandbox_user_id="1001"
-```
-
-### Port use errors
-
-If you see an error about a port being in use or unavailable, try deleting all running Docker Containers (run `docker ps` and `docker rm` relevant containers) and then re-running ```make run``` .
-
-## Discuss
-
-For other issues or questions join the [Slack](https://join.slack.com/t/opendevin/shared_invite/zt-2oikve2hu-UDxHeo8nsE69y6T7yFX_BA) or [Discord](https://discord.gg/ESHStjSjD4) and ask!

docs/modules/usage/how-to/evaluation-harness.md

@@ -158,7 +158,7 @@ To create an evaluation workflow for your benchmark, follow these steps:
            instruction=instruction,
            test_result=evaluation_result,
            metadata=metadata,
-           history=state.history.compatibility_for_eval_history_pairs(),
+           history=compatibility_for_eval_history_pairs(state.history),
            metrics=state.metrics.get() if state.metrics else None,
            error=state.last_error if state and state.last_error else None,
        )
@@ -257,7 +257,7 @@ def codeact_user_response(state: State | None) -> str:
         # check if the agent has tried to talk to the user 3 times, if so, let the agent know it can give up
         user_msgs = [
             event
-            for event in state.history.get_events()
+            for event in state.history
             if isinstance(event, MessageAction) and event.source == 'user'
         ]
         if len(user_msgs) >= 2:

docs/modules/usage/how-to/github-action.md

@@ -4,12 +4,92 @@ This guide explains how to use the OpenHands GitHub Action, both within the Open
 
 ## Using the Action in the OpenHands Repository
 
-To use the OpenHands GitHub Action in the OpenHands repository, an OpenHands maintainer can:
+To use the OpenHands GitHub Action in a repository, you can:
 
 1. Create an issue in the repository.
-2. Add the `fix-me` label to the issue.
-3. The action will automatically trigger and attempt to resolve the issue.
+2. Add the `fix-me` label to the issue or leave a comment on the issue starting with `@openhands-agent`.
+
+The action will automatically trigger and attempt to resolve the issue.
 
 ## Installing the Action in a New Repository
 
-To install the OpenHands GitHub Action in your own repository, follow the [directions in the OpenHands Resolver repo](https://github.com/All-Hands-AI/OpenHands-resolver?tab=readme-ov-file#using-the-github-actions-workflow).
+To install the OpenHands GitHub Action in your own repository, follow
+the [README for the OpenHands Resolver](https://github.com/All-Hands-AI/OpenHands/blob/main/openhands/resolver/README.md).
+
+## Usage Tips
+
+### Iterative resolution
+
+1. Create an issue in the repository.
+2. Add the `fix-me` label to the issue, or leave a comment starting with `@openhands-agent`
+3. Review the attempt to resolve the issue by checking the pull request
+4. Follow up with feedback through general comments, review comments, or inline thread comments
+5. Add the `fix-me` label to the pull request, or address a specific comment by starting with `@openhands-agent`
+
+### Label versus Macro
+
+- Label (`fix-me`): Requests OpenHands to address the **entire** issue or pull request.
+- Macro (`@openhands-agent`): Requests OpenHands to consider only the issue/pull request description and **the specific comment**.
+
+## Advanced Settings
+
+### Add custom repository settings
+
+You can provide custom directions for OpenHands by following the [README for the resolver](https://github.com/All-Hands-AI/OpenHands/blob/main/openhands/resolver/README.md#providing-custom-instructions).
+
+### Configure custom macro
+
+To customize the default macro (`@openhands-agent`):
+
+1. [Create a repository variable](https://docs.github.com/en/actions/writing-workflows/choosing-what-your-workflow-does/store-information-in-variables#creating-configuration-variables-for-a-repository) named `OPENHANDS_MACRO`
+2. Assign the variable a custom value
+
+## Writing Effective .openhands_instructions Files
+
+The `.openhands_instructions` file is a file that you can put in the root directory of your repository to guide OpenHands in understanding and working with your repository effectively. Here are key tips for writing high-quality instructions:
+
+### Core Principles
+
+1. **Concise but Informative**: Provide a clear, focused overview of the repository that emphasizes the most common actions OpenHands will need to perform.
+
+2. **Repository Structure**: Explain the key directories and their purposes, especially highlighting where different types of code (e.g., frontend, backend) are located.
+
+3. **Development Workflows**: Document the essential commands for:
+   - Building and setting up the project
+   - Running tests
+   - Linting and code quality checks
+   - Any environment-specific requirements
+
+4. **Testing Guidelines**: Specify:
+   - Where tests are located
+   - How to run specific test suites
+   - Any testing conventions or requirements
+
+### Example Structure
+
+```markdown
+# Repository Overview
+[Brief description of the project]
+
+## General Setup
+- Main build command
+- Development environment setup
+- Pre-commit checks
+
+## Backend
+- Location and structure
+- Testing instructions
+- Environment requirements
+
+## Frontend
+- Setup prerequisites
+- Build and test commands
+- Environment variables
+
+## Additional Guidelines
+- Code style requirements
+- Special considerations
+- Common workflows
+```
+
+For a real-world example, refer to the [OpenHands repository's .openhands_instructions](https://github.com/All-Hands-AI/OpenHands/blob/main/.openhands_instructions).

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

@@ -19,6 +19,15 @@ OpenHands provides a user-friendly Graphical User Interface (GUI) mode for inter
 3. Enter the corresponding `API Key` for your chosen provider.
 4. Click "Save" to apply the settings.
 
+### GitHub Token Setup
+
+OpenHands automatically exports a `GITHUB_TOKEN` to the shell environment if it is available. This can happen in two ways:
+
+1. Locally (OSS): The user directly inputs their GitHub token.
+2. Online (SaaS): The token is obtained through GitHub OAuth authentication.
+
+When you reach the `/app` route, the app checks if a token is present. If it finds one, it sets it in the environment for the agent to use.
+
 ### Advanced Settings
 
 1. Toggle `Advanced Options` to access additional settings.

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

@@ -44,15 +44,16 @@ 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:0.12-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.14-nikolaik \
     -e SANDBOX_USER_ID=$(id -u) \
     -e WORKSPACE_MOUNT_PATH=$WORKSPACE_BASE \
     -e LLM_API_KEY=$LLM_API_KEY \
     -e LLM_MODEL=$LLM_MODEL \
+    -e LOG_ALL_EVENTS=true \
     -v $WORKSPACE_BASE:/opt/workspace_base \
     -v /var/run/docker.sock:/var/run/docker.sock \
     --add-host host.docker.internal:host-gateway \
     --name openhands-app-$(date +%Y%m%d%H%M%S) \
-    docker.all-hands.dev/all-hands-ai/openhands:0.12 \
+    docker.all-hands.dev/all-hands-ai/openhands:0.14 \
     python -m openhands.core.main -t "write a bash script that prints hi"
 ```

docs/modules/usage/installation.mdx

@@ -11,15 +11,16 @@
 The easiest way to run OpenHands is in Docker.
 
 ```bash
-docker pull docker.all-hands.dev/all-hands-ai/runtime:0.12-nikolaik
+docker pull docker.all-hands.dev/all-hands-ai/runtime:0.14-nikolaik
 
 docker run -it --rm --pull=always \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.12-nikolaik \
+    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.14-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:0.12
+    docker.all-hands.dev/all-hands-ai/openhands:0.14
 ```
 
 You can also run OpenHands in a scriptable [headless mode](https://docs.all-hands.dev/modules/usage/how-to/headless-mode), as an [interactive CLI](https://docs.all-hands.dev/modules/usage/how-to/cli-mode), or using the [OpenHands GitHub Action](https://docs.all-hands.dev/modules/usage/how-to/github-action).

docs/modules/usage/llms/litellm-proxy.md

@@ -0,0 +1,20 @@
+# LiteLLM Proxy
+
+OpenHands supports using the [LiteLLM proxy](https://docs.litellm.ai/docs/proxy/quick_start) to access various LLM providers.
+
+## Configuration
+
+To use LiteLLM proxy with OpenHands, you need to:
+
+1. Set up a LiteLLM proxy server (see [LiteLLM documentation](https://docs.litellm.ai/docs/proxy/quick_start))
+2. When running OpenHands, you'll need to set the following in the OpenHands UI through the Settings:
+  * Enable `Advanced Options`
+  * `Custom Model` to the prefix `litellm_proxy/` + the model you will be using (e.g. `litellm_proxy/anthropic.claude-3-5-sonnet-20241022-v2:0`)
+  * `Base URL` to your LiteLLM proxy URL (e.g. `https://your-litellm-proxy.com`)
+  * `API Key` to your LiteLLM proxy API key
+
+## Supported Models
+
+The supported models depend on your LiteLLM proxy configuration. OpenHands supports any model that your LiteLLM proxy is configured to handle.
+
+Refer to your LiteLLM proxy configuration for the list of available models and their names.

docs/modules/usage/llms/llms.md

@@ -4,11 +4,11 @@ OpenHands can connect to any LLM supported by LiteLLM. However, it requires a po
 
 ## Model Recommendations
 
-Based on a recent evaluation of language models for coding tasks (using the SWE-bench dataset), we can provide some recommendations for model selection. The full analysis can be found in [this blog article](https://www.all-hands.dev/blog/evaluation-of-llms-as-coding-agents-on-swe-bench-at-30x-speed).
+Based on our evaluations of language models for coding tasks (using the SWE-bench dataset), we can provide some recommendations for model selection. Some analyses can be found in [this blog article comparing LLMs](https://www.all-hands.dev/blog/evaluation-of-llms-as-coding-agents-on-swe-bench-at-30x-speed) and [this blog article with some more recent results](https://www.all-hands.dev/blog/openhands-codeact-21-an-open-state-of-the-art-software-development-agent).
 
 When choosing a model, consider both the quality of outputs and the associated costs. Here's a summary of the findings:
 
-- Claude 3.5 Sonnet is the best by a fair amount, achieving a 27% resolve rate with the default agent in OpenHands.
+- Claude 3.5 Sonnet is the best by a fair amount, achieving a 53% resolve rate on SWE-Bench Verified with the default agent in OpenHands.
 - GPT-4o lags behind, and o1-mini actually performed somewhat worse than GPT-4o. We went in and analyzed the results a little, and briefly it seemed like o1 was sometimes "overthinking" things, performing extra environment configuration tasks when it could just go ahead and finish the task.
 - Finally, the strongest open models were Llama 3.1 405 B and deepseek-v2.5, and they performed reasonably, even besting some of the closed models.
 
@@ -63,6 +63,7 @@ We have a few guides for running OpenHands with specific model providers:
 - [Azure](llms/azure-llms)
 - [Google](llms/google-llms)
 - [Groq](llms/groq)
+- [LiteLLM Proxy](llms/litellm-proxy)
 - [OpenAI](llms/openai-llms)
 - [OpenRouter](llms/openrouter)
 

docs/modules/usage/runtimes.md

@@ -49,7 +49,7 @@ but seems to work well on most systems.
 
 ## All Hands Runtime
 The All Hands Runtime is currently in beta. You can request access by joining
-the #remote-runtime-limited-beta channel on Slack (see the README for an invite).
+the #remote-runtime-limited-beta channel on Slack ([see the README](https://github.com/All-Hands-AI/OpenHands?tab=readme-ov-file#-join-our-community) for an invite).
 
 To use the All Hands Runtime, set the following environment variables when
 starting OpenHands:
@@ -59,15 +59,14 @@ docker run # ...
     -e RUNTIME=remote \
     -e SANDBOX_REMOTE_RUNTIME_API_URL="https://runtime.app.all-hands.dev" \
     -e SANDBOX_API_KEY="your-all-hands-api-key" \
-    -e SANDBOX_KEEP_REMOTE_RUNTIME_ALIVE="true" \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.11-nikolaik \
+    -e SANDBOX_KEEP_RUNTIME_ALIVE="true" \
     # ...
 ```
 
 ## Modal Runtime
 Our partners at [Modal](https://modal.com/) have also provided a runtime for OpenHands.
 
-To use the Modal Runtime, create an account, and then [create an API key](https://modal.com/settings)
+To use the Modal Runtime, create an account, and then [create an API key.](https://modal.com/settings)
 
 You'll then need to set the following environment variables when starting OpenHands:
 ```bash
@@ -75,5 +74,4 @@ docker run # ...
     -e RUNTIME=modal \
     -e MODAL_API_TOKEN_ID="your-id" \
     -e MODAL_API_TOKEN_SECRET="your-secret" \
-    -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:0.11-nikolaik \
 ```

docs/package.json

@@ -15,10 +15,10 @@
     "typecheck": "tsc"
   },
   "dependencies": {
-    "@docusaurus/core": "^3.5.2",
-    "@docusaurus/plugin-content-pages": "^3.5.2",
-    "@docusaurus/preset-classic": "^3.5.2",
-    "@docusaurus/theme-mermaid": "^3.5.2",
+    "@docusaurus/core": "^3.6.2",
+    "@docusaurus/plugin-content-pages": "^3.6.2",
+    "@docusaurus/preset-classic": "^3.6.2",
+    "@docusaurus/theme-mermaid": "^3.6.2",
     "@mdx-js/react": "^3.1.0",
     "clsx": "^2.0.0",
     "prism-react-renderer": "^2.4.0",
@@ -29,7 +29,7 @@
   },
   "devDependencies": {
     "@docusaurus/module-type-aliases": "^3.5.1",
-    "@docusaurus/tsconfig": "^3.5.2",
+    "@docusaurus/tsconfig": "^3.6.2",
     "@docusaurus/types": "^3.5.1",
     "typescript": "~5.6.3"
   },

docs/sidebars.ts

@@ -76,6 +76,11 @@ const sidebars: SidebarsConfig = {
                   label: 'Groq',
                   id: 'usage/llms/groq',
                 },
+                {
+                  type: 'doc',
+                  label: 'LiteLLM Proxy',
+                  id: 'usage/llms/litellm-proxy',
+                },
                 {
                   type: 'doc',
                   label: 'OpenAI',

evaluation/EDA/game.py

@@ -87,9 +87,7 @@ class Q20Game:
         # others
         bingo, anwser_reply = self.judge_winner(response)
         if bingo:
-            return (
-                'You are bingo! quit now, run: <execute_bash> exit </execute_bash>.\n'
-            )
+            return 'You are bingo! Use the "finish" tool to finish the interaction.\n'
         if self.curr_turn == self.num_turns - 2:
             anwser_reply += " You must guess now, what's it?"
         return anwser_reply

evaluation/EDA/run_infer.py

@@ -8,6 +8,7 @@ from evaluation.EDA.game import Q20Game, Q20GameCelebrity
 from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -34,7 +35,8 @@ def codeact_user_response_eda(state: State) -> str:
 
     # retrieve the latest model message from history
     if state.history:
-        model_guess = state.history.get_last_agent_message()
+        last_agent_message = state.get_last_agent_message()
+        model_guess = last_agent_message.content if last_agent_message else ''
 
     assert game is not None, 'Game is not initialized.'
     msg = game.generate_user_response(model_guess)
@@ -139,7 +141,8 @@ def process_instance(
     if state is None:
         raise ValueError('State should not be None.')
 
-    final_message = state.history.get_last_agent_message()
+    last_agent_message = state.get_last_agent_message()
+    final_message = last_agent_message.content if last_agent_message else ''
 
     logger.info(f'Final message: {final_message} | Ground truth: {instance["text"]}')
     test_result = game.reward()
@@ -148,7 +151,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # Save the output
     output = EvalOutput(

evaluation/README.md

@@ -84,4 +84,3 @@ all the preprocessing/evaluation/analysis scripts.
 - Raw data and experimental records should not be stored within this repo.
 - For model outputs, they should be stored at [this huggingface space](https://huggingface.co/spaces/OpenHands/evaluation) for visualization.
 - Important data files of manageable size and analysis scripts (e.g., jupyter notebooks) can be directly uploaded to this repo.
-

evaluation/agent_bench/run_infer.py

@@ -16,6 +16,7 @@ from evaluation.agent_bench.helper import (
 from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -242,7 +243,7 @@ def process_instance(
         raw_ans = ''
 
         # retrieve the last agent message or thought
-        for event in state.history.get_events(reverse=True):
+        for event in reversed(state.history):
             if event.source == 'agent':
                 if isinstance(event, AgentFinishAction):
                     raw_ans = event.thought
@@ -271,7 +272,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     metrics = state.metrics.get() if state.metrics else None
 

evaluation/aider_bench/README.md

@@ -56,6 +56,20 @@ You can update the arguments in the script
 ./evaluation/aider_bench/scripts/run_infer.sh eval_gpt35_turbo HEAD CodeActAgent 100 1 "1,3,10"
 ```
 
+### Run Inference on `RemoteRuntime` (experimental)
+
+This is in limited beta. Contact Xingyao over slack if you want to try this out!
+
+```bash
+./evaluation/aider_bench/scripts/run_infer.sh [model_config] [git-version] [agent] [eval_limit] [eval-num-workers] [eval_ids]
+
+# Example - This runs evaluation on CodeActAgent for 133 instances on aider_bench test set, with 2 workers running in parallel
+export ALLHANDS_API_KEY="YOUR-API-KEY"
+export RUNTIME=remote
+export SANDBOX_REMOTE_RUNTIME_API_URL="https://runtime.eval.all-hands.dev"
+./evaluation/aider_bench/scripts/run_infer.sh llm.eval HEAD CodeActAgent 133 2
+```
+
 ## Summarize Results
 
 ```bash

evaluation/aider_bench/run_infer.py

@@ -15,6 +15,7 @@ from evaluation.aider_bench.helper import (
 from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -57,6 +58,9 @@ def get_config(
             use_host_network=False,
             timeout=100,
             api_key=os.environ.get('ALLHANDS_API_KEY', None),
+            remote_runtime_api_url=os.environ.get('SANDBOX_REMOTE_RUNTIME_API_URL'),
+            keep_runtime_alive=False,
+            remote_runtime_init_timeout=1800,
         ),
         # do not mount workspace
         workspace_base=None,
@@ -250,7 +254,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
     metrics = state.metrics.get() if state.metrics else None
 
     # Save the output

evaluation/biocoder/run_infer.py

@@ -13,6 +13,7 @@ from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
     codeact_user_response,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -39,7 +40,7 @@ AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {
 }
 
 AGENT_CLS_TO_INST_SUFFIX = {
-    'CodeActAgent': 'When you think you have fixed the issue through code changes, please run the following command: <execute_bash> exit </execute_bash>.\n'
+    'CodeActAgent': 'When you think you have fixed the issue through code changes, please finish the interaction using the "finish" tool.\n'
 }
 
 FILE_EXT_MAP = {
@@ -299,7 +300,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     test_result['generated'] = test_result['metadata']['1_copy_change_code']
 

evaluation/bird/run_infer.py

@@ -16,6 +16,7 @@ from tqdm import tqdm
 from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -39,21 +40,21 @@ from openhands.utils.async_utils import call_async_from_sync
 def codeact_user_response(state: State) -> str:
     msg = (
         'Please continue working on the task on whatever approach you think is suitable.\n'
-        'If you think you have completed the SQL, please run the following command: <execute_bash> exit </execute_bash>.\n'
+        'If you think you have completed the SQL, please finish the interaction using the "finish" tool.\n'
         'IMPORTANT: YOU SHOULD NEVER ASK FOR HUMAN HELP OR USE THE INTERNET TO SOLVE THIS TASK.\n'
     )
     if state.history:
         # check if the agent has tried to talk to the user 3 times, if so, let the agent know it can give up
         user_msgs = [
             event
-            for event in state.history.get_events()
+            for event in state.history
             if isinstance(event, MessageAction) and event.source == 'user'
         ]
         if len(user_msgs) > 2:
             # let the agent know that it can give up when it has tried 3 times
             return (
                 msg
-                + 'If you want to give up, run: <execute_bash> exit </execute_bash>.\n'
+                + 'If you want to give up, use the "finish" tool to finish the interaction.\n'
             )
     return msg
 
@@ -63,7 +64,7 @@ AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {
 }
 
 AGENT_CLS_TO_INST_SUFFIX = {
-    'CodeActAgent': 'When you think you have fixed the issue through code changes, please run the following command: <execute_bash> exit </execute_bash>.\n'
+    'CodeActAgent': 'When you think you have fixed the issue through code changes, please finish the interaction using the "finish" tool.\n'
 }
 
 
@@ -431,7 +432,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # Save the output
     output = EvalOutput(

evaluation/browsing_delegation/run_infer.py

@@ -9,6 +9,7 @@ from datasets import load_dataset
 from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -89,7 +90,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # find the last delegate action
     last_delegate_action = None

evaluation/discoverybench/run_infer.py

@@ -15,6 +15,7 @@ from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
     codeact_user_response,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -54,7 +55,7 @@ AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {
 }
 
 AGENT_CLS_TO_INST_SUFFIX = {
-    'CodeActAgent': 'When you think you have fixed the issue through code changes, please run the following command: <execute_bash> exit </execute_bash>.\n'
+    'CodeActAgent': 'When you think you have fixed the issue through code changes, please finish the interaction using the "finish" tool.\n'
 }
 
 
@@ -173,14 +174,14 @@ def initialize_runtime(runtime: Runtime, data_files: list[str]):
 
 
 def get_last_agent_finish_action(state: State) -> AgentFinishAction:
-    for event in state.history.get_events(reverse=True):
+    for event in reversed(state.history):
         if isinstance(event, AgentFinishAction):
             return event
     return None
 
 
 def get_last_message_action(state: State) -> MessageAction:
-    for event in state.history.get_events(reverse=True):
+    for event in reversed(state.history):
         if isinstance(event, MessageAction):
             return event
     return None
@@ -249,9 +250,6 @@ def process_instance(
 
     config = get_config(metadata)
 
-    # use a session id for concurrent evaluation
-    sid = 'ID_' + str(instance.instance_id)
-
     # Setup the logger properly, so you can run
     # multi-processing to parallelize the evaluation
     if reset_logger:
@@ -283,7 +281,7 @@ def process_instance(
     instruction += AGENT_CLS_TO_INST_SUFFIX[metadata.agent_class]
 
     # Here's how you can run the agent (similar to the `main` function) and get the final task state
-    runtime = create_runtime(config, sid=sid)
+    runtime = create_runtime(config)
     call_async_from_sync(runtime.connect)
     initialize_runtime(runtime, instance.data_files)
 
@@ -307,7 +305,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # DiscoveryBench Evaluation
     eval_rec = run_eval_gold_vs_gen_NL_hypo_workflow(

evaluation/gaia/run_infer.py

@@ -12,6 +12,7 @@ from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
     codeact_user_response,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -166,7 +167,7 @@ def process_instance(
 
     model_answer_raw = ''
     # get the last message or thought from the agent
-    for event in state.history.get_events(reverse=True):
+    for event in reversed(state.history):
         if event.source == 'agent':
             if isinstance(event, AgentFinishAction):
                 model_answer_raw = event.thought
@@ -203,7 +204,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # Save the output
     output = EvalOutput(

evaluation/gorilla/run_infer.py

@@ -10,6 +10,7 @@ from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
     codeact_user_response,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -32,7 +33,7 @@ AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {
 }
 
 AGENT_CLS_TO_INST_SUFFIX = {
-    'CodeActAgent': 'When you think you have completed the request, please run the following command: <execute_bash> exit </execute_bash>.\n'
+    'CodeActAgent': 'When you think you have completed the request, please finish the interaction using the "finish" tool.\n'
 }
 
 
@@ -101,7 +102,8 @@ def process_instance(
         raise ValueError('State should not be None.')
 
     # retrieve the last message from the agent
-    model_answer_raw = state.history.get_last_agent_message()
+    last_agent_message = state.get_last_agent_message()
+    model_answer_raw = last_agent_message.content if last_agent_message else ''
 
     # attempt to parse model_answer
     ast_eval_fn = instance['ast_eval']
@@ -114,7 +116,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     output = EvalOutput(
         instance_id=instance_id,

evaluation/gpqa/run_infer.py

@@ -28,6 +28,7 @@ from datasets import load_dataset
 from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -86,11 +87,10 @@ def gpqa_codeact_user_response(
     msg = (
         'Please continue working on the task on whatever approach you think is suitable.\n'
         'Feel free to use all tools for calculations and solving the problem, and web-search for finding relevant facts during the process if needed\n'
-        'If you have finished reporting the answer in the expected format, (and only once that is done), please run the following command to submit: <execute_bash> exit </execute_bash>.\n'
+        'If you have finished reporting the answer in the expected format, (and only once that is done), please use the "finish" tool to finish the interaction.\n'
         'Again you are being told a million times to first report the answer in the requested format (see again below for reference) before exiting. DO NOT EXIT WITHOUT REPORTING THE ANSWER FIRST.\n'
         'That is, when you have decided on the answer report in the following format:\n'
         f'{ACTION_FORMAT}\n'
-        '<execute_bash> exit </execute_bash>\n'
         'IMPORTANT: YOU SHOULD NEVER ASK FOR HUMAN HELP TO SOLVE THIS TASK.\n'
     )
     return msg
@@ -99,7 +99,7 @@ def gpqa_codeact_user_response(
 AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {'CodeActAgent': gpqa_codeact_user_response}
 
 AGENT_CLS_TO_INST_SUFFIX = {
-    'CodeActAgent': '\n\n SUPER IMPORTANT: When you think you have solved the question, first report it back to the user in the requested format. Only once that is done, in the next turn, please run the following command: <execute_bash> exit </execute_bash>.\n'
+    'CodeActAgent': '\n\n SUPER IMPORTANT: When you think you have solved the question, first report it back to the user in the requested format. Only once that is done, in the next turn, please finish the interaction using the "finish" tool.\n'
 }
 
 
@@ -204,12 +204,11 @@ Additional Instructions:
 - Do not try to solve the question in a single step. Break it down into smaller steps.
 - You should ONLY interact with the environment provided to you AND NEVER ASK FOR HUMAN HELP.
 
-- SUPER IMPORTANT: When you have reported the answer to the user in the requested format, (and only once that is done) in the next turn, please run the following command: <execute_bash> exit </execute_bash>.
+- SUPER IMPORTANT: When you have reported the answer to the user in the requested format, (and only once that is done) in the next turn, please finish the interaction using the "finish" tool.
 - Again you are being told a million times to first report the answer in the requested format (see again below for reference) before exiting. DO NOT EXIT WITHOUT REPORTING THE ANSWER FIRST.
     That is, when you have decided on the answer report in the following format:
 
 {ACTION_FORMAT}
-<execute_bash> exit </execute_bash>
 
 Again do not quit without reporting the answer first.
 Ok now its time to start solving the question. Good luck!
@@ -244,7 +243,7 @@ Ok now its time to start solving the question. Good luck!
         'C': False,
         'D': False,
     }
-    for event in state.history.get_events(reverse=True):
+    for event in reversed(state.history):
         if (
             isinstance(event, AgentFinishAction)
             and event.source != 'user'
@@ -300,7 +299,7 @@ Ok now its time to start solving the question. Good luck!
         instance_id=str(instance.instance_id),
         instruction=instruction,
         metadata=metadata,
-        history=state.history.compatibility_for_eval_history_pairs(),
+        history=compatibility_for_eval_history_pairs(state.history),
         metrics=metrics,
         error=state.last_error if state and state.last_error else None,
         test_result={

evaluation/humanevalfix/README.md

@@ -23,7 +23,7 @@ For each problem, OpenHands is given a set number of iterations to fix the faili
 ```
 {
     "task_id": "Python/2",
-    "instruction": "Please fix the function in Python__2.py such that all test cases pass.\nEnvironment has been set up for you to start working. You may assume all necessary tools are installed.\n\n# Problem Statement\ndef truncate_number(number: float) -> float:\n    return number % 1.0 + 1.0\n\n\n\n\n\n\ndef check(truncate_number):\n    assert truncate_number(3.5) == 0.5\n    assert abs(truncate_number(1.33) - 0.33) < 1e-6\n    assert abs(truncate_number(123.456) - 0.456) < 1e-6\n\ncheck(truncate_number)\n\nIMPORTANT: You should ONLY interact with the environment provided to you AND NEVER ASK FOR HUMAN HELP.\nYou should NOT modify any existing test case files. If needed, you can add new test cases in a NEW file to reproduce the issue.\nYou SHOULD INCLUDE PROPER INDENTATION in your edit commands.\nWhen you think you have fixed the issue through code changes, please run the following command: <execute_bash> exit </execute_bash>.\n",
+    "instruction": "Please fix the function in Python__2.py such that all test cases pass.\nEnvironment has been set up for you to start working. You may assume all necessary tools are installed.\n\n# Problem Statement\ndef truncate_number(number: float) -> float:\n    return number % 1.0 + 1.0\n\n\n\n\n\n\ndef check(truncate_number):\n    assert truncate_number(3.5) == 0.5\n    assert abs(truncate_number(1.33) - 0.33) < 1e-6\n    assert abs(truncate_number(123.456) - 0.456) < 1e-6\n\ncheck(truncate_number)\n\nIMPORTANT: You should ONLY interact with the environment provided to you AND NEVER ASK FOR HUMAN HELP.\nYou should NOT modify any existing test case files. If needed, you can add new test cases in a NEW file to reproduce the issue.\nYou SHOULD INCLUDE PROPER INDENTATION in your edit commands.\nWhen you think you have fixed the issue through code changes, please finish the interaction using the "finish" tool.\n",
     "metadata": {
         "agent_class": "CodeActAgent",
         "model_name": "gpt-4",
@@ -38,10 +38,10 @@ For each problem, OpenHands is given a set number of iterations to fix the faili
                 "id": 27,
                 "timestamp": "2024-05-22T20:57:24.688651",
                 "source": "user",
-                "message": "Please fix the function in Python__2.py such that all test cases pass.\nEnvironment has been set up for you to start working. You may assume all necessary tools are installed.\n\n# Problem Statement\ndef truncate_number(number: float) -> float:\n    return number % 1.0 + 1.0\n\n\n\n\n\n\ndef check(truncate_number):\n    assert truncate_number(3.5) == 0.5\n    assert abs(truncate_number(1.33) - 0.33) < 1e-6\n    assert abs(truncate_number(123.456) - 0.456) < 1e-6\n\ncheck(truncate_number)\n\nIMPORTANT: You should ONLY interact with the environment provided to you AND NEVER ASK FOR HUMAN HELP.\nYou should NOT modify any existing test case files. If needed, you can add new test cases in a NEW file to reproduce the issue.\nYou SHOULD INCLUDE PROPER INDENTATION in your edit commands.\nWhen you think you have fixed the issue through code changes, please run the following command: <execute_bash> exit </execute_bash>.\n",
+                "message": "Please fix the function in Python__2.py such that all test cases pass.\nEnvironment has been set up for you to start working. You may assume all necessary tools are installed.\n\n# Problem Statement\ndef truncate_number(number: float) -> float:\n    return number % 1.0 + 1.0\n\n\n\n\n\n\ndef check(truncate_number):\n    assert truncate_number(3.5) == 0.5\n    assert abs(truncate_number(1.33) - 0.33) < 1e-6\n    assert abs(truncate_number(123.456) - 0.456) < 1e-6\n\ncheck(truncate_number)\n\nIMPORTANT: You should ONLY interact with the environment provided to you AND NEVER ASK FOR HUMAN HELP.\nYou should NOT modify any existing test case files. If needed, you can add new test cases in a NEW file to reproduce the issue.\nYou SHOULD INCLUDE PROPER INDENTATION in your edit commands.\nWhen you think you have fixed the issue through code changes, please finish the interaction using the "finish" tool.\n",
                 "action": "message",
                 "args": {
-                    "content": "Please fix the function in Python__2.py such that all test cases pass.\nEnvironment has been set up for you to start working. You may assume all necessary tools are installed.\n\n# Problem Statement\ndef truncate_number(number: float) -> float:\n    return number % 1.0 + 1.0\n\n\n\n\n\n\ndef check(truncate_number):\n    assert truncate_number(3.5) == 0.5\n    assert abs(truncate_number(1.33) - 0.33) < 1e-6\n    assert abs(truncate_number(123.456) - 0.456) < 1e-6\n\ncheck(truncate_number)\n\nIMPORTANT: You should ONLY interact with the environment provided to you AND NEVER ASK FOR HUMAN HELP.\nYou should NOT modify any existing test case files. If needed, you can add new test cases in a NEW file to reproduce the issue.\nYou SHOULD INCLUDE PROPER INDENTATION in your edit commands.\nWhen you think you have fixed the issue through code changes, please run the following command: <execute_bash> exit </execute_bash>.\n",
+                    "content": "Please fix the function in Python__2.py such that all test cases pass.\nEnvironment has been set up for you to start working. You may assume all necessary tools are installed.\n\n# Problem Statement\ndef truncate_number(number: float) -> float:\n    return number % 1.0 + 1.0\n\n\n\n\n\n\ndef check(truncate_number):\n    assert truncate_number(3.5) == 0.5\n    assert abs(truncate_number(1.33) - 0.33) < 1e-6\n    assert abs(truncate_number(123.456) - 0.456) < 1e-6\n\ncheck(truncate_number)\n\nIMPORTANT: You should ONLY interact with the environment provided to you AND NEVER ASK FOR HUMAN HELP.\nYou should NOT modify any existing test case files. If needed, you can add new test cases in a NEW file to reproduce the issue.\nYou SHOULD INCLUDE PROPER INDENTATION in your edit commands.\nWhen you think you have fixed the issue through code changes, please finish the interaction using the "finish" tool.\n",
                     "wait_for_response": false
                 }
             },

evaluation/humanevalfix/run_infer.py

@@ -21,6 +21,7 @@ from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
     codeact_user_response,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -74,7 +75,7 @@ AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {
 }
 
 AGENT_CLS_TO_INST_SUFFIX = {
-    'CodeActAgent': 'When you think you have fixed the issue through code changes, please run the following command: <execute_bash> exit </execute_bash>.\n'
+    'CodeActAgent': 'When you think you have fixed the issue through code changes, please finish the interaction using the "finish" tool.\n'
 }
 
 
@@ -255,7 +256,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # Save the output
     output = EvalOutput(

evaluation/integration_tests/run_infer.py

@@ -13,6 +13,7 @@ from evaluation.utils.shared import (
     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 (
@@ -55,18 +56,14 @@ def get_config(
         workspace_base=None,
         workspace_mount_path=None,
     )
-    if metadata.llm_config.log_completions:
-        metadata.llm_config.log_completions_folder = os.path.join(
-            metadata.eval_output_dir, 'llm_completions', instance_id
+    config.set_llm_config(
+        update_llm_config_for_completions_logging(
+            metadata.llm_config, metadata.eval_output_dir, instance_id
         )
-        logger.info(
-            f'Logging LLM completions for instance {instance_id} to '
-            f'{metadata.llm_config.log_completions_folder}'
-        )
-    config.set_llm_config(metadata.llm_config)
+    )
     agent_config = AgentConfig(
         codeact_enable_jupyter=True,
-        codeact_enable_browsing_delegate=True,
+        codeact_enable_browsing=True,
         codeact_enable_llm_editor=False,
     )
     config.set_agent_config(agent_config)
@@ -132,7 +129,7 @@ def process_instance(
     # # result evaluation
     # # =============================================
 
-    histories = [event_to_dict(event) for event in state.history.get_events()]
+    histories = [event_to_dict(event) for event in state.history]
     test_result: TestResult = test_class.verify_result(runtime, histories)
     metrics = state.metrics.get() if state.metrics else None
 

evaluation/integration_tests/tests/t06_github_pr_browsing.py

@@ -0,0 +1,44 @@
+from evaluation.integration_tests.tests.base import BaseIntegrationTest, TestResult
+from openhands.events.action import AgentFinishAction, MessageAction
+from openhands.events.event import Event
+from openhands.events.observation import AgentDelegateObservation
+from openhands.runtime.base import Runtime
+
+
+class Test(BaseIntegrationTest):
+    INSTRUCTION = 'Look at https://github.com/All-Hands-AI/OpenHands/pull/8, and tell me what is happening there and what did @asadm suggest.'
+
+    @classmethod
+    def initialize_runtime(cls, runtime: Runtime) -> None:
+        pass
+
+    @classmethod
+    def verify_result(cls, runtime: Runtime, histories: list[Event]) -> TestResult:
+        # check if the "The answer is OpenHands is all you need!" is in any message
+        message_actions = [
+            event
+            for event in histories
+            if isinstance(
+                event, (MessageAction, AgentFinishAction, AgentDelegateObservation)
+            )
+        ]
+        for event in message_actions:
+            if isinstance(event, AgentDelegateObservation):
+                content = event.content
+            elif isinstance(event, AgentFinishAction):
+                content = event.outputs.get('content', '')
+            elif isinstance(event, MessageAction):
+                content = event.content
+            else:
+                raise ValueError(f'Unknown event type: {type(event)}')
+
+            if (
+                'non-commercial' in content
+                or 'MIT' in content
+                or 'Apache 2.0' in content
+            ):
+                return TestResult(success=True)
+        return TestResult(
+            success=False,
+            reason=f'The answer is not found in any message. Total messages: {len(message_actions)}. Messages: {message_actions}',
+        )

evaluation/logic_reasoning/run_infer.py

@@ -8,6 +8,7 @@ from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
     codeact_user_response,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -225,7 +226,7 @@ def process_instance(
         raise ValueError('State should not be None.')
 
     final_message = ''
-    for event in state.history.get_events(reverse=True):
+    for event in reversed(state.history):
         if isinstance(event, AgentFinishAction):
             final_message = event.thought
             break
@@ -247,7 +248,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # Save the output
     output = EvalOutput(

evaluation/miniwob/README.md

@@ -16,6 +16,20 @@ Access with browser the above MiniWoB URLs and see if they load correctly.
 ./evaluation/miniwob/scripts/run_infer.sh llm.claude-35-sonnet-eval
 ```
 
+### Run Inference on `RemoteRuntime` (experimental)
+
+This is in limited beta. Contact Xingyao over slack if you want to try this out!
+
+```bash
+./evaluation/miniwob/scripts/run_infer.sh [model_config] [git-version] [agent] [note] [eval_limit] [num_workers]
+
+# Example - This runs evaluation on BrowsingAgent for 125 instances on miniwob, with 2 workers running in parallel
+export ALLHANDS_API_KEY="YOUR-API-KEY"
+export RUNTIME=remote
+export SANDBOX_REMOTE_RUNTIME_API_URL="https://runtime.eval.all-hands.dev"
+./evaluation/miniwob/scripts/run_infer.sh llm.eval HEAD BrowsingAgent "" 125 2
+```
+
 Results will be in `evaluation/evaluation_outputs/outputs/miniwob/`
 
 To calculate the average reward, run:

evaluation/miniwob/get_avg_reward.py

@@ -23,7 +23,7 @@ if __name__ == '__main__':
             data = json.loads(line)
             actual_num += 1
             total_cost += data['metrics']['accumulated_cost']
-            total_reward += data['test_result']
+            total_reward += data['test_result']['reward']
 
     avg_reward = total_reward / total_num
     print('Avg Reward: ', avg_reward)

evaluation/miniwob/run_infer.py

@@ -10,10 +10,13 @@ import pandas as pd
 from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
+    codeact_user_response,
+    compatibility_for_eval_history_pairs,
     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 (
@@ -29,7 +32,10 @@ from openhands.events.action import (
     CmdRunAction,
     MessageAction,
 )
-from openhands.events.observation import CmdOutputObservation
+from openhands.events.observation import (
+    BrowserOutputObservation,
+    CmdOutputObservation,
+)
 from openhands.runtime.base import Runtime
 from openhands.runtime.browser.browser_env import (
     BROWSER_EVAL_GET_GOAL_ACTION,
@@ -37,7 +43,12 @@ from openhands.runtime.browser.browser_env import (
 )
 from openhands.utils.async_utils import call_async_from_sync
 
-SUPPORTED_AGENT_CLS = {'BrowsingAgent'}
+SUPPORTED_AGENT_CLS = {'BrowsingAgent', 'CodeActAgent'}
+
+AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {
+    'CodeActAgent': codeact_user_response,
+    'BrowsingAgent': 'Continue the task. IMPORTANT: do not talk to the user until you have finished the task',
+}
 
 
 def get_config(
@@ -47,25 +58,34 @@ def get_config(
     config = AppConfig(
         default_agent=metadata.agent_class,
         run_as_openhands=False,
-        runtime='eventstream',
+        runtime=os.environ.get('RUNTIME', 'eventstream'),
         max_iterations=metadata.max_iterations,
         sandbox=SandboxConfig(
             base_container_image='xingyaoww/od-eval-miniwob:v1.0',
             enable_auto_lint=True,
             use_host_network=False,
             browsergym_eval_env=env_id,
+            api_key=os.environ.get('ALLHANDS_API_KEY', None),
+            remote_runtime_api_url=os.environ.get('SANDBOX_REMOTE_RUNTIME_API_URL'),
+            remote_runtime_init_timeout=1800,
+            keep_runtime_alive=False,
+            timeout=120,
         ),
         # do not mount workspace
         workspace_base=None,
         workspace_mount_path=None,
     )
-    config.set_llm_config(metadata.llm_config)
+    config.set_llm_config(
+        update_llm_config_for_completions_logging(
+            metadata.llm_config, metadata.eval_output_dir, env_id
+        )
+    )
     return config
 
 
 def initialize_runtime(
     runtime: Runtime,
-) -> str:
+) -> tuple[str, BrowserOutputObservation]:
     """Initialize the runtime for the agent.
 
     This function is called before the runtime is used to run the agent.
@@ -85,8 +105,14 @@ def initialize_runtime(
     logger.info(obs, extra={'msg_type': 'OBSERVATION'})
     goal = obs.content
 
+    # Run noop to get the initial browser observation (e.g., the page URL & content)
+    action = BrowseInteractiveAction(browser_actions='noop(1000)')
+    logger.info(action, extra={'msg_type': 'ACTION'})
+    obs = runtime.run_action(action)
+    logger.info(obs, extra={'msg_type': 'OBSERVATION'})
+
     logger.info(f"{'-' * 50} END Runtime Initialization Fn {'-' * 50}")
-    return goal
+    return goal, obs
 
 
 def complete_runtime(
@@ -117,7 +143,7 @@ def process_instance(
     metadata: EvalMetadata,
     reset_logger: bool = True,
 ) -> EvalOutput:
-    env_id = instance.id
+    env_id = instance.instance_id
     config = get_config(metadata, env_id)
 
     # Setup the logger properly, so you can run multi-processing to parallelize the evaluation
@@ -129,7 +155,12 @@ def process_instance(
 
     runtime = create_runtime(config)
     call_async_from_sync(runtime.connect)
-    task_str = initialize_runtime(runtime)
+    task_str, obs = initialize_runtime(runtime)
+
+    task_str += (
+        f'\nInitial browser state (output of `noop(1000)`):\n{obs.get_agent_obs_text()}'
+    )
+
     state: State | None = asyncio.run(
         run_controller(
             config=config,
@@ -137,6 +168,9 @@ def process_instance(
                 content=task_str
             ),  # take output from initialize_runtime
             runtime=runtime,
+            fake_user_response_fn=AGENT_CLS_TO_FAKE_USER_RESPONSE_FN[
+                metadata.agent_class
+            ],
         )
     )
 
@@ -152,19 +186,19 @@ def process_instance(
 
     # Instruction is the first message from the USER
     instruction = ''
-    for event in state.history.get_events():
+    for event in state.history:
         if isinstance(event, MessageAction):
             instruction = event.content
             break
 
     return_val = complete_runtime(runtime)
     logger.info(f'Return value from complete_runtime: {return_val}')
-    reward = max(return_val['rewards'])
+    reward = max(return_val['rewards'], default=0)
 
     # 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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # Save the output
     output = EvalOutput(

evaluation/miniwob/scripts/run_infer.sh

@@ -33,7 +33,7 @@ echo "MODEL_CONFIG: $MODEL_CONFIG"
 
 EVAL_NOTE="${AGENT_VERSION}_${NOTE}"
 
-COMMAND="poetry run python evaluation/miniwob/run_infer.py \
+COMMAND="export PYTHONPATH=evaluation/miniwob:\$PYTHONPATH && poetry run python evaluation/miniwob/run_infer.py \
   --agent-cls $AGENT \
   --llm-config $MODEL_CONFIG \
   --max-iterations 10 \

evaluation/mint/run_infer.py

@@ -13,6 +13,7 @@ from evaluation.mint.tasks import Task
 from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -28,6 +29,7 @@ from openhands.core.config import (
 from openhands.core.logger import openhands_logger as logger
 from openhands.core.main import create_runtime, run_controller
 from openhands.events.action import (
+    Action,
     CmdRunAction,
     MessageAction,
 )
@@ -45,7 +47,10 @@ def codeact_user_response_mint(state: State, task: Task, task_config: dict[str,
         task=task,
         task_config=task_config,
     )
-    last_action = state.history.get_last_action()
+    last_action = next(
+        (event for event in reversed(state.history) if isinstance(event, Action)),
+        None,
+    )
     result_state: TaskState = env.step(last_action.message or '')
 
     state.extra_data['task_state'] = result_state
@@ -65,7 +70,7 @@ AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {
 }
 
 AGENT_CLS_TO_INST_SUFFIX = {
-    'CodeActAgent': '\nIMPORTANT: When your answer is confirmed by the user to be correct, you can exit using the following command: <execute_bash> exit </execute_bash>.\n'
+    'CodeActAgent': 'IMPORTANT: When your answer is confirmed by the user to be correct, you can use the "finish" tool to finish the interaction.\n'
 }
 
 with open(os.path.join(os.path.dirname(__file__), 'requirements.txt'), 'r') as f:
@@ -202,7 +207,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # Save the output
     output = EvalOutput(

evaluation/ml_bench/README.md

@@ -55,7 +55,7 @@ Here's an example of the evaluation output for a single task instance:
 {
   "instance_id": 3,
   "repo": "https://github.com/dmlc/dgl",
-  "instruction": "Please complete the Machine Learning task in the following repository: dgl\n\nThe task is: DGL Implementation of NGCF model\n\nI have a deep desire to embark on a journey brimming with knowledge and expertise. My objective is to train a cutting-edge NGCF Model, known for its unparalleled capabilities, on the illustrious dataset known as gowalla. To ensure swift execution, I kindly request your assistance in crafting the code, making use of the powerful GPU #3 and an embedding size of 32. Can you lend a helping hand to transform this dream into a reality?\n\nYou should create a script named `run.sh` under the specified path in the repo to run the task.\n\nYou can find the task repo at: /workspace/dgl/examples/pytorch/NGCF/NGCF\n\nYou should terminate the subprocess after running the task (e.g., call subprocess.Popen(args).wait()).When you think you have completed the task, please run the following command: <execute_bash> exit </execute_bash>.\n",
+  "instruction": "Please complete the Machine Learning task in the following repository: dgl\n\nThe task is: DGL Implementation of NGCF model\n\nI have a deep desire to embark on a journey brimming with knowledge and expertise. My objective is to train a cutting-edge NGCF Model, known for its unparalleled capabilities, on the illustrious dataset known as gowalla. To ensure swift execution, I kindly request your assistance in crafting the code, making use of the powerful GPU #3 and an embedding size of 32. Can you lend a helping hand to transform this dream into a reality?\n\nYou should create a script named `run.sh` under the specified path in the repo to run the task.\n\nYou can find the task repo at: /workspace/dgl/examples/pytorch/NGCF/NGCF\n\nYou should terminate the subprocess after running the task (e.g., call subprocess.Popen(args).wait()).When you think you have completed the task, please finish the interaction using the "finish" tool.\n",
   "metadata": {
     "agent_class": "CodeActAgent",
     "model_name": "gpt-4-1106-preview",
@@ -70,10 +70,10 @@ Here's an example of the evaluation output for a single task instance:
         "id": 0,
         "timestamp": "2024-05-26T17:40:41.060009",
         "source": "user",
-        "message": "Please complete the Machine Learning task in the following repository: dgl\n\nThe task is: DGL Implementation of NGCF model\n\nI have a deep desire to embark on a journey brimming with knowledge and expertise. My objective is to train a cutting-edge NGCF Model, known for its unparalleled capabilities, on the illustrious dataset known as gowalla. To ensure swift execution, I kindly request your assistance in crafting the code, making use of the powerful GPU #3 and an embedding size of 32. Can you lend a helping hand to transform this dream into a reality?\n\nYou should create a script named `run.sh` under the specified path in the repo to run the task.\n\nYou can find the task repo at: /workspace/dgl/examples/pytorch/NGCF/NGCF\n\nYou should terminate the subprocess after running the task (e.g., call subprocess.Popen(args).wait()).When you think you have completed the task, please run the following command: <execute_bash> exit </execute_bash>.\n",
+        "message": "Please complete the Machine Learning task in the following repository: dgl\n\nThe task is: DGL Implementation of NGCF model\n\nI have a deep desire to embark on a journey brimming with knowledge and expertise. My objective is to train a cutting-edge NGCF Model, known for its unparalleled capabilities, on the illustrious dataset known as gowalla. To ensure swift execution, I kindly request your assistance in crafting the code, making use of the powerful GPU #3 and an embedding size of 32. Can you lend a helping hand to transform this dream into a reality?\n\nYou should create a script named `run.sh` under the specified path in the repo to run the task.\n\nYou can find the task repo at: /workspace/dgl/examples/pytorch/NGCF/NGCF\n\nYou should terminate the subprocess after running the task (e.g., call subprocess.Popen(args).wait()).When you think you have completed the task, please finish the interaction using the "finish" tool.\n",
         "action": "message",
         "args": {
-          "content": "Please complete the Machine Learning task in the following repository: dgl\n\nThe task is: DGL Implementation of NGCF model\n\nI have a deep desire to embark on a journey brimming with knowledge and expertise. My objective is to train a cutting-edge NGCF Model, known for its unparalleled capabilities, on the illustrious dataset known as gowalla. To ensure swift execution, I kindly request your assistance in crafting the code, making use of the powerful GPU #3 and an embedding size of 32. Can you lend a helping hand to transform this dream into a reality?\n\nYou should create a script named `run.sh` under the specified path in the repo to run the task.\n\nYou can find the task repo at: /workspace/dgl/examples/pytorch/NGCF/NGCF\n\nYou should terminate the subprocess after running the task (e.g., call subprocess.Popen(args).wait()).When you think you have completed the task, please run the following command: <execute_bash> exit </execute_bash>.\n",
+          "content": "Please complete the Machine Learning task in the following repository: dgl\n\nThe task is: DGL Implementation of NGCF model\n\nI have a deep desire to embark on a journey brimming with knowledge and expertise. My objective is to train a cutting-edge NGCF Model, known for its unparalleled capabilities, on the illustrious dataset known as gowalla. To ensure swift execution, I kindly request your assistance in crafting the code, making use of the powerful GPU #3 and an embedding size of 32. Can you lend a helping hand to transform this dream into a reality?\n\nYou should create a script named `run.sh` under the specified path in the repo to run the task.\n\nYou can find the task repo at: /workspace/dgl/examples/pytorch/NGCF/NGCF\n\nYou should terminate the subprocess after running the task (e.g., call subprocess.Popen(args).wait()).When you think you have completed the task, please finish the interaction using the "finish" tool.\n",
           "wait_for_response": false
         }
       },

evaluation/ml_bench/run_infer.py

@@ -24,6 +24,7 @@ from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
     codeact_user_response,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -51,7 +52,7 @@ AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {
 }
 
 AGENT_CLS_TO_INST_SUFFIX = {
-    'CodeActAgent': 'When you think you have completed the task, please run the following command: <execute_bash> exit </execute_bash>.\n'
+    'CodeActAgent': 'When you think you have completed the task, please finish the interaction using the "finish" tool.\n'
 }
 
 ID2CONDA = {
@@ -256,7 +257,7 @@ def process_instance(instance: Any, metadata: EvalMetadata, reset_logger: bool =
     # 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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # Save the output
     output = EvalOutput(

evaluation/scienceagentbench/run_infer.py

@@ -10,10 +10,12 @@ from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
     codeact_user_response,
+    compatibility_for_eval_history_pairs,
     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 (
@@ -70,21 +72,19 @@ def get_config(
             timeout=300,
             api_key=os.environ.get('ALLHANDS_API_KEY', None),
             remote_runtime_api_url=os.environ.get('SANDBOX_REMOTE_RUNTIME_API_URL'),
-            keep_remote_runtime_alive=False,
+            keep_runtime_alive=False,
         ),
         # do not mount workspace
         workspace_base=None,
         workspace_mount_path=None,
     )
-    config.set_llm_config(metadata.llm_config)
-    if metadata.llm_config.log_completions:
-        metadata.llm_config.log_completions_folder = os.path.join(
-            metadata.eval_output_dir, 'llm_completions', instance_id
-        )
-        logger.info(
-            f'Logging LLM completions for instance {instance_id} to '
-            f'{metadata.llm_config.log_completions_folder}'
+    config.set_llm_config(
+        update_llm_config_for_completions_logging(
+            metadata.llm_config,
+            metadata.eval_output_dir,
+            instance_id,
         )
+    )
     return config
 
 
@@ -233,7 +233,7 @@ If the program uses some packages that are incompatible, please figure out alter
     # 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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # Save the output
     output = EvalOutput(

evaluation/swe_bench/eval_infer.py

@@ -1,6 +1,7 @@
 import os
 import tempfile
 import time
+from functools import partial
 
 import pandas as pd
 from swebench.harness.grading import get_eval_report
@@ -83,6 +84,7 @@ def get_config(instance: pd.Series) -> AppConfig:
             timeout=1800,
             api_key=os.environ.get('ALLHANDS_API_KEY', None),
             remote_runtime_api_url=os.environ.get('SANDBOX_REMOTE_RUNTIME_API_URL'),
+            remote_runtime_init_timeout=3600,
         ),
         # do not mount workspace
         workspace_base=None,
@@ -93,13 +95,28 @@ def get_config(instance: pd.Series) -> AppConfig:
 
 def process_instance(
     instance: pd.Series,
-    metadata: EvalMetadata | None = None,
+    metadata: EvalMetadata,
     reset_logger: bool = True,
+    log_dir: str | None = None,
 ) -> 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:
-        global output_file
-        log_dir = output_file.replace('.jsonl', '.logs')
+        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:
@@ -126,6 +143,7 @@ def process_instance(
         return EvalOutput(
             instance_id=instance_id,
             test_result=instance['test_result'],
+            metadata=metadata,
         )
 
     runtime = create_runtime(config)
@@ -175,6 +193,7 @@ def process_instance(
             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}')
@@ -244,23 +263,29 @@ def process_instance(
                         test_output_path = os.path.join(log_dir, 'test_output.txt')
                         with open(test_output_path, 'w') as f:
                             f.write(test_output)
-
-                        _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'
-                        ]
+                        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
@@ -268,6 +293,7 @@ def process_instance(
             return EvalOutput(
                 instance_id=instance_id,
                 test_result=instance['test_result'],
+                metadata=metadata,
             )
         else:
             logger.info(
@@ -335,7 +361,7 @@ if __name__ == '__main__':
 
     if 'model_patch' not in predictions.columns:
         predictions['model_patch'] = predictions['test_result'].apply(
-            lambda x: x['git_patch']
+            lambda x: x.get('git_patch', '')
         )
     assert {'instance_id', 'model_patch'}.issubset(
         set(predictions.columns)
@@ -354,12 +380,26 @@ if __name__ == '__main__':
     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)
+
+    # 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=None,
+        metadata=metadata,
         output_file=output_file,
         num_workers=args.eval_num_workers,
-        process_instance_func=process_instance,
+        process_instance_func=process_instance_func,
     )
 
     # Load evaluated predictions & print number of resolved predictions

evaluation/swe_bench/prompt.py

@@ -1,6 +1,6 @@
 CODEACT_SWE_PROMPT = """Now, you're going to solve this issue on your own. Your terminal session has started and you're in the repository's root directory. You can use any bash commands or the special interface to help you. Edit all the files you need to and run any checks or tests that you want.
 Remember, YOU CAN ONLY ENTER ONE COMMAND AT A TIME. You should always wait for feedback after every command.
-When you're satisfied with all of the changes you've made, you can run the following command: <execute_bash> exit </execute_bash>.
+When you're satisfied with all of the changes you've made, you can use the "finish" tool to finish the interaction.
 Note however that you cannot use any interactive session commands (e.g. vim) in this environment, but you can write scripts and run them. E.g. you can write a python script and then run it with `python <script_name>.py`.
 
 NOTE ABOUT THE EDIT COMMAND: Indentation really matters! When editing a file, make sure to insert appropriate indentation before each line!

evaluation/swe_bench/run_infer.py

@@ -20,6 +20,7 @@ from evaluation.utils.shared import (
     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 (
@@ -35,11 +36,12 @@ from openhands.events.action import CmdRunAction, 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.runtime.utils.shutdown_listener import sleep_if_should_continue
 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', 'false').lower() == 'true'
+RUN_WITH_BROWSING = os.environ.get('RUN_WITH_BROWSING', 'false').lower() == 'true'
 
 AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {
     'CodeActAgent': codeact_user_response,
@@ -79,7 +81,7 @@ def get_instruction(instance: pd.Series, metadata: EvalMetadata):
             '</pr_description>\n\n'
             'Can you help me implement the necessary changes to the repository so that the requirements specified in the <pr_description> are met?\n'
             "I've already taken care of all changes to any of the test files described in the <pr_description>. This means you DON'T have to modify the testing logic or any of the tests in any way!\n"
-            'Your task is to make the minimal changes to non-tests files in the /repo directory to ensure the <pr_description> is satisfied.\n'
+            'Your task is to make the minimal changes to non-tests files in the /workspace directory to ensure the <pr_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'
@@ -88,6 +90,13 @@ def get_instruction(instance: pd.Series, metadata: EvalMetadata):
             '5. Think about edgecases and make sure your fix handles them as well\n'
             "Your thinking should be thorough and so it's fine if it's very long.\n"
         )
+
+    if RUN_WITH_BROWSING:
+        instruction += (
+            '<IMPORTANT!>\n'
+            'You SHOULD NEVER attempt to browse the web. '
+            '</IMPORTANT!>\n'
+        )
     return instruction
 
 
@@ -136,24 +145,21 @@ def get_config(
             platform='linux/amd64',
             api_key=os.environ.get('ALLHANDS_API_KEY', None),
             remote_runtime_api_url=os.environ.get('SANDBOX_REMOTE_RUNTIME_API_URL'),
-            keep_remote_runtime_alive=False,
+            keep_runtime_alive=False,
+            remote_runtime_init_timeout=3600,
         ),
         # do not mount workspace
         workspace_base=None,
         workspace_mount_path=None,
     )
-    if metadata.llm_config.log_completions:
-        metadata.llm_config.log_completions_folder = os.path.join(
-            metadata.eval_output_dir, 'llm_completions', instance['instance_id']
+    config.set_llm_config(
+        update_llm_config_for_completions_logging(
+            metadata.llm_config, metadata.eval_output_dir, instance['instance_id']
         )
-        logger.info(
-            f'Logging LLM completions for instance {instance["instance_id"]} to '
-            f'{metadata.llm_config.log_completions_folder}'
-        )
-    config.set_llm_config(metadata.llm_config)
+    )
     agent_config = AgentConfig(
         codeact_enable_jupyter=False,
-        codeact_enable_browsing_delegate=False,
+        codeact_enable_browsing=RUN_WITH_BROWSING,
         codeact_enable_llm_editor=False,
     )
     config.set_agent_config(agent_config)
@@ -438,7 +444,8 @@ def process_instance(
     if state is None:
         raise ValueError('State should not be None.')
 
-    histories = [event_to_dict(event) for event in state.history.get_events()]
+    # 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 = state.metrics.get() if state.metrics else None
 
     # Save the output
@@ -527,5 +534,10 @@ if __name__ == '__main__':
             instances[col] = instances[col].apply(lambda x: str(x))
 
     run_evaluation(
-        instances, metadata, output_file, args.eval_num_workers, process_instance
+        instances,
+        metadata,
+        output_file,
+        args.eval_num_workers,
+        process_instance,
+        timeout_seconds=120 * 60,  # 2 hour PER instance should be more than enough
     )

evaluation/swe_bench/scripts/run_infer.sh

@@ -34,6 +34,11 @@ if [ -z "$USE_INSTANCE_IMAGE" ]; then
   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"
@@ -47,6 +52,8 @@ fi
 
 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"
 
 get_agent_version
 
@@ -67,6 +74,10 @@ 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

evaluation/toolqa/run_infer.py

@@ -9,6 +9,7 @@ from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
     codeact_user_response,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -33,7 +34,7 @@ AGENT_CLS_TO_FAKE_USER_RESPONSE_FN = {
 }
 
 AGENT_CLS_TO_INST_SUFFIX = {
-    'CodeActAgent': 'When you think you have completed the request, please run the following command: <execute_bash> exit </execute_bash>.\n'
+    'CodeActAgent': 'When you think you have completed the request, please finish the interaction using the "finish" tool.\n'
 }
 
 
@@ -126,7 +127,8 @@ def process_instance(instance: Any, metadata: EvalMetadata, reset_logger: bool =
         raise ValueError('State should not be None.')
 
     # retrieve the last message from the agent
-    model_answer_raw = state.history.get_last_agent_message()
+    last_agent_message = state.get_last_agent_message()
+    model_answer_raw = last_agent_message.content if last_agent_message else ''
 
     # attempt to parse model_answer
     correct = eval_answer(str(model_answer_raw), str(answer))
@@ -137,7 +139,7 @@ def process_instance(instance: Any, metadata: EvalMetadata, reset_logger: bool =
     # 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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # Save the output
     output = EvalOutput(

evaluation/utils/shared.py

@@ -3,9 +3,11 @@ import logging
 import multiprocessing as mp
 import os
 import pathlib
+import signal
 import subprocess
 import time
 import traceback
+from contextlib import contextmanager
 from typing import Any, Awaitable, Callable, TextIO
 
 import pandas as pd
@@ -18,6 +20,9 @@ from openhands.core.logger import get_console_handler
 from openhands.core.logger import openhands_logger as logger
 from openhands.events.action import Action
 from openhands.events.action.message import MessageAction
+from openhands.events.event import Event
+from openhands.events.serialization.event import event_to_dict
+from openhands.events.utils import get_pairs_from_events
 
 
 class EvalMetadata(BaseModel):
@@ -89,6 +94,27 @@ class EvalException(Exception):
     pass
 
 
+class EvalTimeoutException(Exception):
+    pass
+
+
+@contextmanager
+def timeout(seconds: int):
+    def timeout_handler(signum, frame):
+        raise EvalTimeoutException(f'Function timed out after {seconds} seconds')
+
+    # Set up the signal handler
+    original_handler = signal.signal(signal.SIGALRM, timeout_handler)
+    signal.alarm(seconds)
+
+    try:
+        yield
+    finally:
+        # Restore the original handler and disable the alarm
+        signal.alarm(0)
+        signal.signal(signal.SIGALRM, original_handler)
+
+
 def codeact_user_response(
     state: State,
     encapsulate_solution: bool = False,
@@ -112,7 +138,14 @@ def codeact_user_response(
     if state.history:
         # check if the last action has an answer, if so, early exit
         if try_parse is not None:
-            last_action = state.history.get_last_action()
+            last_action = next(
+                (
+                    event
+                    for event in reversed(state.history)
+                    if isinstance(event, Action)
+                ),
+                None,
+            )
             ans = try_parse(last_action)
             if ans is not None:
                 return '/exit'
@@ -120,14 +153,14 @@ def codeact_user_response(
         # check if the agent has tried to talk to the user 3 times, if so, let the agent know it can give up
         user_msgs = [
             event
-            for event in state.history.get_events()
+            for event in state.history
             if isinstance(event, MessageAction) and event.source == 'user'
         ]
         if len(user_msgs) >= 2:
             # let the agent know that it can give up when it has tried 3 times
             return (
                 msg
-                + 'If you want to give up, run: <execute_bash> exit </execute_bash>.\n'
+                + 'If you want to give up, use the "finish" tool to finish the interaction.\n'
             )
     return msg
 
@@ -270,15 +303,33 @@ def _process_instance_wrapper(
     metadata: EvalMetadata,
     use_mp: bool,
     max_retries: int = 5,
+    timeout_seconds: int | None = None,
 ) -> EvalOutput:
-    """Wrap the process_instance_func to handle retries and errors.
-
-    Retry an instance up to max_retries times if it fails (e.g., due to transient network/runtime issues).
-    """
+    """Wrap the process_instance_func to handle retries and errors."""
     for attempt in range(max_retries + 1):
         try:
-            result = process_instance_func(instance, metadata, use_mp)
+            if timeout_seconds is not None:
+                with timeout(timeout_seconds):
+                    result = process_instance_func(instance, metadata, use_mp)
+            else:
+                result = process_instance_func(instance, metadata, use_mp)
             return result
+        except EvalTimeoutException as e:
+            error = f'Timeout after {timeout_seconds} seconds'
+            stacktrace = traceback.format_exc()
+            msg = (
+                '-' * 10
+                + '\n'
+                + f'Timeout ({timeout_seconds} seconds) in instance [{instance.instance_id}], Stopped evaluation for this instance.'
+                + '\n'
+                + '-' * 10
+            )
+            logger.exception(e)
+            return EvalOutput(
+                instance_id=instance.instance_id,
+                test_result={},
+                error=error,
+            )
         except Exception as e:
             error = str(e)
             stacktrace = traceback.format_exc()
@@ -327,6 +378,7 @@ def run_evaluation(
         [pd.Series, EvalMetadata, bool], Awaitable[EvalOutput]
     ],
     max_retries: int = 5,  # number of retries for each instance
+    timeout_seconds: int | None = None,
 ):
     use_multiprocessing = num_workers > 1
 
@@ -336,6 +388,7 @@ def run_evaluation(
             f'model {metadata.llm_config.model}, max iterations {metadata.max_iterations}.\n'
         )
     else:
+        logger.warning('Running evaluation without metadata.')
         logger.info(f'Evaluation started with {num_workers} workers.')
 
     total_instances = len(dataset)
@@ -346,7 +399,14 @@ def run_evaluation(
         if use_multiprocessing:
             with mp.Pool(num_workers) as pool:
                 args_iter = (
-                    (process_instance_func, instance, metadata, True, max_retries)
+                    (
+                        process_instance_func,
+                        instance,
+                        metadata,
+                        True,
+                        max_retries,
+                        timeout_seconds,
+                    )
                     for _, instance in dataset.iterrows()
                 )
                 results = pool.imap_unordered(_process_instance_wrapper_mp, args_iter)
@@ -411,3 +471,35 @@ def reset_logger_for_multiprocessing(
     )
     file_handler.setLevel(logging.INFO)
     logger.addHandler(file_handler)
+
+
+def update_llm_config_for_completions_logging(
+    llm_config: LLMConfig,
+    eval_output_dir: str,
+    instance_id: str,
+) -> LLMConfig:
+    """Update the LLM config for logging completions."""
+    if llm_config.log_completions:
+        llm_config.log_completions_folder = os.path.join(
+            eval_output_dir, 'llm_completions', instance_id
+        )
+        logger.info(
+            f'Logging LLM completions for instance {instance_id} to '
+            f'{llm_config.log_completions_folder}'
+        )
+    return llm_config
+
+
+# history is now available as a filtered stream of events, rather than list of pairs of (Action, Observation)
+# we rebuild the pairs here
+# for compatibility with the existing output format in evaluations
+# remove this when it's no longer necessary
+def compatibility_for_eval_history_pairs(
+    history: list[Event],
+) -> list[tuple[dict, dict]]:
+    history_pairs = []
+
+    for action, observation in get_pairs_from_events(history):
+        history_pairs.append((event_to_dict(action), event_to_dict(observation)))
+
+    return history_pairs

evaluation/webarena/run_infer.py

@@ -10,6 +10,7 @@ import pandas as pd
 from evaluation.utils.shared import (
     EvalMetadata,
     EvalOutput,
+    compatibility_for_eval_history_pairs,
     make_metadata,
     prepare_dataset,
     reset_logger_for_multiprocessing,
@@ -166,7 +167,7 @@ def process_instance(
 
     # Instruction is the first message from the USER
     instruction = ''
-    for event in state.history.get_events():
+    for event in state.history:
         if isinstance(event, MessageAction):
             instruction = event.content
             break
@@ -178,7 +179,7 @@ def process_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 = state.history.compatibility_for_eval_history_pairs()
+    histories = compatibility_for_eval_history_pairs(state.history)
 
     # Save the output
     output = EvalOutput(

frontend/.eslintrc

@@ -84,4 +84,4 @@
       }
     }
   ]
-}
+}

frontend/.gitignore

@@ -1,4 +1,9 @@
 # i18n translation files make by script using `make build`
 public/locales/**/*
 src/i18n/declaration.ts
-.env
+.env
+node_modules/
+/test-results/
+/playwright-report/
+/blob-report/
+/playwright/.cache/

frontend/__tests__/clear-session.test.ts

@@ -0,0 +1,40 @@
+import { describe, it, expect, beforeEach, vi } from "vitest";
+import { clearSession } from "../src/utils/clear-session";
+import store from "../src/store";
+import { initialState as browserInitialState } from "../src/state/browserSlice";
+
+describe("clearSession", () => {
+  beforeEach(() => {
+    // Mock localStorage
+    const localStorageMock = {
+      getItem: vi.fn(),
+      setItem: vi.fn(),
+      removeItem: vi.fn(),
+      clear: vi.fn(),
+    };
+    vi.stubGlobal("localStorage", localStorageMock);
+
+    // Set initial browser state to non-default values
+    store.dispatch({
+      type: "browser/setUrl",
+      payload: "https://example.com",
+    });
+    store.dispatch({
+      type: "browser/setScreenshotSrc",
+      payload: "base64screenshot",
+    });
+  });
+
+  it("should clear localStorage and reset browser state", () => {
+    clearSession();
+
+    // Verify localStorage items were removed
+    expect(localStorage.removeItem).toHaveBeenCalledWith("token");
+    expect(localStorage.removeItem).toHaveBeenCalledWith("repo");
+
+    // Verify browser state was reset
+    const state = store.getState();
+    expect(state.browser.url).toBe(browserInitialState.url);
+    expect(state.browser.screenshotSrc).toBe(browserInitialState.screenshotSrc);
+  });
+});

frontend/__tests__/components/chat/chat-input.test.tsx

@@ -1,5 +1,5 @@
 import userEvent from "@testing-library/user-event";
-import { render, screen } from "@testing-library/react";
+import { fireEvent, render, screen } from "@testing-library/react";
 import { describe, afterEach, vi, it, expect } from "vitest";
 import { ChatInput } from "#/components/chat-input";
 
@@ -158,4 +158,46 @@ describe("ChatInput", () => {
     await user.tab();
     expect(onBlurMock).toHaveBeenCalledOnce();
   });
+
+  it("should handle text paste correctly", () => {
+    const onSubmit = vi.fn();
+    const onChange = vi.fn();
+
+    render(<ChatInput onSubmit={onSubmit} onChange={onChange} />);
+
+    const input = screen.getByTestId("chat-input").querySelector("textarea");
+    expect(input).toBeTruthy();
+
+    // Fire paste event with text data
+    fireEvent.paste(input!, {
+      clipboardData: {
+        getData: (type: string) => type === 'text/plain' ? 'test paste' : '',
+        files: []
+      }
+    });
+  });
+
+  it("should handle image paste correctly", () => {
+    const onSubmit = vi.fn();
+    const onImagePaste = vi.fn();
+
+    render(<ChatInput onSubmit={onSubmit} onImagePaste={onImagePaste} />);
+
+    const input = screen.getByTestId("chat-input").querySelector("textarea");
+    expect(input).toBeTruthy();
+
+    // Create a paste event with an image file
+    const file = new File(["dummy content"], "image.png", { type: "image/png" });
+
+    // Fire paste event with image data
+    fireEvent.paste(input!, {
+      clipboardData: {
+        getData: () => '',
+        files: [file]
+      }
+    });
+
+    // Verify image paste was handled
+    expect(onImagePaste).toHaveBeenCalledWith([file]);
+  });
 });

frontend/__tests__/components/chat/chat-interface.test.tsx

@@ -1,19 +1,161 @@
-import { afterEach, describe, expect, it, vi } from "vitest";
-import { render, screen, within } from "@testing-library/react";
+import { afterEach, beforeAll, describe, expect, it, vi } from "vitest";
+import { act, screen, waitFor, within } from "@testing-library/react";
 import userEvent from "@testing-library/user-event";
+import { renderWithProviders } from "test-utils";
 import { ChatInterface } from "#/components/chat-interface";
-import { SocketProvider } from "#/context/socket";
+import { addUserMessage } from "#/state/chatSlice";
+import { SUGGESTIONS } from "#/utils/suggestions";
+import * as ChatSlice from "#/state/chatSlice";
 
 // eslint-disable-next-line @typescript-eslint/no-unused-vars
 const renderChatInterface = (messages: (Message | ErrorMessage)[]) =>
-  render(<ChatInterface />, { wrapper: SocketProvider });
+  renderWithProviders(<ChatInterface />);
+
+describe("Empty state", () => {
+  const { send: sendMock } = vi.hoisted(() => ({
+    send: vi.fn(),
+  }));
+
+  const { useWsClient: useWsClientMock } = vi.hoisted(() => ({
+    useWsClient: vi.fn(() => ({ send: sendMock, runtimeActive: true })),
+  }));
+
+  beforeAll(() => {
+    vi.mock("@remix-run/react", async (importActual) => ({
+      ...(await importActual<typeof import("@remix-run/react")>()),
+      useRouteLoaderData: vi.fn(() => ({})),
+    }));
+
+    vi.mock("#/context/socket", async (importActual) => ({
+      ...(await importActual<typeof import("#/context/ws-client-provider")>()),
+      useWsClient: useWsClientMock,
+    }));
+  });
 
-describe.skip("ChatInterface", () => {
   afterEach(() => {
     vi.clearAllMocks();
   });
 
-  it.todo("should render suggestions if empty");
+  it("should render suggestions if empty", () => {
+    const { store } = renderWithProviders(<ChatInterface />, {
+      preloadedState: {
+        chat: { messages: [] },
+      },
+    });
+
+    expect(screen.getByTestId("suggestions")).toBeInTheDocument();
+
+    act(() => {
+      store.dispatch(
+        addUserMessage({
+          content: "Hello",
+          imageUrls: [],
+          timestamp: new Date().toISOString(),
+        }),
+      );
+    });
+
+    expect(screen.queryByTestId("suggestions")).not.toBeInTheDocument();
+  });
+
+  it("should render the default suggestions", () => {
+    renderWithProviders(<ChatInterface />, {
+      preloadedState: {
+        chat: { messages: [] },
+      },
+    });
+
+    const suggestions = screen.getByTestId("suggestions");
+    const repoSuggestions = Object.keys(SUGGESTIONS.repo);
+
+    // check that there are at most 4 suggestions displayed
+    const displayedSuggestions = within(suggestions).getAllByRole("button");
+    expect(displayedSuggestions.length).toBeLessThanOrEqual(4);
+
+    // Check that each displayed suggestion is one of the repo suggestions
+    displayedSuggestions.forEach((suggestion) => {
+      expect(repoSuggestions).toContain(suggestion.textContent);
+    });
+  });
+
+  it.fails(
+    "should load the a user message to the input when selecting",
+    async () => {
+      // this is to test that the message is in the UI before the socket is called
+      useWsClientMock.mockImplementation(() => ({
+        send: sendMock,
+        runtimeActive: false, // mock an inactive runtime setup
+      }));
+      const addUserMessageSpy = vi.spyOn(ChatSlice, "addUserMessage");
+      const user = userEvent.setup();
+      const { store } = renderWithProviders(<ChatInterface />, {
+        preloadedState: {
+          chat: { messages: [] },
+        },
+      });
+
+      const suggestions = screen.getByTestId("suggestions");
+      const displayedSuggestions = within(suggestions).getAllByRole("button");
+      const input = screen.getByTestId("chat-input");
+
+      await user.click(displayedSuggestions[0]);
+
+      // user message loaded to input
+      expect(addUserMessageSpy).not.toHaveBeenCalled();
+      expect(screen.queryByTestId("suggestions")).toBeInTheDocument();
+      expect(store.getState().chat.messages).toHaveLength(0);
+      expect(input).toHaveValue(displayedSuggestions[0].textContent);
+    },
+  );
+
+  it.fails(
+    "should send the message to the socket only if the runtime is active",
+    async () => {
+      useWsClientMock.mockImplementation(() => ({
+        send: sendMock,
+        runtimeActive: false, // mock an inactive runtime setup
+      }));
+      const user = userEvent.setup();
+      const { rerender } = renderWithProviders(<ChatInterface />, {
+        preloadedState: {
+          chat: { messages: [] },
+        },
+      });
+
+      const suggestions = screen.getByTestId("suggestions");
+      const displayedSuggestions = within(suggestions).getAllByRole("button");
+
+      await user.click(displayedSuggestions[0]);
+      expect(sendMock).not.toHaveBeenCalled();
+
+      useWsClientMock.mockImplementation(() => ({
+        send: sendMock,
+        runtimeActive: true, // mock an active runtime setup
+      }));
+      rerender(<ChatInterface />);
+
+      await waitFor(() =>
+        expect(sendMock).toHaveBeenCalledWith(expect.any(String)),
+      );
+    },
+  );
+});
+
+describe.skip("ChatInterface", () => {
+  beforeAll(() => {
+    // mock useScrollToBottom hook
+    vi.mock("#/hooks/useScrollToBottom", () => ({
+      useScrollToBottom: vi.fn(() => ({
+        scrollDomToBottom: vi.fn(),
+        onChatBodyScroll: vi.fn(),
+        hitBottom: vi.fn(),
+      })),
+    }));
+  });
+
+  afterEach(() => {
+    vi.clearAllMocks();
+  });
 
   it("should render messages", () => {
     const messages: Message[] = [
@@ -128,14 +270,14 @@ describe.skip("ChatInterface", () => {
         timestamp: new Date().toISOString(),
       },
       {
-        error: "Woops!",
+        error: true,
+        id: "",
         message: "Something went wrong",
       },
     ];
     renderChatInterface(messages);
 
     const error = screen.getByTestId("error-message");
-    expect(within(error).getByText("Woops!")).toBeInTheDocument();
     expect(within(error).getByText("Something went wrong")).toBeInTheDocument();
   });
 

frontend/__tests__/components/interactive-chat-box.test.tsx

@@ -25,6 +25,21 @@ describe("InteractiveChatBox", () => {
     within(chatBox).getByTestId("upload-image-input");
   });
 
+  it.fails("should set custom values", () => {
+    render(
+      <InteractiveChatBox
+        onSubmit={onSubmitMock}
+        onStop={onStopMock}
+        value="Hello, world!"
+      />,
+    );
+
+    const chatBox = screen.getByTestId("interactive-chat-box");
+    const chatInput = within(chatBox).getByTestId("chat-input");
+
+    expect(chatInput).toHaveValue("Hello, world!");
+  });
+
   it("should display the image previews when images are uploaded", async () => {
     const user = userEvent.setup();
     render(<InteractiveChatBox onSubmit={onSubmitMock} onStop={onStopMock} />);

frontend/__tests__/components/suggestion-item.test.tsx

@@ -0,0 +1,30 @@
+import { render, screen } from "@testing-library/react";
+import userEvent from "@testing-library/user-event";
+import { afterEach, describe, expect, it, vi } from "vitest";
+import { SuggestionItem } from "#/components/suggestion-item";
+
+describe("SuggestionItem", () => {
+  const suggestionItem = { label: "suggestion1", value: "a long text value" };
+  const onClick = vi.fn();
+
+  afterEach(() => {
+    vi.clearAllMocks();
+  });
+
+  it("should render a suggestion", () => {
+    render(<SuggestionItem suggestion={suggestionItem} onClick={onClick} />);
+
+    expect(screen.getByTestId("suggestion")).toBeInTheDocument();
+    expect(screen.getByText(/suggestion1/i)).toBeInTheDocument();
+  });
+
+  it("should call onClick when clicking a suggestion", async () => {
+    const user = userEvent.setup();
+    render(<SuggestionItem suggestion={suggestionItem} onClick={onClick} />);
+
+    const suggestion = screen.getByTestId("suggestion");
+    await user.click(suggestion);
+
+    expect(onClick).toHaveBeenCalledWith("a long text value");
+  });
+});

frontend/__tests__/components/suggestions.test.tsx

@@ -0,0 +1,60 @@
+import { render, screen } from "@testing-library/react";
+import userEvent from "@testing-library/user-event";
+import { afterEach, describe, expect, it, vi } from "vitest";
+import { Suggestions } from "#/components/suggestions";
+
+describe("Suggestions", () => {
+  const firstSuggestion = {
+    label: "first-suggestion",
+    value: "value-of-first-suggestion",
+  };
+  const secondSuggestion = {
+    label: "second-suggestion",
+    value: "value-of-second-suggestion",
+  };
+  const suggestions = [firstSuggestion, secondSuggestion];
+
+  const onSuggestionClickMock = vi.fn();
+
+  afterEach(() => {
+    vi.clearAllMocks();
+  });
+
+  it("should render suggestions", () => {
+    render(
+      <Suggestions
+        suggestions={suggestions}
+        onSuggestionClick={onSuggestionClickMock}
+      />,
+    );
+
+    expect(screen.getByTestId("suggestions")).toBeInTheDocument();
+    const suggestionElements = screen.getAllByTestId("suggestion");
+
+    expect(suggestionElements).toHaveLength(2);
+    expect(suggestionElements[0]).toHaveTextContent("first-suggestion");
+    expect(suggestionElements[1]).toHaveTextContent("second-suggestion");
+  });
+
+  it("should call onSuggestionClick when clicking a suggestion", async () => {
+    const user = userEvent.setup();
+    render(
+      <Suggestions
+        suggestions={suggestions}
+        onSuggestionClick={onSuggestionClickMock}
+      />,
+    );
+
+    const suggestionElements = screen.getAllByTestId("suggestion");
+
+    await user.click(suggestionElements[0]);
+    expect(onSuggestionClickMock).toHaveBeenCalledWith(
+      "value-of-first-suggestion",
+    );
+
+    await user.click(suggestionElements[1]);
+    expect(onSuggestionClickMock).toHaveBeenCalledWith(
+      "value-of-second-suggestion",
+    );
+  });
+});

frontend/__tests__/hooks/use-rate.test.ts

@@ -0,0 +1,93 @@
+import { act, renderHook } from "@testing-library/react";
+import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
+import { useRate } from "#/utils/use-rate";
+
+describe("useRate", () => {
+  beforeEach(() => {
+    vi.useFakeTimers();
+  });
+
+  afterEach(() => {
+    vi.useRealTimers();
+  });
+
+  it("should initialize", () => {
+    const { result } = renderHook(() => useRate());
+
+    expect(result.current.items).toHaveLength(0);
+    expect(result.current.rate).toBeNull();
+    expect(result.current.lastUpdated).toBeNull();
+    expect(result.current.isUnderThreshold).toBe(true);
+  });
+
+  it("should handle the case of a single element", () => {
+    const { result } = renderHook(() => useRate());
+
+    act(() => {
+      result.current.record(123);
+    });
+
+    expect(result.current.items).toHaveLength(1);
+    expect(result.current.lastUpdated).not.toBeNull();
+  });
+
+  it("should return the difference between the last two elements", () => {
+    const { result } = renderHook(() => useRate());
+
+    vi.setSystemTime(500);
+    act(() => {
+      result.current.record(4);
+    });
+
+    vi.advanceTimersByTime(500);
+    act(() => {
+      result.current.record(9);
+    });
+
+    expect(result.current.items).toHaveLength(2);
+    expect(result.current.rate).toBe(5);
+    expect(result.current.lastUpdated).toBe(1000);
+  });
+
+  it("should update isUnderThreshold after [threshold]ms of no activity", () => {
+    const { result } = renderHook(() => useRate({ threshold: 500 }));
+
+    expect(result.current.isUnderThreshold).toBe(true);
+
+    act(() => {
+       // not sure if fake timers is buggy with intervals,
+       // but I need to call it twice to register
+      vi.advanceTimersToNextTimer();
+      vi.advanceTimersToNextTimer();
+    });
+
+    expect(result.current.isUnderThreshold).toBe(false);
+  });
+
+  it("should return an isUnderThreshold boolean", () => {
+    const { result } = renderHook(() => useRate({ threshold: 500 }));
+
+    vi.setSystemTime(500);
+    act(() => {
+      result.current.record(400);
+    });
+    act(() => {
+      result.current.record(1000);
+    });
+
+    expect(result.current.isUnderThreshold).toBe(false);
+
+    act(() => {
+      result.current.record(1500);
+    });
+
+    expect(result.current.isUnderThreshold).toBe(true);
+
+    act(() => {
+      vi.advanceTimersToNextTimer();
+      vi.advanceTimersToNextTimer();
+    });
+
+    expect(result.current.isUnderThreshold).toBe(false);
+  });
+});

frontend/__tests__/hooks/use-terminal.test.tsx

@@ -2,8 +2,9 @@ import { beforeAll, describe, expect, it, vi } from "vitest";
 import { render } from "@testing-library/react";
 import { afterEach } from "node:test";
 import { useTerminal } from "#/hooks/useTerminal";
-import { SocketProvider } from "#/context/socket";
 import { Command } from "#/state/commandSlice";
+import { WsClientProvider } from "#/context/ws-client-provider";
+import { ReactNode } from "react";
 
 interface TestTerminalComponentProps {
   commands: Command[];
@@ -18,6 +19,17 @@ function TestTerminalComponent({
   return <div ref={ref} />;
 }
 
+interface WrapperProps {
+  children: ReactNode;
+}
+
+
+function Wrapper({children}: WrapperProps) {
+  return (
+    <WsClientProvider enabled={true} token="NO_JWT" ghToken="NO_GITHUB" settings={null}>{children}</WsClientProvider>
+  )
+}
+
 describe("useTerminal", () => {
   const mockTerminal = vi.hoisted(() => ({
     loadAddon: vi.fn(),
@@ -50,7 +62,7 @@ describe("useTerminal", () => {
 
   it("should render", () => {
     render(<TestTerminalComponent commands={[]} secrets={[]} />, {
-      wrapper: SocketProvider,
+      wrapper: Wrapper,
     });
   });
 
@@ -61,7 +73,7 @@ describe("useTerminal", () => {
     ];
 
     render(<TestTerminalComponent commands={commands} secrets={[]} />, {
-      wrapper: SocketProvider,
+      wrapper: Wrapper,
     });
 
     expect(mockTerminal.writeln).toHaveBeenNthCalledWith(1, "echo hello");
@@ -85,7 +97,7 @@ describe("useTerminal", () => {
         secrets={[secret, anotherSecret]}
       />,
       {
-        wrapper: SocketProvider,
+        wrapper: Wrapper,
       },
     );
 

frontend/__tests__/initial-query.test.tsx

@@ -0,0 +1,17 @@
+import { describe, it, expect } from "vitest";
+import store from "../src/store";
+import { setInitialQuery, clearInitialQuery } from "../src/state/initial-query-slice";
+
+describe("Initial Query Behavior", () => {
+  it("should clear initial query when clearInitialQuery is dispatched", () => {
+    // Set up initial query in the store
+    store.dispatch(setInitialQuery("test query"));
+    expect(store.getState().initalQuery.initialQuery).toBe("test query");
+
+    // Clear the initial query
+    store.dispatch(clearInitialQuery());
+
+    // Verify initial query is cleared
+    expect(store.getState().initalQuery.initialQuery).toBeNull();
+  });
+});

frontend/__tests__/routes/_oh.app.test.tsx

@@ -0,0 +1,5 @@
+import { describe, it } from "vitest";
+
+describe("App", () => {
+  it.todo("should render");
+});

frontend/__tests__/utils/cache.test.ts

@@ -0,0 +1,53 @@
+import { afterEach } from "node:test";
+import { beforeEach, describe, expect, it, vi } from "vitest";
+import { cache } from "#/utils/cache";
+
+describe("Cache", () => {
+  const testKey = "key";
+  const testData = { message: "Hello, world!" };
+  const testTTL = 1000; // 1 second
+
+  beforeEach(() => {
+    vi.useFakeTimers();
+  });
+
+  afterEach(() => {
+    vi.useRealTimers();
+  });
+
+  it("gets data from memory if not expired", () => {
+    cache.set(testKey, testData, testTTL);
+
+    expect(cache.get(testKey)).toEqual(testData);
+  });
+
+  it("should expire after 5 minutes by default", () => {
+    cache.set(testKey, testData);
+    expect(cache.get(testKey)).not.toBeNull();
+
+    vi.advanceTimersByTime(5 * 60 * 1000 + 1);
+
+    expect(cache.get(testKey)).toBeNull();
+  });
+
+  it("returns null if cached data is expired", () => {
+    cache.set(testKey, testData, testTTL);
+
+    vi.advanceTimersByTime(testTTL + 1);
+    expect(cache.get(testKey)).toBeNull();
+  });
+
+  it("deletes data from memory", () => {
+    cache.set(testKey, testData, testTTL);
+    cache.delete(testKey);
+    expect(cache.get(testKey)).toBeNull();
+  });
+
+  it("clears all data with the app prefix from memory", () => {
+    cache.set(testKey, testData, testTTL);
+    cache.set("anotherKey", { data: "More data" }, testTTL);
+    cache.clearAll();
+    expect(cache.get(testKey)).toBeNull();
+    expect(cache.get("anotherKey")).toBeNull();
+  });
+});

frontend/__tests__/utils/extractModelAndProvider.test.ts

@@ -59,9 +59,9 @@ describe("extractModelAndProvider", () => {
       separator: "/",
     });
 
-    expect(extractModelAndProvider("claude-3-5-sonnet-20241022")).toEqual({
+    expect(extractModelAndProvider("claude-3-5-sonnet-20240620")).toEqual({
       provider: "anthropic",
-      model: "claude-3-5-sonnet-20241022",
+      model: "claude-3-5-sonnet-20240620",
       separator: "/",
     });
 
@@ -78,4 +78,3 @@ describe("extractModelAndProvider", () => {
     });
   });
 });
-

frontend/__tests__/utils/organizeModelsAndProviders.test.ts

@@ -15,7 +15,7 @@ test("organizeModelsAndProviders", () => {
     "gpt-4o",
     "together-ai-21.1b-41b",
     "gpt-4o-mini",
-    "claude-3-5-sonnet-20241022",
+    "anthropic/claude-3-5-sonnet-20241022",
     "claude-3-haiku-20240307",
     "claude-2",
     "claude-2.1",
@@ -63,4 +63,3 @@ test("organizeModelsAndProviders", () => {
     },
   });
 });
-

frontend/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "openhands-frontend",
-  "version": "0.12.0",
+  "version": "0.14.1",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "openhands-frontend",
-      "version": "0.12.0",
+      "version": "0.14.1",
       "dependencies": {
         "@monaco-editor/react": "^4.6.0",
         "@nextui-org/react": "^2.4.8",
@@ -26,6 +26,7 @@
         "isbot": "^5.1.17",
         "jose": "^5.9.4",
         "monaco-editor": "^0.52.0",
+        "posthog-js": "^1.184.1",
         "react": "^18.3.1",
         "react-dom": "^18.3.1",
         "react-highlight": "^0.15.0",
@@ -45,6 +46,7 @@
         "ws": "^8.18.0"
       },
       "devDependencies": {
+        "@playwright/test": "^1.48.2",
         "@remix-run/dev": "^2.11.2",
         "@remix-run/testing": "^2.11.2",
         "@tailwindcss/typography": "^0.5.15",
@@ -61,6 +63,7 @@
         "@typescript-eslint/parser": "^7.18.0",
         "@vitest/coverage-v8": "^1.6.0",
         "autoprefixer": "^10.4.20",
+        "cross-env": "^7.0.3",
         "eslint": "^8.57.0",
         "eslint-config-airbnb": "^19.0.4",
         "eslint-config-airbnb-typescript": "^18.0.0",
@@ -3378,6 +3381,21 @@
         "url": "https://opencollective.com/unts"
       }
     },
+    "node_modules/@playwright/test": {
+      "version": "1.48.2",
+      "resolved": "https://registry.npmjs.org/@playwright/test/-/test-1.48.2.tgz",
+      "integrity": "sha512-54w1xCWfXuax7dz4W2M9uw0gDyh+ti/0K/MxcCUxChFh37kkdxPdfZDw5QBbuPUJHr1CiHJ1hXgSs+GgeQc5Zw==",
+      "dev": true,
+      "dependencies": {
+        "playwright": "1.48.2"
+      },
+      "bin": {
+        "playwright": "cli.js"
+      },
+      "engines": {
+        "node": ">=18"
+      }
+    },
     "node_modules/@polka/url": {
       "version": "1.0.0-next.28",
       "resolved": "https://registry.npmjs.org/@polka/url/-/url-1.0.0-next.28.tgz",
@@ -7864,6 +7882,16 @@
         "node": ">=6.6.0"
       }
     },
+    "node_modules/core-js": {
+      "version": "3.38.1",
+      "resolved": "https://registry.npmjs.org/core-js/-/core-js-3.38.1.tgz",
+      "integrity": "sha512-OP35aUorbU3Zvlx7pjsFdu1rGNnD4pgw/CWoYzRY3t2EzoVT7shKHY1dlAy3f41cGIO7ZDPQimhGFTlEYkG/Hw==",
+      "hasInstallScript": true,
+      "funding": {
+        "type": "opencollective",
+        "url": "https://opencollective.com/core-js"
+      }
+    },
     "node_modules/core-util-is": {
       "version": "1.0.3",
       "resolved": "https://registry.npmjs.org/core-util-is/-/core-util-is-1.0.3.tgz",
@@ -7896,6 +7924,24 @@
         }
       }
     },
+    "node_modules/cross-env": {
+      "version": "7.0.3",
+      "resolved": "https://registry.npmjs.org/cross-env/-/cross-env-7.0.3.tgz",
+      "integrity": "sha512-+/HKd6EgcQCJGh2PSjZuUitQBQynKor4wrFbRg4DtAgS1aWO+gU52xpH7M9ScGgXSYmAVS9bIJ8EzuaGw0oNAw==",
+      "dev": true,
+      "dependencies": {
+        "cross-spawn": "^7.0.1"
+      },
+      "bin": {
+        "cross-env": "src/bin/cross-env.js",
+        "cross-env-shell": "src/bin/cross-env-shell.js"
+      },
+      "engines": {
+        "node": ">=10.14",
+        "npm": ">=6",
+        "yarn": ">=1"
+      }
+    },
     "node_modules/cross-fetch": {
       "version": "4.0.0",
       "resolved": "https://registry.npmjs.org/cross-fetch/-/cross-fetch-4.0.0.tgz",
@@ -9666,6 +9712,11 @@
         "url": "https://github.com/sponsors/wooorm"
       }
     },
+    "node_modules/fflate": {
+      "version": "0.4.8",
+      "resolved": "https://registry.npmjs.org/fflate/-/fflate-0.4.8.tgz",
+      "integrity": "sha512-FJqqoDBR00Mdj9ppamLa/Y7vxm+PRmNWA67N846RvsoYVMKB4q3y/de5PA7gUmRMYK/8CMz2GDZQmCRN1wBcWA=="
+    },
     "node_modules/file-entry-cache": {
       "version": "6.0.1",
       "resolved": "https://registry.npmjs.org/file-entry-cache/-/file-entry-cache-6.0.1.tgz",
@@ -19406,6 +19457,50 @@
         "pathe": "^1.1.2"
       }
     },
+    "node_modules/playwright": {
+      "version": "1.48.2",
+      "resolved": "https://registry.npmjs.org/playwright/-/playwright-1.48.2.tgz",
+      "integrity": "sha512-NjYvYgp4BPmiwfe31j4gHLa3J7bD2WiBz8Lk2RoSsmX38SVIARZ18VYjxLjAcDsAhA+F4iSEXTSGgjua0rrlgQ==",
+      "dev": true,
+      "dependencies": {
+        "playwright-core": "1.48.2"
+      },
+      "bin": {
+        "playwright": "cli.js"
+      },
+      "engines": {
+        "node": ">=18"
+      },
+      "optionalDependencies": {
+        "fsevents": "2.3.2"
+      }
+    },
+    "node_modules/playwright-core": {
+      "version": "1.48.2",
+      "resolved": "https://registry.npmjs.org/playwright-core/-/playwright-core-1.48.2.tgz",
+      "integrity": "sha512-sjjw+qrLFlriJo64du+EK0kJgZzoQPsabGF4lBvsid+3CNIZIYLgnMj9V6JY5VhM2Peh20DJWIVpVljLLnlawA==",
+      "dev": true,
+      "bin": {
+        "playwright-core": "cli.js"
+      },
+      "engines": {
+        "node": ">=18"
+      }
+    },
+    "node_modules/playwright/node_modules/fsevents": {
+      "version": "2.3.2",
+      "resolved": "https://registry.npmjs.org/fsevents/-/fsevents-2.3.2.tgz",
+      "integrity": "sha512-xiqMQR4xAeHTuB9uWm+fFRcIOgKBMiOBP+eXiyT7jsgVCq1bkVygt00oASowB7EdtpOHaaPgKt812P9ab+DDKA==",
+      "dev": true,
+      "hasInstallScript": true,
+      "optional": true,
+      "os": [
+        "darwin"
+      ],
+      "engines": {
+        "node": "^8.16.0 || ^10.6.0 || >=11.0.0"
+      }
+    },
     "node_modules/possible-typed-array-names": {
       "version": "1.0.0",
       "resolved": "https://registry.npmjs.org/possible-typed-array-names/-/possible-typed-array-names-1.0.0.tgz",
@@ -19653,6 +19748,31 @@
       "resolved": "https://registry.npmjs.org/postcss-value-parser/-/postcss-value-parser-4.2.0.tgz",
       "integrity": "sha512-1NNCs6uurfkVbeXG4S8JFT9t19m45ICnif8zWLd5oPSZ50QnwMfK+H3jv408d4jw/7Bttv5axS5IiHoLaVNHeQ=="
     },
+    "node_modules/posthog-js": {
+      "version": "1.184.1",
+      "resolved": "https://registry.npmjs.org/posthog-js/-/posthog-js-1.184.1.tgz",
+      "integrity": "sha512-q/1Kdard5SZnL2smrzeKcD+RuUi2PnbidiN4D3ThK20bNrhy5Z2heIy9SnRMvEiARY5lcQ7zxmDCAKPBKGSOtQ==",
+      "dependencies": {
+        "core-js": "^3.38.1",
+        "fflate": "^0.4.8",
+        "preact": "^10.19.3",
+        "web-vitals": "^4.2.0"
+      }
+    },
+    "node_modules/posthog-js/node_modules/web-vitals": {
+      "version": "4.2.4",
+      "resolved": "https://registry.npmjs.org/web-vitals/-/web-vitals-4.2.4.tgz",
+      "integrity": "sha512-r4DIlprAGwJ7YM11VZp4R884m0Vmgr6EAKe3P+kO0PPj3Unqyvv59rczf6UiGcb9Z8QxZVcqKNwv/g0WNdWwsw=="
+    },
+    "node_modules/preact": {
+      "version": "10.24.3",
+      "resolved": "https://registry.npmjs.org/preact/-/preact-10.24.3.tgz",
+      "integrity": "sha512-Z2dPnBnMUfyQfSQ+GBdsGa16hz35YmLmtTLhM169uW944hYL6xzTYkJjC07j+Wosz733pMWx0fgON3JNw1jJQA==",
+      "funding": {
+        "type": "opencollective",
+        "url": "https://opencollective.com/preact"
+      }
+    },
     "node_modules/prelude-ls": {
       "version": "1.2.1",
       "resolved": "https://registry.npmjs.org/prelude-ls/-/prelude-ls-1.2.1.tgz",

frontend/package.json

@@ -1,6 +1,6 @@
 {
   "name": "openhands-frontend",
-  "version": "0.12.0",
+  "version": "0.14.1",
   "private": true,
   "type": "module",
   "engines": {
@@ -25,6 +25,7 @@
     "isbot": "^5.1.17",
     "jose": "^5.9.4",
     "monaco-editor": "^0.52.0",
+    "posthog-js": "^1.184.1",
     "react": "^18.3.1",
     "react-dom": "^18.3.1",
     "react-highlight": "^0.15.0",
@@ -44,11 +45,12 @@
     "ws": "^8.18.0"
   },
   "scripts": {
-    "dev": "npm run make-i18n && VITE_MOCK_API=false remix vite:dev",
-    "dev:mock": "npm run make-i18n && VITE_MOCK_API=true remix vite:dev",
+    "dev": "npm run make-i18n && cross-env VITE_MOCK_API=false remix vite:dev",
+    "dev:mock": "npm run make-i18n && cross-env VITE_MOCK_API=true remix vite:dev",
     "build": "npm run make-i18n && tsc && remix vite:build",
     "start": "npx sirv-cli build/ --single",
     "test": "vitest run",
+    "test:e2e": "playwright test",
     "test:coverage": "npm run make-i18n && vitest run --coverage",
     "dev_wsl": "VITE_WATCH_USE_POLLING=true vite",
     "preview": "vite preview",
@@ -70,6 +72,7 @@
     ]
   },
   "devDependencies": {
+    "@playwright/test": "^1.48.2",
     "@remix-run/dev": "^2.11.2",
     "@remix-run/testing": "^2.11.2",
     "@tailwindcss/typography": "^0.5.15",
@@ -86,6 +89,7 @@
     "@typescript-eslint/parser": "^7.18.0",
     "@vitest/coverage-v8": "^1.6.0",
     "autoprefixer": "^10.4.20",
+    "cross-env": "^7.0.3",
     "eslint": "^8.57.0",
     "eslint-config-airbnb": "^19.0.4",
     "eslint-config-airbnb-typescript": "^18.0.0",

frontend/playwright.config.ts

@@ -0,0 +1,79 @@
+import { defineConfig, devices } from "@playwright/test";
+
+/**
+ * Read environment variables from file.
+ * https://github.com/motdotla/dotenv
+ */
+// import dotenv from 'dotenv';
+// import path from 'path';
+// dotenv.config({ path: path.resolve(__dirname, '.env') });
+
+/**
+ * See https://playwright.dev/docs/test-configuration.
+ */
+export default defineConfig({
+  testDir: "./tests",
+  /* Run tests in files in parallel */
+  fullyParallel: true,
+  /* Fail the build on CI if you accidentally left test.only in the source code. */
+  forbidOnly: !!process.env.CI,
+  /* Retry on CI only */
+  retries: process.env.CI ? 2 : 0,
+  /* Opt out of parallel tests on CI. */
+  workers: process.env.CI ? 1 : undefined,
+  /* Reporter to use. See https://playwright.dev/docs/test-reporters */
+  reporter: "html",
+  /* Shared settings for all the projects below. See https://playwright.dev/docs/api/class-testoptions. */
+  use: {
+    /* Base URL to use in actions like `await page.goto('/')`. */
+    baseURL: "http://127.0.0.1:3000",
+
+    /* Collect trace when retrying the failed test. See https://playwright.dev/docs/trace-viewer */
+    trace: "on-first-retry",
+  },
+
+  /* Configure projects for major browsers */
+  projects: [
+    {
+      name: "chromium",
+      use: { ...devices["Desktop Chrome"] },
+    },
+
+    {
+      name: "firefox",
+      use: { ...devices["Desktop Firefox"] },
+    },
+
+    {
+      name: "webkit",
+      use: { ...devices["Desktop Safari"] },
+    },
+
+    /* Test against mobile viewports. */
+    // {
+    //   name: 'Mobile Chrome',
+    //   use: { ...devices['Pixel 5'] },
+    // },
+    // {
+    //   name: 'Mobile Safari',
+    //   use: { ...devices['iPhone 12'] },
+    // },
+
+    /* Test against branded browsers. */
+    // {
+    //   name: 'Microsoft Edge',
+    //   use: { ...devices['Desktop Edge'], channel: 'msedge' },
+    // },
+    // {
+    //   name: 'Google Chrome',
+    //   use: { ...devices['Desktop Chrome'], channel: 'chrome' },
+    // },
+  ],
+
+  /* Run your local dev server before starting the tests */
+  webServer: {
+    command: "npm run dev:mock -- --port 3000",
+    url: "http://127.0.0.1:3000",
+    reuseExistingServer: !process.env.CI,
+  },
+});

frontend/public/config.json

@@ -1,4 +1,5 @@
 {
   "APP_MODE": "oss",
-  "GITHUB_CLIENT_ID": ""
-}
+  "GITHUB_CLIENT_ID": "",
+  "POSTHOG_CLIENT_KEY": "phc_3ESMmY9SgqEAGBB6sMGK5ayYHkeUuknH2vP6FmWH9RA"
+}

frontend/src/api/github.ts

@@ -122,6 +122,9 @@ export const retrieveGitHubUser = async (
       id: data.id,
       login: data.login,
       avatar_url: data.avatar_url,
+      company: data.company,
+      name: data.name,
+      email: data.email,
     };
 
     return user;
@@ -136,33 +139,6 @@ export const retrieveGitHubUser = async (
   return error;
 };
 
-/**
- * Given a GitHub token and a repository name, creates a repository for the authenticated user
- * @param token The GitHub token
- * @param repositoryName Name of the repository to create
- * @param description Description of the repository
- * @param isPrivate Boolean indicating if the repository should be private
- * @returns The created repository or an error response
- */
-export const createGitHubRepository = async (
-  token: string,
-  repositoryName: string,
-  description?: string,
-  isPrivate = true,
-): Promise<GitHubRepository | GitHubErrorReponse> => {
-  const response = await fetch("https://api.github.com/user/repos", {
-    method: "POST",
-    headers: generateGitHubAPIHeaders(token),
-    body: JSON.stringify({
-      name: repositoryName,
-      description,
-      private: isPrivate,
-    }),
-  });
-
-  return response.json();
-};
-
 export const retrieveLatestGitHubCommit = async (
   token: string,
   repository: string,

frontend/src/api/open-hands.ts

@@ -1,4 +1,5 @@
 import { request } from "#/services/api";
+import { cache } from "#/utils/cache";
 import {
   SaveFileSuccessResponse,
   FileUploadSuccessResponse,
@@ -7,6 +8,7 @@ import {
   GitHubAccessTokenResponse,
   ErrorResponse,
   GetConfigResponse,
+  GetVSCodeUrlResponse,
 } from "./open-hands.types";
 
 class OpenHands {
@@ -15,7 +17,13 @@ class OpenHands {
    * @returns List of models available
    */
   static async getModels(): Promise<string[]> {
-    return request("/api/options/models");
+    const cachedData = cache.get<string[]>("models");
+    if (cachedData) return cachedData;
+
+    const data = await request("/api/options/models");
+    cache.set("models", data);
+
+    return data;
   }
 
   /**
@@ -23,7 +31,13 @@ class OpenHands {
    * @returns List of agents available
    */
   static async getAgents(): Promise<string[]> {
-    return request(`/api/options/agents`);
+    const cachedData = cache.get<string[]>("agents");
+    if (cachedData) return cachedData;
+
+    const data = await request(`/api/options/agents`);
+    cache.set("agents", data);
+
+    return data;
   }
 
   /**
@@ -31,11 +45,23 @@ class OpenHands {
    * @returns List of security analyzers available
    */
   static async getSecurityAnalyzers(): Promise<string[]> {
-    return request(`/api/options/security-analyzers`);
+    const cachedData = cache.get<string[]>("agents");
+    if (cachedData) return cachedData;
+
+    const data = await request(`/api/options/security-analyzers`);
+    cache.set("security-analyzers", data);
+
+    return data;
   }
 
   static async getConfig(): Promise<GetConfigResponse> {
-    return request("/config.json");
+    const cachedData = cache.get<GetConfigResponse>("config");
+    if (cachedData) return cachedData;
+
+    const data = await request("/config.json");
+    cache.set("config", data);
+
+    return data;
   }
 
   /**
@@ -149,6 +175,20 @@ class OpenHands {
       true,
     );
   }
+
+  /**
+   * Get the VSCode URL
+   * @returns VSCode URL
+   */
+  static async getVSCodeUrl(): Promise<GetVSCodeUrlResponse> {
+    return request(`/api/vscode-url`, {}, false, false, 1);
+  }
+
+  static async getRuntimeId(): Promise<{ runtime_id: string }> {
+    const data = await request("/api/conversation");
+
+    return data;
+  }
 }
 
 export default OpenHands;

frontend/src/api/open-hands.types.ts

@@ -43,5 +43,11 @@ export interface Feedback {
 
 export interface GetConfigResponse {
   APP_MODE: "saas" | "oss";
-  GITHUB_CLIENT_ID: string | null;
+  GITHUB_CLIENT_ID: string;
+  POSTHOG_CLIENT_KEY: string;
+}
+
+export interface GetVSCodeUrlResponse {
+  vscode_url: string | null;
+  error?: string;
 }

frontend/src/assets/branding/all-hands-logo-spark.svg

@@ -32,4 +32,4 @@
       <rect width="69" height="46" fill="white" transform="translate(0.5)" />
     </clipPath>
   </defs>
-</svg>
+</svg>

frontend/src/assets/branding/github-logo.svg

@@ -2,4 +2,4 @@
   <path
     d="M15.359 21V17.319C15.3974 16.8654 15.3314 16.4095 15.1651 15.9814C14.9989 15.5534 14.7363 15.1631 14.3949 14.8364C17.6154 14.5035 21 13.3716 21 8.17826C20.9997 6.85027 20.4489 5.57321 19.4615 4.61139C19.9291 3.44954 19.896 2.16532 19.3692 1.02548C19.3692 1.02548 18.159 0.692576 15.359 2.43321C13.0082 1.84237 10.5302 1.84237 8.17949 2.43321C5.37949 0.692576 4.16923 1.02548 4.16923 1.02548C3.64244 2.16532 3.60938 3.44954 4.07692 4.61139C3.08218 5.58034 2.53079 6.86895 2.53846 8.2068C2.53846 13.3621 5.92308 14.494 9.14359 14.865C8.80615 15.1883 8.54591 15.574 8.3798 15.9968C8.2137 16.4196 8.14544 16.8701 8.17949 17.319V21M8.17949 18.1465C3.05128 19.5732 3.05128 15.7686 1 15.293L8.17949 18.1465Z"
     stroke="white" stroke-linecap="round" stroke-linejoin="round" />
-</svg>
+</svg>

frontend/src/assets/vscode-alt.svg

@@ -0,0 +1,57 @@
+<svg width="100" height="100" viewBox="0 0 100 100" fill="none" xmlns="http://www.w3.org/2000/svg">
+<g clip-path="url(#clip0)">
+<g filter="url(#filter0_d)">
+<mask id="mask0" mask-type="alpha" maskUnits="userSpaceOnUse" x="0" y="0" width="100" height="100">
+<path fill-rule="evenodd" clip-rule="evenodd" d="M70.9119 99.5723C72.4869 100.189 74.2828 100.15 75.8725 99.3807L96.4604 89.4231C98.624 88.3771 100 86.1762 100 83.7616V16.2392C100 13.8247 98.624 11.6238 96.4604 10.5774L75.8725 0.619067C73.7862 -0.389991 71.3446 -0.142885 69.5135 1.19527C69.252 1.38636 69.0028 1.59985 68.769 1.83502L29.3551 37.9795L12.1872 24.88C10.5891 23.6607 8.35365 23.7606 6.86938 25.1178L1.36302 30.1525C-0.452603 31.8127 -0.454583 34.6837 1.35854 36.3466L16.2471 50.0001L1.35854 63.6536C-0.454583 65.3164 -0.452603 68.1876 1.36302 69.8477L6.86938 74.8824C8.35365 76.2395 10.5891 76.34 12.1872 75.1201L29.3551 62.0207L68.769 98.1651C69.3925 98.7923 70.1246 99.2645 70.9119 99.5723ZM75.0152 27.1813L45.1092 50.0001L75.0152 72.8189V27.1813Z" fill="white"/>
+</mask>
+<g mask="url(#mask0)">
+<path d="M96.4614 10.593L75.8567 0.62085C73.4717 -0.533437 70.6215 -0.0465506 68.7498 1.83492L1.29834 63.6535C-0.515935 65.3164 -0.513852 68.1875 1.30281 69.8476L6.8125 74.8823C8.29771 76.2395 10.5345 76.339 12.1335 75.1201L93.3604 13.18C96.0854 11.102 100 13.0557 100 16.4939V16.2535C100 13.84 98.6239 11.64 96.4614 10.593Z" fill="#D9D9D9"/>
+<g filter="url(#filter1_d)">
+<path d="M96.4614 89.4074L75.8567 99.3797C73.4717 100.534 70.6215 100.047 68.7498 98.1651L1.29834 36.3464C-0.515935 34.6837 -0.513852 31.8125 1.30281 30.1524L6.8125 25.1177C8.29771 23.7605 10.5345 23.6606 12.1335 24.88L93.3604 86.8201C96.0854 88.8985 100 86.9447 100 83.5061V83.747C100 86.1604 98.6239 88.3603 96.4614 89.4074Z" fill="#E6E6E6"/>
+</g>
+<g filter="url(#filter2_d)">
+<path d="M75.8578 99.3807C73.4721 100.535 70.6219 100.047 68.75 98.1651C71.0564 100.483 75 98.8415 75 95.5631V4.43709C75 1.15852 71.0565 -0.483493 68.75 1.83492C70.6219 -0.0467614 73.4721 -0.534276 75.8578 0.618963L96.4583 10.5773C98.6229 11.6237 100 13.8246 100 16.2391V83.7616C100 86.1762 98.6229 88.3761 96.4583 89.4231L75.8578 99.3807Z" fill="white"/>
+</g>
+<g style="mix-blend-mode:overlay" opacity="0.25">
+<path style="mix-blend-mode:overlay" opacity="0.25" fill-rule="evenodd" clip-rule="evenodd" d="M70.8508 99.5723C72.4258 100.189 74.2218 100.15 75.8115 99.3807L96.4 89.4231C98.5635 88.3771 99.9386 86.1762 99.9386 83.7616V16.2391C99.9386 13.8247 98.5635 11.6239 96.4 10.5774L75.8115 0.618974C73.7252 -0.390085 71.2835 -0.142871 69.4525 1.19518C69.1909 1.38637 68.9418 1.59976 68.7079 1.83493L29.2941 37.9795L12.1261 24.88C10.528 23.6606 8.2926 23.7605 6.80833 25.1177L1.30198 30.1524C-0.51354 31.8126 -0.515625 34.6837 1.2975 36.3465L16.186 50L1.2975 63.6536C-0.515625 65.3164 -0.51354 68.1875 1.30198 69.8476L6.80833 74.8824C8.2926 76.2395 10.528 76.339 12.1261 75.1201L29.2941 62.0207L68.7079 98.1651C69.3315 98.7923 70.0635 99.2645 70.8508 99.5723ZM74.9542 27.1812L45.0481 50L74.9542 72.8188V27.1812Z" fill="url(#paint0_linear)"/>
+</g>
+</g>
+</g>
+</g>
+<defs>
+<filter id="filter0_d" x="-6.25" y="-4.16667" width="112.5" height="112.5" filterUnits="userSpaceOnUse" color-interpolation-filters="sRGB">
+<feFlood flood-opacity="0" result="BackgroundImageFix"/>
+<feColorMatrix in="SourceAlpha" type="matrix" values="0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 127 0"/>
+<feOffset dy="2.08333"/>
+<feGaussianBlur stdDeviation="3.125"/>
+<feColorMatrix type="matrix" values="0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.15 0"/>
+<feBlend mode="normal" in2="BackgroundImageFix" result="effect1_dropShadow"/>
+<feBlend mode="normal" in="SourceGraphic" in2="effect1_dropShadow" result="shape"/>
+</filter>
+<filter id="filter1_d" x="-8.39436" y="15.6951" width="116.728" height="92.6376" filterUnits="userSpaceOnUse" color-interpolation-filters="sRGB">
+<feFlood flood-opacity="0" result="BackgroundImageFix"/>
+<feColorMatrix in="SourceAlpha" type="matrix" values="0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 127 0"/>
+<feOffset/>
+<feGaussianBlur stdDeviation="4.16667"/>
+<feColorMatrix type="matrix" values="0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.25 0"/>
+<feBlend mode="overlay" in2="BackgroundImageFix" result="effect1_dropShadow"/>
+<feBlend mode="normal" in="SourceGraphic" in2="effect1_dropShadow" result="shape"/>
+</filter>
+<filter id="filter2_d" x="60.4167" y="-8.33346" width="47.9167" height="116.667" filterUnits="userSpaceOnUse" color-interpolation-filters="sRGB">
+<feFlood flood-opacity="0" result="BackgroundImageFix"/>
+<feColorMatrix in="SourceAlpha" type="matrix" values="0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 127 0"/>
+<feOffset/>
+<feGaussianBlur stdDeviation="4.16667"/>
+<feColorMatrix type="matrix" values="0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.25 0"/>
+<feBlend mode="overlay" in2="BackgroundImageFix" result="effect1_dropShadow"/>
+<feBlend mode="normal" in="SourceGraphic" in2="effect1_dropShadow" result="shape"/>
+</filter>
+<linearGradient id="paint0_linear" x1="49.939" y1="-5.19792e-05" x2="49.939" y2="100.001" gradientUnits="userSpaceOnUse">
+<stop stop-color="white"/>
+<stop offset="1" stop-color="white" stop-opacity="0"/>
+</linearGradient>
+<clipPath id="clip0">
+<rect width="100" height="100" fill="white"/>
+</clipPath>
+</defs>
+</svg>

frontend/src/components/AgentControlBar.tsx

@@ -6,7 +6,7 @@ import PlayIcon from "#/assets/play";
 import { generateAgentStateChangeEvent } from "#/services/agentStateService";
 import { RootState } from "#/store";
 import AgentState from "#/types/AgentState";
-import { useSocket } from "#/context/socket";
+import { useWsClient } from "#/context/ws-client-provider";
 
 const IgnoreTaskStateMap: Record<string, AgentState[]> = {
   [AgentState.PAUSED]: [
@@ -72,7 +72,7 @@ function ActionButton({
 }
 
 function AgentControlBar() {
-  const { send } = useSocket();
+  const { send } = useWsClient();
   const { curAgentState } = useSelector((state: RootState) => state.agent);
 
   const handleAction = (action: AgentState) => {

frontend/src/components/AgentStatusBar.tsx

@@ -1,6 +1,7 @@
 import React, { useEffect } from "react";
 import { useTranslation } from "react-i18next";
 import { useSelector } from "react-redux";
+import toast from "react-hot-toast";
 import { I18nKey } from "#/i18n/declaration";
 import { RootState } from "#/store";
 import AgentState from "#/types/AgentState";
@@ -16,7 +17,7 @@ enum IndicatorColor {
 }
 
 function AgentStatusBar() {
-  const { t } = useTranslation();
+  const { t, i18n } = useTranslation();
   const { curAgentState } = useSelector((state: RootState) => state.agent);
   const { curStatusMessage } = useSelector((state: RootState) => state.status);
 
@@ -94,15 +95,27 @@ function AgentStatusBar() {
   const [statusMessage, setStatusMessage] = React.useState<string>("");
 
   React.useEffect(() => {
-    if (curAgentState === AgentState.LOADING) {
-      const trimmedCustomMessage = curStatusMessage.status.trim();
-      if (trimmedCustomMessage) {
-        setStatusMessage(t(trimmedCustomMessage));
-        return;
+    let message = curStatusMessage.message || "";
+    if (curStatusMessage?.id) {
+      const id = curStatusMessage.id.trim();
+      if (i18n.exists(id)) {
+        message = t(curStatusMessage.id.trim()) || message;
       }
     }
+    if (curStatusMessage?.type === "error") {
+      toast.error(message);
+      return;
+    }
+    if (curAgentState === AgentState.LOADING && message.trim()) {
+      setStatusMessage(message);
+    } else {
+      setStatusMessage(AgentStatusMap[curAgentState].message);
+    }
+  }, [curStatusMessage.id]);
+
+  React.useEffect(() => {
     setStatusMessage(AgentStatusMap[curAgentState].message);
-  }, [curAgentState, curStatusMessage.status]);
+  }, [curAgentState]);
 
   return (
     <div className="flex flex-col items-center">

frontend/src/components/Jupyter.tsx

@@ -25,7 +25,11 @@ function JupyterCell({ cell }: IJupyterCell): JSX.Element {
           className="scrollbar-custom scrollbar-thumb-gray-500 hover:scrollbar-thumb-gray-400 dark:scrollbar-thumb-white/10 dark:hover:scrollbar-thumb-white/20 overflow-auto px-5"
           style={{ padding: 0, marginBottom: 0, fontSize: "0.75rem" }}
         >
-          <SyntaxHighlighter language="python" style={atomOneDark}>
+          <SyntaxHighlighter
+            language="python"
+            style={atomOneDark}
+            wrapLongLines
+          >
             {code}
           </SyntaxHighlighter>
         </pre>
@@ -78,7 +82,11 @@ function JupyterCell({ cell }: IJupyterCell): JSX.Element {
   );
 }
 
-function JupyterEditor(): JSX.Element {
+interface JupyterEditorProps {
+  maxWidth: number;
+}
+
+function JupyterEditor({ maxWidth }: JupyterEditorProps) {
   const { t } = useTranslation();
 
   const { cells } = useSelector((state: RootState) => state.jupyter);
@@ -88,7 +96,7 @@ function JupyterEditor(): JSX.Element {
     useScrollToBottom(jupyterRef);
 
   return (
-    <div className="flex-1">
+    <div className="flex-1" style={{ maxWidth }}>
       <div
         className="overflow-y-auto h-full"
         ref={jupyterRef}

frontend/src/components/analytics-consent-form-modal.tsx

@@ -0,0 +1,42 @@
+import { useFetcher } from "@remix-run/react";
+import { ModalBackdrop } from "./modals/modal-backdrop";
+import ModalBody from "./modals/ModalBody";
+import ModalButton from "./buttons/ModalButton";
+import {
+  BaseModalTitle,
+  BaseModalDescription,
+} from "./modals/confirmation-modals/BaseModal";
+
+export function AnalyticsConsentFormModal() {
+  const fetcher = useFetcher({ key: "set-consent" });
+
+  return (
+    <ModalBackdrop>
+      <fetcher.Form
+        method="POST"
+        action="/set-consent"
+        className="flex flex-col gap-2"
+      >
+        <ModalBody>
+          <BaseModalTitle title="Your Privacy Preferences" />
+          <BaseModalDescription>
+            We use tools to understand how our application is used to improve
+            your experience. You can enable or disable analytics. Your
+            preferences will be stored and can be updated anytime.
+          </BaseModalDescription>
+
+          <label className="flex gap-2 items-center self-start">
+            <input name="analytics" type="checkbox" defaultChecked />
+            Send anonymous usage data
+          </label>
+
+          <ModalButton
+            type="submit"
+            text="Confirm Preferences"
+            className="bg-primary text-white w-full hover:opacity-80"
+          />
+        </ModalBody>
+      </fetcher.Form>
+    </ModalBackdrop>
+  );
+}

frontend/src/components/attach-image-label.tsx

@@ -1,4 +1,4 @@
-import Clip from "#/assets/clip.svg?react";
+import Clip from "#/icons/clip.svg?react";
 
 export function AttachImageLabel() {
   return (

frontend/src/components/buttons/ModalButton.tsx

@@ -2,6 +2,7 @@ import clsx from "clsx";
 import React from "react";
 
 interface ModalButtonProps {
+  testId?: string;
   variant?: "default" | "text-like";
   onClick?: () => void;
   text: string;
@@ -13,6 +14,7 @@ interface ModalButtonProps {
 }
 
 function ModalButton({
+  testId,
   variant = "default",
   onClick,
   text,
@@ -24,6 +26,7 @@ function ModalButton({
 }: ModalButtonProps) {
   return (
     <button
+      data-testid={testId}
       type={type === "submit" ? "submit" : "button"}
       disabled={disabled}
       onClick={onClick}