Fix issue #5076: Integration test github action (#5077)

Co-authored-by: Engel Nyst <[email protected]>

.github/workflows/eval-runner.yml

@@ -1,4 +1,4 @@
-name: Run Evaluation
+name: Run SWE-Bench Evaluation
 
 on:
   pull_request:
@@ -58,24 +58,6 @@ jobs:
           echo "api_key = \"$DEEPSEEK_API_KEY\"" >> config.toml
           echo "temperature = 0.0" >> config.toml
 
-      - name: Run integration test evaluation
-        env:
-          ALLHANDS_API_KEY: ${{ secrets.ALLHANDS_EVAL_RUNTIME_API_KEY }}
-          RUNTIME: remote
-          SANDBOX_REMOTE_RUNTIME_API_URL: https://runtime.eval.all-hands.dev
-          EVAL_DOCKER_IMAGE_PREFIX: us-central1-docker.pkg.dev/evaluation-092424/swe-bench-images
-
-        run: |
-          poetry run ./evaluation/integration_tests/scripts/run_infer.sh llm.eval HEAD CodeActAgent '' $N_PROCESSES
-
-          # get evaluation report
-          REPORT_FILE=$(find evaluation/evaluation_outputs/outputs/integration_tests/CodeActAgent/deepseek-chat_maxiter_10_N* -name "report.md" -type f | head -n 1)
-          echo "REPORT_FILE: $REPORT_FILE"
-          echo "INTEGRATION_TEST_REPORT<<EOF" >> $GITHUB_ENV
-          cat $REPORT_FILE >> $GITHUB_ENV
-          echo >> $GITHUB_ENV
-          echo "EOF" >> $GITHUB_ENV
-
       - name: Run SWE-Bench evaluation
         env:
           ALLHANDS_API_KEY: ${{ secrets.ALLHANDS_EVAL_RUNTIME_API_KEY }}
@@ -143,9 +125,6 @@ jobs:
               **SWE-Bench Evaluation Report**
               ${{ env.SWEBENCH_REPORT }}
               ---
-              **Integration Tests Evaluation Report**
-              ${{ env.INTEGRATION_TEST_REPORT }}
-              ---
               You can download the full evaluation outputs [here](${{ env.ARTIFACT_URL }}).
 
       - name: Post to a Slack channel

.github/workflows/integration-runner.yml

@@ -0,0 +1,158 @@
+name: Run Integration Tests
+
+on:
+  pull_request:
+    types: [labeled]
+  workflow_dispatch:
+    inputs:
+      reason:
+        description: 'Reason for manual trigger'
+        required: true
+        default: ''
+  schedule:
+    - cron: '30 22 * * *'  # Runs at 10:30pm UTC every day
+
+env:
+  N_PROCESSES: 10 # Global configuration for number of parallel processes for evaluation
+
+jobs:
+  run-integration-tests:
+    if: github.event.label.name == 'integration-test' || github.event_name == 'workflow_dispatch' || github.event_name == 'schedule'
+    runs-on: ubuntu-latest
+    permissions:
+      contents: "read"
+      id-token: "write"
+      pull-requests: "write"
+      issues: "write"
+    strategy:
+      matrix:
+        python-version: ["3.12"]
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Install poetry via pipx
+        run: pipx install poetry
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: "poetry"
+
+      - name: Comment on PR if 'integration-test' label is present
+        if: github.event_name == 'pull_request' && github.event.label.name == 'integration-test'
+        uses: KeisukeYamashita/create-comment@v1
+        with:
+          unique: false
+          comment: |
+            Hi! I started running the integration tests on your PR. You will receive a comment with the results shortly.
+
+      - name: Install Python dependencies using Poetry
+        run: poetry install --without evaluation,llama-index
+
+      - name: Configure config.toml for testing with Haiku
+        env:
+          LLM_MODEL: "litellm_proxy/claude-3-5-haiku-20241022"
+          LLM_API_KEY: ${{ secrets.LLM_API_KEY }}
+          LLM_BASE_URL: ${{ secrets.LLM_BASE_URL }}
+        run: |
+          echo "[llm.eval]" > config.toml
+          echo "model = \"$LLM_MODEL\"" >> config.toml
+          echo "api_key = \"$LLM_API_KEY\"" >> config.toml
+          echo "base_url = \"$LLM_BASE_URL\"" >> config.toml
+          echo "temperature = 0.0" >> config.toml
+
+      - name: Build environment
+        run: make build
+
+      - name: Run integration test evaluation for Haiku
+        env:
+          SANDBOX_FORCE_REBUILD_RUNTIME: True
+        run: |
+          poetry run ./evaluation/integration_tests/scripts/run_infer.sh llm.eval HEAD CodeActAgent '' $N_PROCESSES '' 'haiku_run'
+
+          # get integration tests report
+          REPORT_FILE_HAIKU=$(find evaluation/evaluation_outputs/outputs/integration_tests/CodeActAgent/*haiku*_maxiter_10_N* -name "report.md" -type f | head -n 1)
+          echo "REPORT_FILE: $REPORT_FILE_HAIKU"
+          echo "INTEGRATION_TEST_REPORT_HAIKU<<EOF" >> $GITHUB_ENV
+          cat $REPORT_FILE_HAIKU >> $GITHUB_ENV
+          echo >> $GITHUB_ENV
+          echo "EOF" >> $GITHUB_ENV
+
+      - name: Wait a little bit
+        run: sleep 10
+
+      - name: Configure config.toml for testing with DeepSeek
+        env:
+          LLM_MODEL: "litellm_proxy/deepseek-chat"
+          LLM_API_KEY: ${{ secrets.LLM_API_KEY }}
+          LLM_BASE_URL: ${{ secrets.LLM_BASE_URL }}
+        run: |
+          echo "[llm.eval]" > config.toml
+          echo "model = \"$LLM_MODEL\"" >> config.toml
+          echo "api_key = \"$LLM_API_KEY\"" >> config.toml
+          echo "base_url = \"$LLM_BASE_URL\"" >> config.toml
+          echo "temperature = 0.0" >> config.toml
+
+      - name: Run integration test evaluation for DeepSeek
+        env:
+          SANDBOX_FORCE_REBUILD_RUNTIME: True
+        run: |
+          poetry run ./evaluation/integration_tests/scripts/run_infer.sh llm.eval HEAD CodeActAgent '' $N_PROCESSES '' 'deepseek_run'
+
+          # get integration tests report
+          REPORT_FILE_DEEPSEEK=$(find evaluation/evaluation_outputs/outputs/integration_tests/CodeActAgent/deepseek*_maxiter_10_N* -name "report.md" -type f | head -n 1)
+          echo "REPORT_FILE: $REPORT_FILE_DEEPSEEK"
+          echo "INTEGRATION_TEST_REPORT_DEEPSEEK<<EOF" >> $GITHUB_ENV
+          cat $REPORT_FILE_DEEPSEEK >> $GITHUB_ENV
+          echo >> $GITHUB_ENV
+          echo "EOF" >> $GITHUB_ENV
+
+      - name: Create archive of evaluation outputs
+        run: |
+          TIMESTAMP=$(date +'%y-%m-%d-%H-%M')
+          cd evaluation/evaluation_outputs/outputs  # Change to the outputs directory
+          tar -czvf ../../../integration_tests_${TIMESTAMP}.tar.gz integration_tests/CodeActAgent/*  # Only include the actual result directories
+
+      - name: Upload evaluation results as artifact
+        uses: actions/upload-artifact@v4
+        id: upload_results_artifact
+        with:
+          name: integration-test-outputs-${{ github.run_id }}-${{ github.run_attempt }}
+          path: integration_tests_*.tar.gz
+
+      - name: Get artifact URLs
+        run: |
+          echo "ARTIFACT_URL=${{ steps.upload_results_artifact.outputs.artifact-url }}" >> $GITHUB_ENV
+
+      - name: Set timestamp and trigger reason
+        run: |
+          echo "TIMESTAMP=$(date +'%Y-%m-%d-%H-%M')" >> $GITHUB_ENV
+          if [[ "${{ github.event_name }}" == "pull_request" ]]; then
+            echo "TRIGGER_REASON=pr-${{ github.event.pull_request.number }}" >> $GITHUB_ENV
+          elif [[ "${{ github.event_name }}" == "workflow_dispatch" ]]; then
+            echo "TRIGGER_REASON=manual-${{ github.event.inputs.reason }}" >> $GITHUB_ENV
+          else
+            echo "TRIGGER_REASON=nightly-scheduled" >> $GITHUB_ENV
+          fi
+
+      - name: Comment with results and artifact link
+        id: create_comment
+        uses: KeisukeYamashita/create-comment@v1
+        with:
+          # if triggered by PR, use PR number, otherwise use 5077 as fallback issue number for manual triggers
+          number: ${{ github.event_name == 'pull_request' && github.event.pull_request.number || 5077 }}
+          unique: false
+          comment: |
+              Trigger by: ${{ github.event_name == 'pull_request' && format('Pull Request (integration-test label on PR #{0})', github.event.pull_request.number) || (github.event_name == 'workflow_dispatch' && format('Manual Trigger: {0}', github.event.inputs.reason)) || 'Nightly Scheduled Run' }}
+              Commit: ${{ github.sha }}
+              **Integration Tests Report (Haiku)**
+              Haiku LLM Test Results:
+              ${{ env.INTEGRATION_TEST_REPORT_HAIKU }}
+              ---
+              **Integration Tests Report (DeepSeek)**
+              DeepSeek LLM Test Results:
+              ${{ env.INTEGRATION_TEST_REPORT_DEEPSEEK }}
+              ---
+              Download evaluation outputs (includes both Haiku and DeepSeek results): [Download](${{ steps.upload_results_artifact.outputs.artifact-url }})

evaluation/integration_tests/run_infer.py

@@ -48,13 +48,19 @@ def get_config(
             # use default base_container_image
             enable_auto_lint=True,
             use_host_network=False,
-            timeout=100,
+            timeout=300,
+            # Add platform to the sandbox config to solve issue 4401
+            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_runtime_alive=False,
+            remote_runtime_init_timeout=3600,
         ),
         # do not mount workspace
         workspace_base=None,
         workspace_mount_path=None,
+        # debug
+        debug=True,
     )
     config.set_llm_config(
         update_llm_config_for_completions_logging(
@@ -129,7 +135,12 @@ def process_instance(
     # # result evaluation
     # # =============================================
 
-    histories = [event_to_dict(event) for event in state.history]
+    histories = state.history
+
+    # some basic check
+    logger.info(f'Total events in history: {len(histories)}')
+    assert len(histories) > 0, 'History should not be empty'
+
     test_result: TestResult = test_class.verify_result(runtime, histories)
     metrics = state.metrics.get() if state.metrics else None
 
@@ -139,7 +150,7 @@ def process_instance(
         instance=instance.to_dict(),
         instruction=instruction,
         metadata=metadata,
-        history=histories,
+        history=[event_to_dict(event) for event in histories],
         metrics=metrics,
         error=state.last_error if state and state.last_error else None,
         test_result=test_result.model_dump(),

evaluation/integration_tests/tests/t05_simple_browsing.py

@@ -108,6 +108,8 @@ def initialize_runtime(cls, runtime: Runtime) -> None:
 
     @classmethod
     def verify_result(cls, runtime: Runtime, histories: list[Event]) -> TestResult:
+        from openhands.core.logger import openhands_logger as logger
+
         # check if the "The answer is OpenHands is all you need!" is in any message
         message_actions = [
             event
@@ -116,19 +118,29 @@ def verify_result(cls, runtime: Runtime, histories: list[Event]) -> TestResult:
                 event, (MessageAction, AgentFinishAction, AgentDelegateObservation)
             )
         ]
+        logger.debug(f'Total message-like events: {len(message_actions)}')
+
         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)}')
+            try:
+                if isinstance(event, AgentDelegateObservation):
+                    content = event.content
+                elif isinstance(event, AgentFinishAction):
+                    content = event.outputs.get('content', '')
+                elif isinstance(event, MessageAction):
+                    content = event.content
+                else:
+                    logger.warning(f'Unexpected event type: {type(event)}')
+                    continue
 
-            if 'OpenHands is all you need!' in content:
-                return TestResult(success=True)
+                if 'OpenHands is all you need!' in content:
+                    return TestResult(success=True)
+            except Exception as e:
+                logger.error(f'Error processing event: {e}')
+
+        logger.debug(
+            f'Total messages: {len(message_actions)}. Messages: {message_actions}'
+        )
         return TestResult(
             success=False,
-            reason=f'The answer is not found in any message. Total messages: {len(message_actions)}. Messages: {message_actions}',
+            reason=f'The answer is not found in any message. Total messages: {len(message_actions)}.',
         )

evaluation/integration_tests/tests/t06_github_pr_browsing.py

@@ -14,31 +14,43 @@ def initialize_runtime(cls, runtime: Runtime) -> None:
 
     @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
+        from openhands.core.logger import openhands_logger as logger
+
+        # check if the license information is in any message
         message_actions = [
             event
             for event in histories
             if isinstance(
                 event, (MessageAction, AgentFinishAction, AgentDelegateObservation)
             )
         ]
+        logger.info(f'Total message-like events: {len(message_actions)}')
+
         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)
+            try:
+                if isinstance(event, AgentDelegateObservation):
+                    content = event.content
+                elif isinstance(event, AgentFinishAction):
+                    content = event.outputs.get('content', '')
+                elif isinstance(event, MessageAction):
+                    content = event.content
+                else:
+                    logger.warning(f'Unexpected event type: {type(event)}')
+                    continue
+
+                if (
+                    'non-commercial' in content
+                    or 'MIT' in content
+                    or 'Apache 2.0' in content
+                ):
+                    return TestResult(success=True)
+            except Exception as e:
+                logger.error(f'Error processing event: {e}')
+
+        logger.debug(
+            f'Total messages: {len(message_actions)}. Messages: {message_actions}'
+        )
         return TestResult(
             success=False,
-            reason=f'The answer is not found in any message. Total messages: {len(message_actions)}. Messages: {message_actions}',
+            reason=f'The answer is not found in any message. Total messages: {len(message_actions)}.',
         )