Test: use default agent preset for integration tests

[xingyaoww]

We have a default preset for agents here: https://github.com/All-Hands-AI/agent-sdk/blob/main/openhands/sdk/preset/default.py

Our downstream product (GUI/CLI) will use these preset in production, and it will be nice if we can directly import these presets in integration tests so we can test the agent we will ship to production.

[all-hands-bot]

This issue is stale because it has been open for 40 days with no activity. Remove the stale label or leave a comment, otherwise it will be closed in 10 days.

[simonrosenberg]

@OpenHands locate where all the integration tests are defined and make sure they use the default agent when applicable.

[openhands-ai]

I'm on it! simonrosenberg can track my progress at all-hands.dev

[openhands-ai]

Summary

I've successfully updated the integration tests to use the default agent preset from openhands.tools.preset.default, ensuring tests validate the same agent configuration shipped to production (GUI/CLI).

Changes Made (12 files, -54 lines net):

Core Changes:

1. tests/integration/base.py - Updated BaseIntegrationTest to:

   • Import and use get_default_tools() and get_default_condenser() from the preset
   • Add enable_browser property (default False) for tests needing browser support
   • Changed tools property from abstract to concrete using the default preset
   • Changed condenser property to use the default condenser

2. tests/integration/utils/behavior_helpers.py - Updated default_behavior_tools() to use get_default_tools(enable_browser=False) instead of manual tool registration

Functional Tests (t01-t08):

• Removed custom tools property from all tests - they now inherit default tools from base class
• Tests t05 and t06 (browsing tests) override enable_browser to return True

Special Cases:

• t09 keeps its custom condenser for token-based condensation testing, but now uses get_default_tools() for tools
• b05 uses default tools from base class

Checklist:

• ✅ Integration tests now use the default agent preset
• ✅ Browser tests (t05, t06) properly enable browser via enable_browser property
• ✅ t09 maintains custom condenser for testing token-based condensation
• ✅ Pre-commit hooks pass on all modified files
• ✅ Changes are concise (net reduction of 54 lines by removing redundant code)
• ✅ Branch pushed and PR created

Pull Request:

#1565

View full conversation

[all-hands-bot]

This issue is stale because it has been open for 40 days with no activity. Remove the stale label or leave a comment, otherwise it will be closed in 10 days.

[xingyaoww]

@OpenHands please help me check if we've already implemented this

[openhands-ai]

I'm on it! xingyaoww can track my progress at all-hands.dev