llama-stack-mirror/llama_stack/providers/inline/agents/meta_reference
skamenan7 533c1323b8 feat: add proper BuiltinTool support for both websearch providers
- Add web_search to BuiltinTool enum for consistency with brave_search
- Add WEB_SEARCH_BRAVE_TOOL constant following existing patterns
- Update agent logic to properly map both websearch tools:
  * web_search → BuiltinTool.web_search → WEB_SEARCH_TOOL → Tavily
  * web_search_brave → BuiltinTool.brave_search → WEB_SEARCH_BRAVE_TOOL → Brave
- Add tool encoding support for BuiltinTool.web_search in Llama model
- Replace hardcoded strings with constants for maintainability
- Fix tool conflicts that caused non-deterministic provider selection

This resolves the architectural issues where both websearch providers
exposed the same tool name, causing unpredictable behavior and making
it impossible for users to reliably choose their preferred provider.

Fixes #2606
2025-07-14 10:30:21 -04:00
..
__init__.py feat: fine grained access control policy (#2264) 2025-06-03 14:51:12 -07:00
agent_instance.py feat: add proper BuiltinTool support for both websearch providers 2025-07-14 10:30:21 -04:00
agents.py feat: Add webmethod for deleting openai responses (#2160) 2025-06-30 11:28:02 +02:00
config.py feat: add list responses API (#2233) 2025-05-23 13:16:48 -07:00
openai_responses.py fix: only load mcp when enabled in tool_group (#2621) 2025-07-04 20:27:05 +05:30
persistence.py fix: AccessDeniedError leads to HTTP 500 instead of error 403 (#2595) 2025-07-03 10:50:49 -07:00
safety.py chore(api): add mypy coverage to meta_reference_safety (#2661) 2025-07-09 10:22:34 +02:00