Devin AI 06e2683fd8 feat: implement fallback LLMs for agent execution ...

- Add fallback_llms field to Agent class to support multiple LLM fallbacks
- Modify get_llm_response in agent_utils.py to try fallback LLMs when primary fails
- Update CrewAgentExecutor and LiteAgent to pass fallback LLMs to get_llm_response
- Add smart error handling that skips fallbacks for auth errors but tries them for other failures
- Add comprehensive tests covering all fallback scenarios
- Maintain full backward compatibility for agents without fallback LLMs

Addresses GitHub Issue #3032: Support Fallback LLMs for Agent Execution

Co-Authored-By: João <joao@crewai.com>

2025-06-19 06:02:39 +00:00

9.4 KiB

Raw Blame History

View Raw