mirror of
https://github.com/crewAIInc/crewAI.git
synced 2025-12-16 04:18:35 +00:00
02f9a36acb
Make max_tokens optional and compute dynamically when not set by user. Previously, max_tokens defaulted to 4096 and was always passed to the Anthropic API, causing the model to generate up to 4096 tokens even for simple queries that should only need a few tokens. This resulted in extremely slow execution times. Changes: - Changed max_tokens parameter from int (default 4096) to int | None (default None) - Added dynamic computation in _prepare_completion_params(): - Default: 1024 tokens (much more reasonable for most queries) - Large context models (200k+): 2048 tokens - User-specified values are always respected - Updated docstring to reflect that max_tokens is now optional - Added comprehensive tests covering: - Explicit max_tokens values are passed through unchanged - Default behavior computes reasonable max_tokens dynamically - max_tokens=None uses dynamic computation - Dynamic values are appropriate for model context window size - User-provided values are always respected This fix aligns with the v0.203.1 behavior where max_tokens was optional and only passed when explicitly set, while maintaining compatibility with the Anthropic SDK requirement that max_tokens must be provided. Fixes #3807 Co-Authored-By: João <joao@crewai.com>