Authored by Farukest. Fixes #435. The retry summary in _handle_max_iterations() hardcoded max_tokens instead of using _max_tokens_param(), which returns max_completion_tokens for direct OpenAI API (required by gpt-4o, o-series). The first attempt already used _max_tokens_param correctly — only the retry path was wrong. Includes 4 tests for _max_tokens_param provider detection.
201 KiB
201 KiB