test_litellm_text_completion_metrics[sync] and [async] in py/src/braintrust/integrations/litellm/test_litellm.py call gpt-3.5-turbo-instruct, which OpenAI has retired. Re-recording returns a 404, so every litellm latest re-record (most recently #849, litellm 1.104.0) has to restore these cassettes from git.
Fix
Point the test at a model that still supports the legacy completions endpoint (or another provider that litellm routes through text_completion), or retire the test if the path is covered elsewhere. Then re-record cassettes/latest/test_litellm_text_completion_metrics*.
Parent: #858 (found while reviewing the 2026-10-05 weekly dependency update, #849).
🤖 Generated with Claude Code
test_litellm_text_completion_metrics[sync]and[async]inpy/src/braintrust/integrations/litellm/test_litellm.pycallgpt-3.5-turbo-instruct, which OpenAI has retired. Re-recording returns a 404, so every litellmlatestre-record (most recently #849, litellm 1.104.0) has to restore these cassettes from git.Fix
Point the test at a model that still supports the legacy completions endpoint (or another provider that litellm routes through
text_completion), or retire the test if the path is covered elsewhere. Then re-recordcassettes/latest/test_litellm_text_completion_metrics*.Parent: #858 (found while reviewing the 2026-10-05 weekly dependency update, #849).
🤖 Generated with Claude Code