yashwantbezawada:fix/anthropic-cache-control-cleanup-issue-33709 - Branch - langchain-ai/langchain

fix(anthropic): clean up cache_control in middleware to prevent fallback errors

#33894

Comparing

yashwantbezawada:fix/anthropic-cache-control-cleanup-issue-33709

(

2cb902d

) with

master

(

cc3af82

)

Untouched: 3

Skipped: 31

No successful run was found on master ( 94d5271) during the generation of this report, so cc3af82 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report.

Benchmarks

Skipped (31)

Passed

test_init_time_with_client

libs/partners/anthropic/tests/unit_tests/test_standard.py

+1%

2.3 ms*2.3 ms

test_init_time

libs/partners/anthropic/tests/unit_tests/test_standard.py::TestAnthropicStandard

764.4 µs*762.5 µs

test_stream_time

libs/partners/anthropic/tests/integration_tests/test_standard.py::TestAnthropicStandard

34.8 ms*34.9 ms

Commits

Click on a commit to change the comparison range

Base

master

cc3af82

+0.13%

fix(anthropic): clean up cache_control in middleware to prevent fallback errors Fixes issue where AnthropicPromptCachingMiddleware's cache_control parameter persisted in model_settings when ModelFallbackMiddleware switched to non-Anthropic models (OpenAI, Google), causing TypeError. The fix uses try/finally blocks to ensure cache_control is always removed from model_settings after handler execution, regardless of success or failure. This prevents the Anthropic-specific parameter from being passed to fallback models. Changes: - Added cleanup logic in wrap_model_call() and awrap_model_call() - Updated existing tests to verify cleanup behavior - Added comprehensive tests for both success and error cases Fixes #33709

2cb902d

22 hours ago

by yashwantbezawada

Home Terms Privacy Docs