Replace retired Gemini 2.5 Flash Lite default and pin deepseek-v4-flash to Baidu fp8 - #876
Merged
Merged
Conversation
llm_factory replaced the whole additional_kwargs dict when injecting the OpenRouter app-info headers, so any extra_body set in llm_config was silently dropped. It now merges the headers into the config's additional_kwargs. baseline.json sets provider.order to baidu/fp8 with allow_fallbacks false on openrouter-deepseek-deepseek-v4-flash-0731 so OpenRouter stops routing between the ~30 providers hosting that model. Regression test added in worker_plan/tests. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
baidu/fp8with fallbacks disabled, so OpenRouter stops load-balancing across the ~30 providers hosting that model.llm_factory.py: the OpenRouter app-info header injection replaced the wholeadditional_kwargsdict, silently dropping anyextra_bodyset in the JSON config. It now merges.worker_plan/tests/test_llm_factory.pycovering the merge.Verification
test.py: 399 tests, only the 23 pre-existing mcp_cloud errors that need PostgreSQL.ruff check --select=E9,F63,F7,F82 .clean.🤖 Generated with Claude Code