fix(anthropic): map refusal and context-window stop reasons#316
Open
sven2718 wants to merge 1 commit into
Open
fix(anthropic): map refusal and context-window stop reasons#316sven2718 wants to merge 1 commit into
sven2718 wants to merge 1 commit into
Conversation
The Anthropic API can end a response with stop_reason "refusal" (safety classifier intervention) or "model_context_window_exceeded", but both fell through to FinishReasonUnknown, leaving clients unable to distinguish them from a missing stop reason. Map "refusal" to FinishReasonContentFilter, matching how the OpenAI and Google providers already surface their native content-filter stops, and map "model_context_window_exceeded" to FinishReasonLength alongside "max_tokens". 💘 Generated with Crush Assisted-by: Kimi K3 (max thinking) via Crush <crush@charm.land>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
One thing that's been annoying me about working with fable-5 in crush is that if you trip over anthropic's content classifier (so fable is refusing to process the API calls), crush will just end the turn with no indication of what's gone wrong.
This is easy to fix, because FinishReasonContentFilter is a thing that fantasy already supports -- it just doesn't understand Anthropic's flavor of the error.
The core fix is a 2-line diff. While we're here it's probably also worth writing a unit test and extending the coverage of FinishReasonLength.