Reduce malformed managed ranking output within the existing token budget - #43
Merged
Merged
Conversation
This was referenced Oct 5, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Live full-batch ranking intermittently returned invalid structured JSON or exhausted its 15-second deadline even after Gemini access-denial recovery. Request compact numeric score tuples instead of repeating field names for every candidate, and decode them into the unchanged normalized classifier response. Existing object answers remain accepted; invalid output still fails validation and uses the existing bounded retry/fallback.
Keep
model: auto, all 30 candidates/labels, the 2,000-token ceiling, original deadline and request count. Add allowlisted finish-reason/token metadata for invalid JSON without logging prompts or provider content. The observed JSON failure is confirmed; provider-side truncation is not yet proven.Validation: 15 focused tests and the full 220-test suite passed, including 30-candidate ordinal and perspective decoding, malformed tuples, truncation recovery and log privacy. Package validation, full public build and Wrangler deployment dry-run passed. Exact-source CI must pass before release; live strict qualification follows deployment. Refs #30 and sass-maker/free-ai#95.