fix: disable thinking mode for structured JSON output requests#34
Open
back2zion wants to merge 1 commit intonikmcfly:mainfrom
Open
fix: disable thinking mode for structured JSON output requests#34back2zion wants to merge 1 commit intonikmcfly:mainfrom
back2zion wants to merge 1 commit intonikmcfly:mainfrom
Conversation
Models with reasoning/thinking capabilities (e.g. Qwen3 with --reasoning-parser) emit <think> tags before generating content. When combined with response_format=json_object and vLLM's guided decoding, this causes an infinite abort-retry loop because the thinking tokens violate the JSON schema constraint. Automatically sets enable_thinking=false via chat_template_kwargs whenever structured JSON output is requested.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Models with reasoning/thinking capabilities (e.g. Qwen3 with
--reasoning-parser) emit<think>tags before generating content. When combined withresponse_format=json_objectand vLLM's guided decoding, this causes an infinite abort-retry loop — the thinking tokens violate the JSON schema, vLLM aborts, the client retries, and the process hangs indefinitely.This blocks persona generation at ~50-53 out of 75 agents every time.
Fix
Automatically sets
enable_thinking=falseviachat_template_kwargswhenever structured JSON output is requested, in both:llm_client.py(general LLM client used for ontology generation etc.)oasis_profile_generator.py(agent persona generation)Test plan
<think>tag contamination