fix: suppress thinking-mode output in ReviewPlanTask#159
Merged
neoneye merged 1 commit intoPlanExeOrg:mainfrom Mar 7, 2026
Merged
fix: suppress thinking-mode output in ReviewPlanTask#159neoneye merged 1 commit intoPlanExeOrg:mainfrom
neoneye merged 1 commit intoPlanExeOrg:mainfrom
Conversation
Qwen 3.5-35B-A3B enters chain-of-thought reasoning mode on complex multi-turn prompts, outputting a large 'Thinking Process:' block before the JSON. This exhausts the num_output token budget, leaving no tokens for the actual JSON response. JSON extraction then fails with 'Could not extract json string from output'. Fix: explicit instruction to output ONLY the JSON object without any reasoning preamble. Confirmed failure: ReviewPlanTask failed 2026-03-07 01:05 EST on Mac Mini M4 Pro 64GB, Qwen 3.5-35B-A3B, num_output=8192.
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
ReviewPlanTaskfails withCould not extract json string from output: Thinking Process:\n\n1. Analyze the Request:...Qwen 3.5-35B-A3B enters chain-of-thought reasoning mode on complex multi-turn prompts and outputs a large
Thinking Process:block before generating JSON. This exhausts thenum_outputtoken budget, leaving no tokens for the actual JSON response. JSON extraction finds no JSON in the output and raisesValueError.This is a different failure class from the Pydantic tail-field truncation fixed in PRs #153/#155/#156. Those were missing required fields. This is missing JSON entirely — the model's reasoning trace consumed the entire output budget.
Fix
One line added to
REVIEW_PLAN_SYSTEM_PROMPT:Changes
worker_plan/worker_plan_internal/plan/review_plan.py— 2 linesThis PR intentionally changes only 1 file.