-
Notifications
You must be signed in to change notification settings - Fork 1.4k
DeepSeek V3.2 Thinking Mode Integration #1134
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: main
Are you sure you want to change the base?
Conversation
📋 Review SummaryThis PR implements support for DeepSeek V3.2's Thinking Mode, enabling Qwen Code to properly exercise and validate the existing thinking-mode pipeline with a real model that exposes reasoning capabilities. The changes span across core components, CLI UI components, and session management to handle reasoning content from DeepSeek V3.2. 🔍 General Feedback
🎯 Specific Feedback🟢 Medium
🔵 Low
✅ Highlights
|
Code Coverage Summary
CLI Package - Full Text ReportCore Package - Full Text ReportFor detailed HTML reports, please see the 'coverage-reports-22.x-ubuntu-latest' artifact from the main CI run. |
Summary
This PR implements support for DeepSeek V3.2's Thinking Mode, enabling Qwen Code to properly exercise and validate our existing thinking-mode pipeline with a real model that exposes reasoning capabilities.
Changes
Core: OpenAI Content Generator
reasoning_contentfield used by DeepSeek V3.2reasoning_contentdelta chunksCore: Turn Processing
getThoughtText()utilityCore: Chat History
stripThoughtsFromHistory()to properly filter out thought parts before sending context back to the APICore: Session Service
CLI: UI Components
GeminiThoughtMessagecomponent with distinctive styling (✦ prefix, dimmed secondary color)GeminiThoughtMessageContentfor rendering thought content blocksHistoryItemDisplayto properly render thought messages in conversation historyCLI: Streaming Hook
useGeminiStreamto handle thought events from the streaming pipelineCLI: Resume History Utils
Telemetry
Testing
reasoning_contentconversion in both streaming and non-streaming scenariosUsage
To use DeepSeek V3.2 with Thinking Mode:
deepseek-reasonerBonus: Kimi K2 Thinking can also work, while other thinking models have not been tested.
Screenshots
Checklist