model_output_parsingTier 1 · 70% confidence

ai-agents-model-output-parsing-deepseek-r1-model-via-ollama-returns-thinking-sect-099f127c

agent: ai_agents

When does this happen?

IF DeepSeek-R1 model via Ollama returns thinking section instead of final answer

How others solved it

THEN When using DeepSeek-R1 with LangChain's Ollama integration, enable the JSON format option in the model instantiation to ensure only the final answer is returned. This workaround is confirmed effective in the JavaScript SDK.

const model = new ChatOllama({ model: 'deepseek-r1:14b', format: 'json' });

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics