langchain-ollama allows ollama<0.6.3, which rejects model-defined thinking levels with a Pydantic ValidationError

当你用 LangChain 的 ChatOllama 连接本地 Ollama 模型,并把 reasoning 设为模型自定义的思考等级(例如 "xhigh" 或 "max" )时,会在请求发往 Ollama 之前就被 Pydantic 拦截报错。核心原因是 langchain-ollama 依赖声明



![[BUG]: AnythingLLM stops working when/while/after mass-removing documents from workspace](https://www.chat-gpts.plus/wp-content/uploads/2026/10/6606-4a404777-768x403.jpg)
![[Bug] Client sends empty _meta:{} on every request; strict servers (Meta Ads MCP) reject with HTTP 400](https://www.chat-gpts.plus/wp-content/uploads/2026/10/3473-a9bea40e-768x403.jpg)

![[Bug]: s3_v2 async 500/503 retries are bypassed by HTTPStatusError](https://www.chat-gpts.plus/wp-content/uploads/2026/10/42868-cdeacc24-768x403.jpg)
![[Bug]: timestamp_granularities=["segment", "word"] only returns the last granularity for Whisper](https://www.chat-gpts.plus/wp-content/uploads/2026/10/35937-6543e9b1-768x403.jpg)
