DeepSeek R1 response lacks thinking process

[Wangwentao1]

I use a custom API provider to integrate DeepSeek R1 in perplexica，but I found the thinking process is missing: Like in the picture

I debuged the raw message in the metasearchagent.ts with docker logger. So the original LLM response lacks the thinking process. I want to figure it out why.

[Wangwentao1]

I think this bug is related to langchain async request?

[Wangwentao1]

langchain-ai/langchain#29513

[ItzCrazyKns]

If thinking tags are returned by the provider it should be displayed like a charm by this PR: #576 so please refer to it for more updates.