[GH-ISSUE #5509] [BUG]: 'reasoning_content' in thinking mode not being passed to Deepseek API #5135

Closed
opened 2026-06-05 14:52:13 -04:00 by yindo · 5 comments
Owner

Originally created by @Beginner50 on GitHub (Apr 24, 2026).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/5509

Originally assigned to: @timothycarambat on GitHub.

How are you running AnythingLLM?

AnythingLLM desktop app

What happened?

The discovered models from Deepseek changed from the soon to be deprecated, 'Deepseek-chat' and 'Deepseek-reasoning' to 'Deepseek-v4-flash' and 'Deepseek-v4-pro' respectively.

The issue: 'The agent model failed to respond: 400 The reasoning_content in the thinking mode must be passed back to the API.' seems to arise after the agent finishes execution of a skill. This problem seems to occur only with the latest Deepseek models and not with any other providers as far as I am concerned; Gemini seems to work fine.

Image

Are there known steps to reproduce?

Set Deepseek API key and select the latest 'Deepseek-v4-flash' or 'Deepseek-v4-pro' model and upload a random document in the workspace.

Instruct the model to summarize the given document in the workspace in the prompt

LLM Provider & Model (if applicable)

Deepseek (Deepseek-v4-flash & Deepseek-v4-pro)

Embedder Provider & Model (if applicable)

AnythingLLM Embedder (all-MiniLM-L6-v2)

Originally created by @Beginner50 on GitHub (Apr 24, 2026). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/5509 Originally assigned to: @timothycarambat on GitHub. ### How are you running AnythingLLM? AnythingLLM desktop app ### What happened? The discovered models from Deepseek changed from the soon to be deprecated, 'Deepseek-chat' and 'Deepseek-reasoning' to 'Deepseek-v4-flash' and 'Deepseek-v4-pro' respectively. The issue: 'The agent model failed to respond: 400 The `reasoning_content` in the thinking mode must be passed back to the API.' seems to arise after the agent finishes execution of a skill. This problem seems to occur only with the latest Deepseek models and not with any other providers as far as I am concerned; Gemini seems to work fine. <img width="922" height="467" alt="Image" src="https://github.com/user-attachments/assets/40e7ab99-070b-47ea-a745-f860fe744133" /> ### Are there known steps to reproduce? Set Deepseek API key and select the latest 'Deepseek-v4-flash' or 'Deepseek-v4-pro' model and upload a random document in the workspace. Instruct the model to summarize the given document in the workspace in the prompt ### LLM Provider & Model (if applicable) Deepseek (Deepseek-v4-flash & Deepseek-v4-pro) ### Embedder Provider & Model (if applicable) AnythingLLM Embedder (all-MiniLM-L6-v2)
yindo added the bug label 2026-06-05 14:52:13 -04:00
yindo closed this issue 2026-06-05 14:52:13 -04:00
Author
Owner

@ShawnWx2019 commented on GitHub (Apr 24, 2026):

Docker version met the same problem, when using rag-memory tool

<!-- gh-comment-id:4311782665 --> @ShawnWx2019 commented on GitHub (Apr 24, 2026): Docker version met the same problem, when using rag-memory tool
Author
Owner

@ohromeo commented on GitHub (May 6, 2026):

has this been fixed?, as it is still happening

<!-- gh-comment-id:4388063332 --> @ohromeo commented on GitHub (May 6, 2026): has this been fixed?, as it is still happening
Author
Owner

@Beginner50 commented on GitHub (May 6, 2026):

has this been fixed?, as it is still happening

The desktop app has not been updated for the fix. I have switched to the docker version instead.

<!-- gh-comment-id:4388291603 --> @Beginner50 commented on GitHub (May 6, 2026): > has this been fixed?, as it is still happening The desktop app has not been updated for the fix. I have switched to the docker version instead.
Author
Owner

@ohromeo commented on GitHub (May 6, 2026):

ok, thanks, strange still happening on my docker...

Image Image
<!-- gh-comment-id:4388491095 --> @ohromeo commented on GitHub (May 6, 2026): ok, thanks, strange still happening on my docker... <img width="1175" height="535" alt="Image" src="https://github.com/user-attachments/assets/46b312a9-0db0-4993-b81e-a087c730941d" /> <img width="191" height="104" alt="Image" src="https://github.com/user-attachments/assets/c60885ed-ed1a-40d1-a334-375e25196837" />
Author
Owner

@Beginner50 commented on GitHub (May 6, 2026):

Yeah my bad, I may have encountered an issue related to this where the APIs switch to deepseek-reasoning and chat when I restart my laptop.
Restarting the docker services seems to fix this on my end but I agree this is a bandaid solution. Didn't have the time or energy to open up another issue

<!-- gh-comment-id:4388562659 --> @Beginner50 commented on GitHub (May 6, 2026): Yeah my bad, I may have encountered an issue related to this where the APIs switch to deepseek-reasoning and chat when I restart my laptop. Restarting the docker services seems to fix this on my end but I agree this is a bandaid solution. Didn't have the time or energy to open up another issue
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Mintplex-Labs/anything-llm#5135