[Bug]: Deepseek-reasoner model fails after first message in LibreChat when added to default models

[kihyunnn]

What happened?

When using LibreChat with the Deepseek API specifically with the "deepseek-reasoner" model added to the default models configuration, the conversation only works for the first message. Subsequent messages fail with a system message error. Other Deepseek models (deepseek-chat, deepseek-coder) are not affected by this issue.

Configuration where issue occurs:

default: ["deepseek-chat", "deepseek-reasoner", "deepseek-coder"]

The error message received is:

2025-01-21 01:58:37 LibreChat         | 2025-01-20 16:58:37 error: 
2025-01-21 01:58:37 LibreChat         | 2025-01-20 16:58:37 error: [handleAbortError] AI response error; aborting request: 400 The system message of deepseek-reasoner must be put on the beginning of the message sequence.

Also tell us, what did you expect to happen?
Expected the deepseek-reasoner model to maintain continuous conversation like the other Deepseek models (deepseek-chat, deepseek-coder), without any system message errors.

Steps to Reproduce

1. Configure LibreChat with deepseek-reasoner in the default models array
2. Start a new conversation using the deepseek-reasoner model
3. Send first message (works correctly)
4. Attempt to send a second message
5. Observe the system message error

What browsers are you seeing the problem on?

Chrome

Relevant log output

2025-01-21 01:58:37 LibreChat         | 2025-01-20 16:58:37 warn: [OpenAIClient.chatCompletion][stream] API error
2025-01-21 01:58:37 LibreChat         | 2025-01-20 16:58:37 error: 
2025-01-21 01:58:37 LibreChat         | 2025-01-20 16:58:37 error: [handleAbortError] AI response error; aborting request: 400 The system message of deepseek-reasoner must be put on the beginning of the message sequence.

Screenshots

Code of Conduct

• I agree to follow this project's Code of Conduct