How to setup librechat with DeepSeek-R1 API endpoint

[AngusHsu-cubewise]

I'm trying to integrate a DeepSeek-R1 API endpoint with LibreChat, but encountering compatibility issues with non-standard response formats and custom parameters that aren't part of the OpenAI Chat Completions specification.

Current Configuration

 custom:
    - name: "Deepseek"
      apiKey: "2025_pa_test_token"
      baseURL: "http://116.1.202.42:5013/v1"
      models:
        default: ["default"]
        fetch: false
      titleConvo: true
      titleModel: "Deepseek/Deepseek-r1"
      dropParams: ["stop"]
      addParams:
        temperature: 0.7
        top_p: 0.8
        max_tokens: 2048
        stream: false
      headers:
        Content-Type: "application/json"

API Behavior

The DeepSeek-R1 API works with direct curl requests but has compatibility issues with LibreChat:
Working curl request:
bashcurl -X POST "http://116.1.202.42:5013/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer 2025_pa_test_token" \
  -d '{
    "messages": [
        {
            "role": "user",
            "content": "Hi, who are you?"
        }
    ],
    "temperature": 0.7,
    "reasoning": false,
    "top_p": 0.8,
    "stop": [
        "[End]"
    ],
    "max_tokens": 2048
}'

API Response:

{
  "object": "chat.completion",
  "id": "21b0845e17538634791297856e3477",
  "choices": [{
    "index": 0,
    "message": {
      "content": "<think>\nI'm DeepSeek-R1, an AI assistant created exclusively by the Chinese Company DeepSeek. I'll do my best to help you.\n</think>\n\nI'm DeepSeek-R1, an AI assistant created exclusively by the Chinese Company DeepSeek. I'll do my best to help you.",
      "role": "assistant"
    },
    "finish_reason": "stop"
  }],
  "created": 1753863479,
  "model": "DeepSeek-R1",
  "usage": {
    "completion_tokens": 63,
    "prompt_tokens": 11,
    "total_tokens": 74,
    "prompt_tokens_details": {
      "cached_tokens": 0
    }
  }
}

Request for Assistance

I would be incredibly grateful if anyone from the LibreChat team or the wonderful community could help me:

1. Refine my configuration to properly integrate this DeepSeek-R1 endpoint
2. Find an elegant solution for the reasoning trace display challenge
3. Confirm the proper approach for handling custom API parameters
4. Share any debugging tips or best practices for custom endpoint integration

I really appreciate the amazing work you've all done with LibreChat, and I understand that supporting every possible API variation can be challenging. Any guidance, configuration examples, or documentation pointers would be tremendously helpful - not just for me, but potentially for other community members interested in integrating similar reasoning-capable models.

Environment Details

LibreChat version: [LibreChat v0.7.9-rc1]
API: DeepSeek-R1 compatible endpoint