DeepSeek - Complete API Reference

##All APIs require Bearer Token authentication## **Get API Key:** Visit [API Key Management Page](https://starmagic.ai/app/api-keys) to get your API Key **Add to request header:** ``` Authorization: Bearer YOUR_API_KEY ```

Authorization: Bearer YOUR_API_KEY

Chat model name - `deepseek-chat`: General conversation model - `deepseek-reasoner`: Deep reasoning model, excels at math, coding and complex logical reasoning **Note**: `deepseek-reasoner` does not support `temperature`, `top_p`, `tools`, `tool_choice`, `response_format` parameters. Passing these will be rejected by upstream

"deepseek-chat"

Conversation message list, supports multi-turn conversation Different roles have different field structures, select the corresponding role to view

[
  null
]

Thinking mode control (Beta) **Details**: - Controls the deep thinking feature of `deepseek-reasoner` model - When enabled, the model will perform deep reasoning before responding

{
  "type": "enabled"
}

Frequency penalty parameter to reduce repetitive content **Details**: - Positive values penalize tokens based on their frequency in the generated text - Higher values make it less likely to repeat existing content - Default: 0 (no penalty)

0

Maximum number of tokens to generate **Details**: - The model will stop generating when this limit is reached - If not set, the model decides the generation length

4096

Presence penalty parameter to encourage new topics **Details**: - Positive values penalize tokens based on whether they have appeared in the text - Higher values encourage discussing new topics - Default: 0 (no penalty)

0

Specify response format **Details**: - Set to `{"type": "json_object"}` to enable JSON mode - In JSON mode, the model will output valid JSON content

{
  "type": "text"
}

Stop sequences. The model will stop generating when encountering these strings **Details**: - Can be a single string or an array of strings - Maximum 16 stop sequences

Whether to stream the response - `true`: Stream via SSE (Server-Sent Events), returning content in real-time chunks - `false`: Wait for the complete response before returning

false

Streaming response options Only effective when `stream=true`

{
  "include_usage": true
}

Sampling temperature, controls output randomness **Details**: - Lower values (e.g. 0.2): More deterministic, focused output - Higher values (e.g. 1.5): More random, creative output - Default: 1

1

Nucleus Sampling parameter **Details**: - Controls sampling from tokens whose cumulative probability reaches the threshold - For example, 0.9 means sampling from tokens reaching 90% cumulative probability - Default: 1.0 (consider all tokens) **Tip**: Avoid adjusting both temperature and top_p simultaneously

1

Tool definition list for Function Calling **Details**: - Maximum 128 tool definitions - Each tool requires a name, description and parameter schema

[
  {
    "type": "function",
    "function": {
      "name": "string",
      "description": "string",
      "parameters": {},
      "strict": false
    }
  }
]

Controls tool calling behavior **Options**: - `none`: Do not call any tools - `auto`: Model decides whether to call tools - `required`: Force the model to call one or more tools **Default**: `none` when no tools provided, `auto` when tools are provided

Whether to return token log probabilities **Details**: - When set to `true`, the response will include log probability information for each token

false

Return log probabilities of the top N most likely tokens **Details**: - Requires `logprobs` to be set to `true` - Range: `[0, 20]`

0