DeepSeek Configuration

DeepSeek

https://platform.deepseek.com/

Configurable Parameter List

Parameter Name	Parameter Description	Default Value
api_base	API base address	https://api.deepseek.com
key	DeepSeek API key	sk-AAAAAAAAAAAAA
model	Large model name: deepseek-reasoner, deepseek-chat	deepseek-reasoner
max_tokens	Maximum input tokens	400
temperature	Use sampling temperature to control the randomness of the output. The value ranges from 0 to 2. A higher temperature (e.g., 0.8) will make the output more random and produce unexpected results. A lower temperature (e.g., 0.2) will make the output more focused and closer to the given input content	0.7
read_thought_enabled	Whether to read out the thinking text	true

Configuration Example

roles.json

{
 "1": {  
     "start_text": "Hello, I'm Bunny. How can I help you?",
     "prompt": "You play the role of a child's buddy named Bunny. You are kind, lively, and lovable. You are always full of love for children, and you often praise and encourage them. Provide interesting and innovative answers in language that a 5-year-old can easily understand. Ask for her opinion on the chat topic each time to stimulate her thinking and curiosity.",
     "llm_type": "deepseek",
     "llm_config": {
         "api_base": "https://api.deepseek.com",
         "key": "sk-TESTAAAAAAAAAAAAa",
         "model": "deepseek-reasoner",
         "max_tokens": 1024,
         "temperature": 0.7,
         "read_thought_enabled": true
     }
 }
}

Or

docker-compose.yml

 DEEPSEEK_API_BASE: "https://api.deepseek.com"
 DEEPSEEK_MODEL: "deepseek-reasoner"
 DEEPSEEK_MAX_TOKENS: 1024
 DEEPSEEK_TEMPERATURE: 1.0
 DEEPSEEK_READ_THOUGHT_ENABLED: "true"