Model Configuration

The models are configured in config/model_config.json.

Specifies the LLMs and embedding models used by the simulator:

{
  "chat": [
    {
      "provider": "openai",                     // LLM provider: OpenAI
      "config_name": "openai-gpt4o",            // Configuration identifier
      "model_name": "gpt-4o",                   // Model name
      "api_key": "sk-xxx",                      // API key
      "generate_args": {
        "temperature": 0                        // Temperature setting (0 = deterministic)
      }
    },
    {
      "provider": "vllm",                       // LLM provider: vLLM for local deployment
      "config_name": "vllm-qwen",               // Configuration identifier
      "model_name": "Qwen2.5-14B-Instruct",     // Model name or path
      "client_args": {
        "base_url": "http://localhost:9889/v1/" // Local API endpoint
      },
      "generate_args": {
        "temperature": 0                        // Temperature setting
      }
    }
  ],
  "embedding": [
    {
      "provider": "vllm",
      "config_name": "embedding-bert",   // Configuration identifier
      "model_name": "bge-base-en-v1.5",         // Embedding model name or path
      "client_args": {
        "base_url": "http://localhost:9890/v1/" // Local API endpoint
      }
    }
  ]
}

Supported Providers

YuLan-OneSim supports the following LLM and embedding providers:

Provider	Description	Chat Models	Embedding Models
openai	OpenAI API (GPT models)	GPT-4, GPT-3.5-turbo	text-embedding-ada-002
aliyun	Alibaba Cloud DashScope API	Qwen series (turbo/plus/max)	text-embedding-v1
deepseek	DeepSeek API	DeepSeek-chat, DeepSeek-coder	DeepSeek-embedding
tencent	Tencent Cloud API	Hunyuan series (lite/standard)	Hunyuan-embedding
ark	ByteDance Ark API	Doubao series (lite/pro)	Doubao-embedding
vllm	Local deployment with vLLM	Llama, Vicuna, ChatGLM	Various open-source models

Field Definitions

Field	Type	Description
chat[].provider	string	Which backend to use ("openai", "vllm", "aliyun", "deepseek", "tencent", "ark")
chat[].config_name	string	Unique key referenced by agents or prompts to select this model
chat[].model_name	string	Model identifier or local path to load balance across providers
chat[].api_key	string	API key for cloud-based providers (e.g., OpenAI, Aliyun, DeepSeek, Tencent, Ark)
chat[].client_args	object	Client arguments (e.g., max_retries, base_url(only for vLLM)) used to customize API request behavior.
chat[].generate_args	object	Inference parameters (e.g., temperature) used to control generation behavior.
embedding[].provider	string	Embedding backend identifier (e.g., "openai")
embedding[].config_name	string	Unique key referenced by the memory manager to select this embedding
embedding[].model_name	string	Embedding model identifier or local path
embedding[].client_args	object	Client arguments (e.g., max_retries, base_url(only for vLLM)) used to customize API request behavior.

Configuration Examples

OpenAI-Only Configuration

With openAI, you can set your model configs like this to use the API

{
  "chat": [
    {
      "provider": "openai",               
      "config_name": "openai-gpt4o", // you may choose any unique strings as identifier, and there's memory config in config.json refering it  
      "model_name": "gpt-4o",    // chat model name, you can check the list on official documentation.         
      "api_key": "sk-xxx",  // put your API key here.              
      "generate_args": {
        "temperature": 0.7    // arguments of generating, you may check it on documentation of API provider.            
      }
    }
  ],
  "embedding": [
    {
      "provider": "openai",
      "config_name": "openai-embedding", 
      "model_name": "text-embedding-3-small",   // embedding model name, you can check the list on official documentation
    }
  ]
}

Aliyun DashScope Configuration

{
  "chat": [
    {
      "provider": "aliyun",
      "config_name": "aliyun-qwen",
      "model_name": "qwen-plus",
      "api_key": "your-dashscope-api-key",
      "generate_args": {
        "temperature": 0.7
      }
    }
  ],
  "embedding": [
    {
      "provider": "aliyun",
      "config_name": "aliyun-embedding",
      "model_name": "text-embedding-v1",
      "api_key": "your-dashscope-api-key"
    }
  ]
}

DeepSeek Configuration

{
  "chat": [
    {
      "provider": "deepseek",
      "config_name": "deepseek-chat",
      "model_name": "deepseek-chat",
      "api_key": "your-deepseek-api-key",
      "generate_args": {
        "temperature": 0.7
      }
    }
  ]
}

Multi-Provider Configuration

{
  "chat": [
    {
      "provider": "openai",
      "config_name": "openai-gpt4o",
      "model_name": "gpt-4o",
      "api_key": "sk-xxx"
    },
    {
      "provider": "aliyun",
      "config_name": "aliyun-qwen",
      "model_name": "qwen-plus",
      "api_key": "your-dashscope-api-key"
    },
    {
      "provider": "vllm",
      "config_name": "local-llama",
      "model_name": "Llama-3-8B-Instruct",
      "client_args": {
        "base_url": "http://localhost:8000/v1/"
      }
    }
  ],
  "embedding": [
    {
      "provider": "openai",
      "config_name": "openai-embedding",
      "model_name": "text-embedding-3-small",
      "api_key": "sk-xxx"
    }
  ]
}