Skip to content

qwen-code.md

Latest commit

 

History

History
105 lines (81 loc) · 3 KB

qwen-code.md

File metadata and controls

105 lines (81 loc) · 3 KB
Raw

Qwen Code

Connect Qwen Code to go-llm-proxy to use your self-hosted or third-party models.

Quick start

Use the built-in config generator (--serve-config-generator). Select Qwen Code from the dropdown, choose a default model and any additional models, and generate a settings.json config file.

How it works

Qwen Code supports multiple model providers with both OpenAI and Anthropic protocols. The proxy is configured as a provider with per-model entries. Qwen Code's /model command lets users switch between available models at runtime.

Configuration file

Save as ~/.qwen/settings.json:

{
  "$version": 3,
  "model": { "name": "MiniMax-M2.5" },
  "security": { "auth": { "selectedType": "openai" } },
  "modelProviders": {
    "openai": [
      {
        "id": "MiniMax-M2.5",
        "name": "MiniMax M2.5",
        "envKey": "PROXY_API_KEY",
        "baseUrl": "https://your-proxy.example.com/v1",
        "generationConfig": { "timeout": 300000, "maxRetries": 1 }
      },
      {
        "id": "qwen-3.5",
        "name": "Qwen 3.5",
        "envKey": "PROXY_API_KEY",
        "baseUrl": "https://your-proxy.example.com/v1",
        "generationConfig": { "timeout": 300000, "maxRetries": 1 }
      }
    ],
    "anthropic": [
      {
        "id": "claude-sonnet-4-20250514",
        "name": "Claude Sonnet 4",
        "envKey": "PROXY_API_KEY",
        "baseUrl": "https://your-proxy.example.com/anthropic",
        "generationConfig": { "timeout": 300000, "maxRetries": 1 }
      }
    ]
  },
  "env": {
    "PROXY_API_KEY": "your-proxy-api-key"
  }
}

Key settings

Field Purpose
model.name Default model on startup
security.auth.selectedType Auth protocol for the default model ("openai" or "anthropic")
modelProviders.openai[] Models using OpenAI protocol (base URL ends with /v1)
modelProviders.anthropic[] Models using Anthropic protocol (base URL ends with /anthropic)
env.PROXY_API_KEY Your proxy API key (referenced by envKey on each model)

Dual-protocol support

Models are grouped by protocol:

  • OpenAI models go in modelProviders.openai with baseUrl pointing to /v1
  • Anthropic models go in modelProviders.anthropic with baseUrl pointing to /anthropic

The config generator handles this automatically based on each model's protocol.

Switching models

After launching Qwen Code, use the /model command to switch between configured models at runtime. All models listed in modelProviders are available.

Web search

The config generator optionally includes Tavily web search:

{
  "webSearch": {
    "provider": [{ "type": "tavily", "apiKey": "tvly-your-key" }],
    "default": "tavily"
  }
}

Installation

macOS / Linux

mkdir -p ~/.qwen
# Save generated settings.json to ~/.qwen/settings.json

Windows

mkdir %USERPROFILE%\.qwen
REM Save generated settings.json to %USERPROFILE%\.qwen\settings.json