Channel Configuration API

Configure API channels for AI model access.

Channel Structure

channels:
  - name: openai-main        # Unique name
    baseUrl: https://api.openai.com/v1
    apiKey: ${OPENAI_API_KEY}
    model: gpt-4o
    priority: 1              # Lower = higher priority
    enabled: true

Required Fields

Field	Type	Description
name	string	Unique channel identifier
baseUrl	string	API endpoint URL
apiKey	string	API key (supports env vars)
model	string	Default model name

Optional Fields

Field	Type	Default	Description
priority	number	1	Channel priority
enabled	boolean	true	Enable/disable
maxTokens	number	4096	Max output tokens
temperature	number	0.7	Response creativity
timeout	number	60000	Request timeout (ms)

Built-in Free Channels

Free Channel Notice

The plugin may come with pre-configured free/demo channels (e.g. free Gemini, GLM). These are provided by third-party relay services:

• No availability guarantee: Free channels may stop, throttle, or change at any time
• No stability guarantee: Response speed and quality may be unstable
• For trial use only: Configure your own API Key for stable service
• Security note: Data may pass through third parties; do not send sensitive information

Supported Providers

International

Provider	baseUrl	Features
OpenAI	https://api.openai.com/v1	Chat, vision, tools, embedding, audio
Anthropic Claude	https://api.anthropic.com	Chat, vision, tools, thinking
Google Gemini	https://generativelanguage.googleapis.com	Chat, vision, tools, grounding
xAI Grok	https://api.x.ai/v1	Chat, tools
Mistral AI	https://api.mistral.ai/v1	Chat, embedding, tools
Groq	https://api.groq.com/openai/v1	Chat, tools (ultra-fast)

Chinese Providers

Provider	baseUrl	Features
DeepSeek	https://api.deepseek.com/v1	Chat, tools, reasoning
Zhipu AI	https://open.bigmodel.cn/api/paas/v4	Chat, vision, tools, embedding
Qwen	https://dashscope.aliyuncs.com/compatible-mode/v1	Chat, vision, tools, embedding
Moonshot Kimi	https://api.moonshot.cn/v1	Chat, tools, file

Relay/Aggregation

Provider	baseUrl	Description
OpenRouter	https://openrouter.ai/api/v1	Multi-model aggregation
SiliconFlow	https://api.siliconflow.cn/v1	Chinese aggregation platform
Together AI	https://api.together.xyz/v1	Open-source model hosting

Compatibility

Most OpenAI-compatible APIs can be connected using the openai type — just change baseUrl and apiKey.

Provider Examples

OpenAI

- name: openai
  baseUrl: https://api.openai.com/v1
  apiKey: ${OPENAI_API_KEY}
  model: gpt-4o

Claude

- name: claude
  baseUrl: https://api.anthropic.com
  apiKey: ${ANTHROPIC_API_KEY}
  model: claude-sonnet-4-20250514
  provider: claude

DeepSeek

- name: deepseek
  baseUrl: https://api.deepseek.com/v1
  apiKey: ${DEEPSEEK_API_KEY}
  model: deepseek-chat

Gemini

- name: gemini
  baseUrl: https://generativelanguage.googleapis.com
  apiKey: ${GOOGLE_API_KEY}
  model: gemini-2.5-flash
  provider: gemini

Load Balancing

channelStrategy:
  mode: priority      # priority, round-robin, random
  failover: true      # Auto switch on failure
  maxRetries: 3       # Retry attempts
  retryDelay: 1000    # Delay between retries (ms)

Environment Variables

Security

Use environment variables for API keys:

apiKey: ${OPENAI_API_KEY}

Set in system or .env file:

export OPENAI_API_KEY=sk-xxx
export ANTHROPIC_API_KEY=sk-ant-xxx

Testing Channels

Via Web Panel

1. Go to Channels tab
2. Click Test Connection

Via Command

#ai测试渠道 openai

Next Steps

• Models - Model parameters
• Proxy - Proxy configuration