Skip to content

GUIDE.md

Latest commit

 

History

History
119 lines (76 loc) · 3.59 KB

GUIDE.md

File metadata and controls

119 lines (76 loc) · 3.59 KB

API Switch User Guide

Quick Start

1. Add a Channel

Go to Channel Management, click Add Channel, and fill in:

Field Description
Name Custom name for easy identification
API Type Select the corresponding provider type
Base URL Upstream API address (no need to include /v1)
API Key Key provided by the provider

After adding, click Fetch Models to get the available model list, then check the models you want to use.

2. Enable Models

Go to API Management and enable the model entries you want to use (green dot = available).

3. Start Using

Set your client's API Base URL to:

http://127.0.0.1:9090/v1

Leave the API Key empty (when access key verification is not enabled).

Use auto as the model name for automatic optimal channel matching, or specify an exact model name.

Recommended Channel Configurations

MiniMax (SiliconFlow)

Field Value
API Type openai or anthropic
Base URL https://api.minimaxi.com or https://api.minimax.chat
API Key Your Key
Add Model Manually MiniMax-M2.7

CODING PLAN

Field Value
API Type openai
Base URL https://api.rcouyi.com
API Key Your Key
Fetch Models Not supported, add manually
Add Model Manually gemini-2.0-flash, gemini-2.5-pro, etc.

The model list API for relay services like this is unavailable, so fetching will fail. After adding your Key in the channel, go directly to API Management and click Add API to manually enter the model name.

FAQ

Proxy Failed to Start: Port in Use

Change the listening port in Settings → Proxy Settings to another port, then restart the proxy.

Request Returns 401

  1. If Enforce Access Key Verification is enabled, the request must include an Access Key: 
    Authorization: Bearer sk-xxx
  2. If not enabled, check that the channel's API Key is correct.

Request Returns "No available provider"

  1. Confirm the channel is enabled.
  2. Confirm the corresponding model entry is enabled (green dot).
  3. Confirm the model is not in cooldown (red dot). Models in cooldown do not participate in routing.

Model Shows Red Dot (Cooldown)

Models that fail requests are automatically cooled down and do not participate in routing during the cooldown period. Default cooldown is 300 seconds, adjustable in Settings → Circuit Breaker. Models recover automatically after cooldown expires.

Failed to Fetch Models

  1. Check that the Base URL is correct (no need to include the /v1 suffix).
  2. Check that the API Key is valid.
  3. Check that the network can access the upstream API.

Claude Channel Configuration

  • Select API type claude
  • Base URL: https://api.anthropic.com
  • Model name must match the upstream, e.g. claude-sonnet-4-20250514

Gemini Channel Configuration

  • Select API type gemini
  • Base URL: https://generativelanguage.googleapis.com
  • API Key: Google AI Studio Key

Azure OpenAI Configuration

  • Select API type azure
  • Base URL: Endpoint address
  • Model name: Deployment Name

Tray Menu Model Order is Incorrect

The tray shows the top 5 enabled AUTO-group entries, sorted by the order in API Management. Drag to adjust the AUTO-group priority in API Management.

Log Shows (auto) Prefix

Indicates that the requested model is auto, and the text after the parentheses is the actual model matched.

This document is continuously updated. If you encounter other issues, feel free to open an Issue.