Skip to content

Troubleshooting en

mars.sven edited this page Jul 12, 2026 · 1 revision

Getting Started

Languages: EN · 中文 · 日本語 · Deutsch · Tiếng Việt · 繁體中文 · Wiki Home

5 steps

  1. Install and launch AIGateway
    Download the package for your platform from Releases and start the app.

  2. Confirm the gateway is running
    Default address: http://127.0.0.1:18080/v1
    Open Gateway in the app and verify the service is running.

  3. Add providers
    Providers → pick a preset (DeepSeek, SiliconFlow, Ollama, Qwen, …) → paste API key → fetch models.

  4. Configure model groups
    Models → review virtual model groups → drag to set failover priority (primary on top, backups below).

  5. One-click tool takeover
    Apps → select ChatGPT / Claude Code / OpenClaw / Harness → Take over.
    After that, switch models/providers only inside AIGateway—no repeated tool config edits or terminal restarts.

Client setup

Scenario Recommendation
After one-click takeover Tools already point at the local gateway; set model to a model group name (e.g. deepseek-chat)
Manual config base_url = http://127.0.0.1:18080/v1, API key may be aigateway
Codex / ChatGPT model_provider = "aigateway", model = group name

Architecture

Tools (ChatGPT / Claude Code / OpenClaw / …)
        │  base_url → http://127.0.0.1:18080/v1
        ▼
   AIGateway (routing / failover / token stats)
        │
        ▼
Upstream providers (DeepSeek / SiliconFlow / Ollama / …)

Next

Clone this wiki locally