Use one base URL for each client. Keep the request format supported by your application, and select an exact model ID available to your API key.

​
Base URLs

  • International: https://infistar.ai/v1
  • Mainland China: https://infistar.cc/v1
  • Use HTTPS for all requests.
Some clients expect the origin only; others expect the /v1 suffix. Follow the instructions for your client rather than adding /v1 twice.

​
Authentication

Send your API key in the HTTP request header: 
Authorization: Bearer YOUR_API_KEY
Keep keys on your server. Do not embed them in public HTML, browser JavaScript, mobile applications or a public repository. An invalid or expired key can produce 401 Unauthorized; see authentication and permissions.

​
Connections and timeouts

  1. Reasoning models can take longer to respond. The existing gateway guidance recommends a client timeout of 60–180 seconds.
  2. Reuse connections with Connection: keep-alive where your client supports it.
  3. Use stream: true when the selected model and client support streaming. This lets users see output as it arrives instead of waiting for the entire response.
The gateway supports browser cross-origin requests for testing. For production, send requests through your own backend so the API key is not exposed.

​
Check connectivity

First call the model list using your own key. Copy an exact model ID returned for that key, then use it in a minimal request: 
curl https://infistar.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "MODEL_ID_FROM_YOUR_MODEL_LIST",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
Replace both placeholders before running the example. Check the model’s supported interface in models and pricing; a model appearing in the list does not mean it supports every API format. See the API overview for other interfaces, or HTTP troubleshooting if the request fails.