Skip to main content
These are optional OpenAI‑style endpoints so existing clients can talk to your gateway. The gateway can expose OpenAI-compatible HTTP endpoints and forward requests to the default AI adapter. See OpenAI-compatible endpoints for usage notes. If you only use Telegram/WhatsApp channels, you can keep these disabled and ignore this section. Note that enabling these endpoints will potentially allow others to send requests to your gateway, so you should only enable them if you are comfortable with that and aware of the security implications. Gateway auth is required by default (even on loopback). HTTP endpoints must be protected with gateway.http.auth_token or the gateway auth token.

​
Enable endpoints

{
  "gateway": {
    "http": {
      "bind": "127.0.0.1",
      "port": 19790,
      "auth_token": "replace-me",
      "endpoints": {
        "openai_chat": true,
        "openresponses": true
      }
    }
  }
}
If auth_token is set, clients must send:
  • Authorization: Bearer <token> or

​
Example

curl http://127.0.0.1:19790/v1/chat/completions \
  -H "Authorization: Bearer replace-me" \
  -H "Content-Type: application/json" \
  -d '{"model":"any","messages":[{"role":"user","content":"hello"}]}'

​
OpenAI Chat Completions

POST /v1/chat/completions Minimal supported fields:
  • model
  • messages (array of { role, content })
  • stream (boolean)
  • user (optional session hint)
The gateway supports stream for clients that expect server-sent events.

​
OpenResponses

POST /v1/responses Minimal supported fields:
  • model
  • input (string or array of { role, content })
  • stream (boolean)
  • user (optional session hint)

​
Sessions

Requests are grouped into sessions so the gateway can keep context. If your client supports the OpenAI user field, providing a stable value can help keep a consistent thread.