Chat Completions Proxy

OpenAI-compatible chat completions proxy. Both paths use the same handler and always target the default AI provider. Only POST is allowed.

POST /api/ai/chat/completions

POST /api/ai/v1/chat/completions

Send an OpenAI-style JSON body.

REQUEST Field	Value Type	Description
`messages`	array	Chat messages with roles such as `system`, `user`, `assistant`, and `tool`.
`stream`	boolean	If `true`, the response is streamed as SSE `text/event-stream`.
`tool_choice`	string	If set to `required`, tool execution is forced.
`tools`	array	Client tools are not forwarded to the upstream provider. If omitted or empty, all server tools remain available. If the list contains the `connection_test` tool, only that server tool is used.
`timeout_ms`	number	Timeout in milliseconds. Default: `120000`.

Non-streaming responses use an OpenAI chat.completion shape with fields such as id, object, created, model, and choices. Streaming responses send chunk events and end with data: [DONE]. Additional SSE events may carry tool progress.

If a custom default provider URL points back to this proxy endpoint, the request returns 409 Conflict. Missing default provider returns 404. Timeout returns 504. Provider failures return 502. Wrong methods return 405.

Example Request

# Request
POST /api/ai/chat/completions HTTP/1.1
Host: helpdesk.example.com
Authorization: Bearer <access_token>
Content-Type: application/json
 
{
  "messages": [
    {
      "role": "user",
      "content": "Hello"
    }
  ],
  "stream": false,
  "timeout_ms": 120000
}
 
# Response
HTTP/1.1 200 OK
Content-Type: application/json
 
{
  "id": "chatcmpl-example",
  "object": "chat.completion",
  "created": 1710000000,
  "model": "inet-ai-proxy",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello! How can I help?"
      },
      "finish_reason": "stop"
    }
  ]
}

Note: POST /api/ai/v1/chat/completions behaves identically to POST /api/ai/chat/completions.