Docs / Reference

Errors & Events

HTTP error codes returned by the API and the SSE event types you'll encounter during a streaming inference response.

HTTP error codes

Status	When it occurs	Resolution
200	Inference started successfully. Response body is an SSE stream.	—
400	Malformed request body, unknown `model` ID, invalid message role, or missing required fields.	Check the `error` field in the response body.
401	Missing `Authorization` header, invalid token, revoked token, or expired token.	Re-authorize the app to obtain a new token.
402	The authorized user's SereChat balance is zero or negative.	Ask the user to top up their balance at serechat.com/plan.
429	Rate limit exceeded.	Back off and retry with exponential backoff.
404	`/api/auth/app/request` — `request_id` not found or expired. `/api/auth/app/exchange` — request not in authorized state.	Start a new authorization flow.
410	Code was entered incorrectly 5 times; the request is permanently invalidated.	Start a new authorization flow.
500	Unexpected server error.	Retry after a short delay. If persistent, contact support.

SSE event reference

The inference stream uses named SSE events. Read the current event: line, then parse the JSON in the following data: line.

content

json

event: content
data: { "content": "Hello, world" }

Text chunk from the model. Accumulate all content events to build the complete response.

reasoning

json

event: reasoning
data: { "content": "Let me think..." }

Internal reasoning from models that support extended thinking. Optional to display to end users.

tool_calls

json

event: tool_calls
data: {
  "tool_calls": [
    {
      "id": "call_abc123",
      "type": "function",
      "function": {
        "name": "get_weather",
        "arguments": "{\"city\":\"Amsterdam\"}"
      }
    }
  ]
}

The model wants to call one or more functions. Execute them, then append the results as tool role messages and call the inference endpoint again. The finish event will have "finish_reason": "tool_calls".

finish

json

event: finish
data: {
  "finish_reason": "stop"
}

finish_reason	Meaning
stop	Model finished normally.
tool_calls	Model issued tool call(s). Handle them and continue.
length	Reached `max_tokens` limit.

error

json

event: error
data: { "error": "Inference failed unexpectedly" }

A non-fatal mid-stream error. The stream may still close normally after this. If the error is unrecoverable, a done event follows immediately.

done

json

event: done
data: {}

Always the last event. Close your connection after receiving it.

Authorization errors

Status	Body	Description
400	`app_name required`	Missing or empty `app_name` in /request.
400	`request_id and code required`	Missing fields in /exchange.
400	`Invalid code`	Wrong code entered. Attempts remaining shown in header.
404	`Request not found or expired`	request_id unknown or past its 10-minute window.
409	`Not yet authorized`	Exchange called before the user approved in the browser.
410	`Request invalidated`	Too many wrong code attempts.