upstream_timeout

A dependency accepted the request and then did not answer in time.

• A model provider was slow past our ceiling.
• A connected tool endpoint of yours did not respond — tool calls have their own timeout.

• Retry with backoff.
• If it is your own tool endpoint, check its latency. Tool calls are made inside a visitor's turn, so a slow endpoint is a slow answer.

{
  "error": {
    "code": "upstream_timeout",
    "message": "A dependency accepted the request and then did not answer in time",
    "request_id": "7cb7f7862a82425d8e4c3fb8a497dfe6"
  },
  "type": "https://integrable.cloud/docs/errors/upstream_timeout",
  "title": "Upstream Timeout",
  "status": 504,
  "detail": "A dependency accepted the request and then did not answer in time",
  "instance": "/api/bots/01a0652b-3713-7ea1-a6c9-2e895389ec34"
}

Branch on error.code, not on the message — the code is stable, the sentence is not. Log request_id either way.

Codes that are easy to confuse with this one, or that show up alongside it:

• upstream_error

Quote the request_id from the response — hello@integrable.cloud. It is what lets us find the exact request. The full list of codes is at Error codes, and the conventions every endpoint shares are in Retries, versioning and limits.