Datalumo
Docs · Errors & limits ▾
View as Markdown →

Errors & limits

#HTTP status codes

Status Typical cause
401 Missing, invalid, or expired API key
403 Wrong organisation, missing permission, source not allowed, widget rejected the request, or conversation does not belong to this visitor
404 Unknown organisation, source, page, widget, or conversation
402 No answer credits left (chat and summarize)
422 Invalid input, identity check failed, or keep limit reached
429 Too many requests

#Invalid input

When a required field is missing or invalid, the body looks like this:

{
  "message": "The content field is required.",
  "errors": {
    "content": ["The content field is required."]
  }
}

message is a human-readable summary. errors maps each field name to one or more messages.

#Rate limits

Limits are counted per minute.

#Ingestion API

Counted per API key (or IP if there is no key).

Area Default
Page push and related routes 120 requests per minute

#Widget API

Each group has a limit for the whole widget and a lower limit for a single visitor (widget + IP), so one visitor cannot use up the entire budget.

Group Endpoints Defaults (widget / visitor)
Search search 60 / 20
AI chat, summarize 20 / 6
Events events 120 / 30
Config config, conversations 300 / 60

Exceeding a limit returns 429.

#Answers and billing

  • Search and page ingestion do not use answer credits.
  • Chat and summarize use your plan's monthly answers.
  • When answers are used up, those endpoints return 402.
  • Chat attaches free search hits in data (same hit shape as the search endpoint) so the official embed can show results without a second request. Example:
{
  "error": "answers_exhausted",
  "message": "AI answers are paused for this site — search still works.",
  "fallback": "search",
  "data": [
    { "id": "…", "external_id": null, "name": "…", "url": "…", "snippet": "…", "score": 0.5, "meta": null }
  ],
  "session_id": "…"
}
  • Summarize returns the same error fields without a result list.

#Tips for reliable page sync

  • Always send a stable external_id so retries update the same page instead of creating duplicates.
  • Re-pushing identical content for an already indexed page is skipped.
  • Batch requests accept at most 50 pages at a time.