Errors & troubleshooting - LangSearch

Check the HTTP status before reading search results. Error responses include a log_id when available; retain it when reporting a failed request.

{
  "code": "400",
  "log_id": "example-request-id",
  "message": "count must not exceed 50"
}

Status	Meaning	What to do
`400`	Invalid request	Check JSON, a non-empty query, and `count` ≤ 50.
`401`	Authentication failed	Check your Bearer token and that the key is active.
`403`	Access denied	Check account access in the Dashboard.
`429`	Rate or daily allowance exceeded	Reduce bursts; check usage and the UTC reset time.
`500`	Server error	Retry a limited number of times with backoff.
`502`	Search provider failure	Retry with backoff; keep the `log_id` if it persists.
`503` / `504`	Unavailable or timeout	Allow time before a bounded retry.

Read the error safely

Depending on the response path, the error text can appear in message or msg, and code can be a string or number. Use HTTP status as the primary success check.

Retry policy

Retry transient failures with an increasing delay and jitter, and set a maximum attempt count. A daily quota error requires waiting until the next automatic reset; repeated immediate retries will not restore allowance. When reporting a problem, include the HTTP status, time in UTC, and log_id. Do not share your API key.