Learning

Submit map feedback into the learning loop

Learning surface — report a wrong/right/moved verdict on a geocoding answer. Seeds an evidence bundle and a pending entry in the validation-gated promotion queue; a human adjudicates before anything can change an answer. Authenticated; never metered (X-Quota-Cost: 0); limited to 20 submissions per day per key. note is retained for human review only and never reaches analytics.

post/v1/feedback

Request

  • Base URL: https://geo.thinair.co
  • URL: https://geo.thinair.co/v1/feedback
  • Auth: HTTP bearer

Request body

querystring

The original query text this feedback is about.

place_idstring

Stable ThinAir place id the feedback is about.

verdict'wrong' | 'right' | 'moved' required

'wrong' — the answer is incorrect; 'right' — confirms the answer; 'moved' — the place exists but sits at different coordinates (pair with suggested).

notestring

Free-text context for human review (≤280 chars). Stored in the evidence bundle only — never emitted to analytics.

Example request

{
  "query": "Santo Domingo & San Martin, Santiago",
  "place_id": "ta_place_8f3a2c9e61c04d58",
  "suggested": {
    "lat": 29.7604,
    "lon": -95.3698
  }
}

Response

Feedback recorded into the promotion queue.

bundle_idstring required
queue_idstring

Promotion-queue reference. Absent when the queue write soft-failed (see warnings) — never a dangling id.

class_key'feedback_contradiction' | 'feedback_confirmation' required
status'recorded' required
warningsstring[]

Soft-failure notices (e.g. evidence store unavailable). The feedback is still accepted.

Example response

{
  "bundle_id": "eb_01J0000000000000000000000A",
  "queue_id": "pq_01J0000000000000000000000B"
}

Changes

No recorded changes to this operation across all 1 revision of this API.