Feedback

POST /api/v1/feedback — rate an image.

Renders are generated, so occasionally one is wrong: the wrong body style, a cropped bumper, a badge that does not exist. Tell us. Feedback is attached to the exact cached asset, and an asset that collects two negative reports is flagged and re-rendered with the next prompt variant — for everyone.

Rate an image

POST/api/v1/feedback

images:readfree

Request body
Parameter	Type	Description
`request_id`	string	The `X-Request-Id` of the image response you are rating. The most precise way to identify the asset.
`vehicle`	object	Alternative to `request_id`: `make`, `model`, `year`, `view` and `color` identifying the rendered variant.
`rating`	integer 1–5	Star rating; 1–2 counts as negative.
`verdict`	"good" \| "bad"	Simpler thumbs up/down when a rating is overkill.
`reason`	string ≤ 500	What is wrong (or right): "shows a sedan, should be a wagon".

Provide request_id or vehicle, and rating or verdict.

bash

curl -X POST https://carimage.dev/api/v1/feedback \
  -H "Authorization: Bearer $CAR_IMAGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"request_id": "req_01j9x…", "verdict": "bad", "reason": "Rear spoiler belongs to the GT3, not the Carrera"}'

201 Created

{
  "data": {
    "asset_id": "3f0c2a7e-…",
    "negative": true,
    "flagged": false,
    "feedback_count": 1
  },
  "request_id": "req_01ja0…"
}

From the CLI and MCP

car-image feedback --request-id req_… --bad --reason "…"
MCP tool rate_image — agents can rate what they just fetched.
The dashboard shows your recent images with one-click ratings.

Feedback is free and appreciated

It is never billed, it does not count toward your rate limit budget in any meaningful way, and it is the fastest path to a better render for the next person who asks for the same car.

Errors

400 — neither request_id nor vehicle given, or neither rating nor verdict; reason too long.
404 — no rendered asset matches the identifier (feedback must target an image that exists).