5 retrievers
Every listing takes the same three calls: find it, read its input schema and example, run it. Anonymous calls are limited to 10 requests a minute per listing. Image and face listings can answer 503 on the first call after idle, so retry once.
# list every retriever
curl https://api.mixpeek.com/v1/marketplace/catalog
# read one: input schema and an example request
curl https://api.mixpeek.com/v1/marketplace/catalog/trademark-slogan-match
# run it with that example
curl -X POST https://api.mixpeek.com/v1/marketplace/catalog/trademark-slogan-match/execute \
-H "Content-Type: application/json" \
-d '{"inputs": {"query": "Just do it"}, "limit": 5}'5 retrievers
Paste a page's text, a transcript or a short description of a video. Get the IAB Content Taxonomy categories it fits, across all four tiers, ranked by score.
Upload a photo. Get every public figure it matches, with a confidence score.
Upload a frame. Get the reference marks that look most like it, ranked, for you to review.
Paste a line of copy. Get the registered slogans it resembles and who owns them.
One frame in. Every public figure and brand mark it contains, scored.
A check Mixpeek runs for you against reference data it maintains, such as registered logos, trademarked slogans and public figures. You send one input and get back what it matches, scored.
Every listing has a free tier and takes anonymous calls, so a person or an agent can try it in one request.
The extractors behind each listing are the ones you can run on your own files in your own namespace.
Features are extracted at ingest, so a query is a search over finished data.
Looking for search demos over sample datasets, like trailers and artworks? See the showcase.