FEVER is an appliance made of multiple systems, deployed easily. It functions as one multimodal media database: images, video, audio in one place.
With FEVER you or your users can search media items by content using natural language, find duplicates and near duplicates, curate training data for your own AI and more.
Tenancy and isolation are easy to manage.
API Concepts
blob - A single indexed media item addressed by blob_id
media kinds - image / video / audio; GET /v1/media/formats to programmatically see what FEVER ingests
job - Ingest and export are asynchronous, so poll GET /v1/jobs/{job_id} to check status
customer_id - YOUR end user's account id to associate ownership of the item
API keys - Two roles, admin or scoped, both use bearer auth
datasets / dataset_items - Collate assets into datasets, queryable over SQL and exportable
search - Text or image queries in, ranked SearchHits out that tell you how strong the match is
duplicates - Duplicate and near duplicate clusters
export - Write your tagged dataset to S3 for training your own model
query - Read only SQL console with fever_search() for easy semantic ranking
