AI Controller

Introduction

Manages AI embeddings for mediacontainers. This endpoint is normally not used actively by users or custom clients. It is only relevant if you want to connect your own AI to AdmiralCloud and write its embeddings back. Embeddings are stored per mediacontainer and model (text_v1 for text, image_v1 for images) and are the basis for semantic search.

Use create to ingest chunks (text content plus embedding) for a mediacontainer. Each batch carries an origin (e.g. transcript, meta_ai_description): re-ingesting replaces only the chunks with the same origin, so several origins can coexist for one model. Use destroy to remove all embeddings of a mediacontainer for a given model.