Search Gen 2
Eluvio search is organized around a search index, which is identified by a content object in the Content Fabric. Access control is governed by having access to that content object.
The Content Search API covers the whole lifecycle: create an index over a set of contents, pick the features to search on, and query it. By default, shot detection is run such that clips align to shot boundaries but other tagger models can be selected to generate the clip boundaries.
See Model Listing API for a list of available features to index. Selecting these to add to your index will automatically tag all contents in the index and make the tags searchable. You may also use any of the listed models to define the aggregation strategy.
The following services run underneath and are only needed for advanced cases, such as supplying your own tags to search:
| Service | What it is | API |
|---|---|---|
| Tagger | Runs AI models over content to produce tags. | Eluvio Tagger API |
| Tagstore | Stores tags and tracks. Tags can be written here directly instead of being generated by a model. | Tagstore API |
| Vectorstore | Low-level REST interface to the vector database: CRUD for collections, indexes, vectors, and documents, plus semantic vector search on top of keyword filters. For clip search use cases it is not recommended to touch this API. | Vectorstore API |
Next steps
- Getting Started — create an index and run a search.
- Search Configuration — features, aggregation, and search defaults.