Skip to navigation

Guides

The platform works in three layers: you upload content as assets, build a knowledge store from that content, and query it through natural language, either by generating a response or by searching.

┌─────────────────────────────────┐
│         Your Application        │  ← You build this
├─────────────────────────────────┤
│   Responses API and Search      │  ← Query with natural language
├─────────────────────────────────┤
│     Knowledge Store             │  ← Videos and images + derived understanding
├─────────────────────────────────┤
│     Assets                      │  ← Video, audio, images
└─────────────────────────────────┘

Use these guides to set up a knowledge store, then generate responses or search it.

Set up a knowledge store

1. Upload content

Upload videos and images to the platform. Choose between direct and multipart uploads.

2. Create a knowledge store

Create a knowledge store. You can also configure what the platform extracts.

3. Add assets to a knowledge store

Add assets to a knowledge store for indexing.

Query a knowledge store

Create a response

Generate a response from a knowledge store, then stream it, structure the output, or continue with multi-turn sessions. Use this when you want Jockey to interpret and explain, including agentic search.

Search a knowledge store

Search a knowledge store with a natural-language query and receive ranked video clips and images. Use this when you want direct retrieval without agentic reasoning.