Knowledge overview

Knowledge is your tenant's library of documents — manuals, procedures, specifications, internal wikis — that the on-glasses AI can search inside a live training session.

What knowledge is

A knowledge document is a file you've uploaded to your tenant. After upload, TrainAR processes it (extracts text, chunks it, generates embeddings) so the AI can semantically search inside it during a session. When an engineer asks a question, the AI calls the built-in search_knowledge skill to find relevant chunks and answers based on what it finds.

What knowledge is NOT

• Knowledge isn't skills. Skills are actions the AI takes (call a webhook, return a result). Knowledge is content the AI reads. The AI uses a skill (search_knowledge) to access knowledge — but knowledge itself is just text + image content.
• Knowledge isn't bundled MCP tools. MCP Tools come from third-party MCP servers you've connected. They're callable tools, not documents.
• Knowledge isn't recordings. Session recordings are stored under Sessions, not Knowledge — they're operational records, not reference material. See Sessions overview.

Where you manage knowledge

Dashboard → Skills & Knowledge → Knowledge tab.

The Knowledge tab uses a folder-and-collection explorer:

• Folders on the left — organisational structure you define for navigation
• Collections + documents in the main area for the selected folder

You can:

• Upload new documents — see Uploading knowledge
• Create collections to group related documents — see Knowledge collections
• Re-process a document if its processing status is wrong
• Delete a document permanently (also removes its chunks + embeddings)

How a session uses knowledge

When a session starts, the AI receives a manifest of all knowledge available to that seat. The manifest lists each document's name, its collection (if any), and a short auto-generated summary.

When the engineer asks a question, the AI:

1. Decides whether to call search_knowledge based on the manifest signals
2. If yes, sends the engineer's question to the vector search
3. Gets back the top chunks ranked by semantic similarity
4. Synthesises a spoken response, quoting or paraphrasing the relevant content

For the full walkthrough, see How the AI uses knowledge.

Where to next

• Uploading knowledge — formats, sizes, processing
• Knowledge collections — group documents for better retrieval
• How the AI uses knowledge — manifest + vector search
• Knowledge processing pipeline — what happens under the hood