All use cases
Use case 06Enterprise knowledge base
Ask in plain language. Get an answer that cites its source.
A document library the agent can search, plus the tools your knowledge already lives in.
01How it works
The mechanisms involved.
- A document library
- Upload PDFs, Word documents, spreadsheets and CSV, images, Markdown, HTML, and audio, or save a URL. Scanned PDFs go through OCR, and a document keeps its version history when its content changes.Where the docs say so: A document library
- Tools where knowledge lives
- Confluence and Jira skills install from the marketplace and let the agent search pages and issues through their connections. Google Drive, OneDrive, Dropbox, and Box connect the same way.Where the docs say so: Tools where knowledge lives
- Cited answers
- Ask one document or the whole library and the answer cites the exact passages it came from. Search is semantic when an embedder is ready and falls back to keywords when it isn't.Where the docs say so: Cited answers
- Isolated by tenant
- In a multi-tenant deployment, documents and the vector index are isolated per tenant, and people ask from Slack, Teams, or the Dashboard.Where the docs say so: Isolated by tenant
02Limits
What this doesn't cover.
- 01
Semantic recall needs an embedder. Until a local one runs or an admin consents to a cloud one, search uses keywords.
Source for limit 1 - 02
OCR runs on up to 20 pages per document by default, and low-contrast scans produce noisy text.
Source for limit 2
Run it on your own infrastructure.
Docker Compose, a model key, a channel. The Personal tier is free for individual, non-commercial use.