Overview
The Knowledge Base is where reusable organisational documents live — standards, vendor catalogues, past project deliverables. It is the layer underneath AI Memory: documents go in here, then memories pin them.
Every document is chunked, embedded with pgvector, and made available for semantic search. You can also browse the library directly without going through the AI.
Tags let you scope search and pinning — "DAS", "BMS", "Client A" — without forcing you into a rigid folder hierarchy.
When to use it
- You want past project deliverables searchable by anyone in the organisation.
- You maintain standards or templates that should be reused across projects.
- You need a single place where vendor catalogues live for AI-aided Asset auto-fill.
How it works
1. Upload
Drag documents in. PDF, DOCX, XLSX, PPTX and plain text are extracted automatically. Drawings (DWG, PDF drawings) are indexed by their text layer.
2. Tag
Tag each document with categories that match how you actually look for things. Tags are flat — no nesting — by design.
3. Search
The search box runs semantic search across the library. You can also filter by tag.
4. Pin into memories
From a document, pin into one or more AI Memories so projects can use it.
Tips
- Decide on a tag vocabulary early and write it down somewhere visible. Free-form tagging degenerates fast.
- Re-upload documents instead of editing in place when you make significant changes — the version history matters.
- Big catalogues (200+ pages) are fine; the chunking and embedding handles them. You do not need to split them yourself.