Knowledge Base

The Knowledge Base lets you save documents, web pages, and notes alongside your meeting transcriptions. All content is indexed for AI-powered semantic search, so your AI assistant can find relevant information across everything you've saved.

What You Can Store

• Web Pages - Save articles, documentation, or any web content
• Documents - Store text documents and notes
• Meeting Transcriptions - Automatically indexed from your meetings
• Bookmarks - Quick references to important URLs
• Notes - Quick thoughts and reminders

How It Works

When you save content to the Knowledge Base:

1. Content is cleaned - HTML is stripped, text is extracted
2. Text is chunked - Content is split into ~500 token segments with smart sentence boundaries
3. AI embeddings are generated - Each chunk gets a semantic fingerprint using OpenAI
4. Stored for search - Content becomes searchable by meaning, not just keywords

AI-Powered Search

Unlike traditional keyword search, the Knowledge Base uses semantic search. This means:

• Search by meaning, not exact words
• Find related content even with different terminology
• Get relevant results ranked by similarity
• Search across all your content with a single query

Example

If you saved an article about "machine learning optimization" and search for "how to improve AI model performance", the system understands these are related and returns the article.

Project Context

Content can be associated with specific projects. When chatting with your AI assistant about a project, it automatically has access to:

• Meeting transcriptions for that project
• Documents and resources you've saved to that project
• Web pages you've bookmarked for reference

This gives your AI assistant the context it needs to provide relevant, informed responses.

Content Types

Managing Your Knowledge Base

Adding Content

Content can be added through:

• Saving web pages directly
• Uploading documents
• Automatic indexing of meeting transcriptions
• Creating notes and bookmarks

Organizing Content

• Tags - Add tags to organize and filter content
• Projects - Associate content with specific projects
• Workspaces - Keep work and personal content separate

Archiving

Content you no longer need can be archived:

• Archived content is hidden from normal searches
• Can be restored at any time
• Permanently delete when you're sure

Privacy

Your Knowledge Base content is:

• Private to your account
• Only searchable by you and your AI assistant
• Never shared with other users
• Scoped to your projects and workspaces
• Stored securely with encryption at rest

Getting Started

Step 1: Access the Knowledge Base

1. Open the sidebar menu
2. Expand the Tools section
3. Click Knowledge Base

Or navigate directly to: /w/{your-workspace}/knowledge-base

Step 2: Index Your Meeting Transcriptions

If you have existing meeting transcriptions:

1. Look for the yellow alert: "Transcriptions need indexing"
2. Click Index Now to process them
3. Wait for completion (processes 20 at a time)
4. Repeat if more remain pending

Step 3: Add Your First Resource

1. Click Add Resource (top right)
2. Choose a type:
   • Web Page - Articles, documentation with URLs
   • Document - Text documents without URLs
   • Note - Quick thoughts and reminders
   • Bookmark - URL references for later
3. Fill in the title and content
4. Click Add Resource

Your content is automatically processed and indexed for search.

Step 4: Test Semantic Search

1. Click the Search tab
2. Type a query (at least 3 characters)
3. View results ranked by relevance

Try these searches:

• Topics from your recent meetings
• Concepts rather than exact words
• Questions like "how to deploy" or "what's the budget"

Tips for Best Results

• Be specific: "Q4 marketing budget" works better than just "budget"
• Use natural language: Search like you'd ask a question
• Add context: More words help find better matches
• Check similarity: Higher percentages mean closer matches

Technical Details

For developers and power users:

• Embedding Model: OpenAI text-embedding-3-small (1536 dimensions)
• Vector Database: pgvector on PostgreSQL
• Chunking: ~500 tokens with sentence boundary awareness and overlap
• Search: Cosine similarity ranking