Skip to main content

Knowledge Base

The Knowledge Base stores information that AI assistants can access across sessions. It's persistent memory for your AI tools.

Why Use Knowledge Base?

AI assistants typically forget everything between sessions. Every time you start a new chat, you have to re-explain:

  • Project architecture
  • Coding conventions
  • Business requirements
  • Past decisions and their reasoning

The Knowledge Base solves this by storing context that AI can retrieve automatically.

Adding Knowledge

From the Web App

  1. Go to Knowledge Base in your project
  2. Click "Add Item"
  3. Enter a key (e.g., "architecture")
  4. Add your content
  5. Click "Save"

From AI

Ask your AI assistant:

"Save this to the knowledge base: Our API uses REST with JSON responses. Authentication is via Bearer tokens."

From CLI

erold kb set architecture "
Our app uses a microservices architecture:
- API Gateway: Kong
- Services: Node.js
- Database: PostgreSQL
- Cache: Redis
"

Knowledge Keys

Keys are identifiers for knowledge items. Use descriptive, hierarchical keys:

  • architecture - Overall system design
  • coding/style - Code style guidelines
  • api/auth - Authentication details
  • decisions/database - Why we chose PostgreSQL

Searching Knowledge

AI assistants automatically search the knowledge base when relevant. You can also explicitly ask: 

"Search the knowledge base for authentication"

Or via CLI:

yet kb search "authentication"

Best Practices

What to Store

  • Project context - Goals, scope, stakeholders
  • Architecture decisions - Why you chose certain technologies
  • Coding standards - Style guides, patterns to use/avoid
  • API documentation - Endpoint summaries, auth flows
  • Business rules - Domain logic that affects implementation

Keep It Concise

AI context windows have limits. Store summaries and key points rather than entire documents.

Update Regularly

When decisions change, update the knowledge base. Outdated information is worse than no information.

Example Knowledge Items

# Key: coding/style
- Use TypeScript strict mode
- Prefer const over let
- Use async/await over callbacks
- Maximum function length: 50 lines
- Always add JSDoc for public functions

# Key: api/endpoints
- GET /users - List users (paginated)
- POST /users - Create user
- GET /users/:id - Get user details
- All responses use { data, meta, error } format

# Key: decisions/database
We chose PostgreSQL over MongoDB because:
1. Strong ACID compliance needed for transactions
2. Complex queries with joins
3. Team has more SQL experience

Plan Limits

Knowledge base items are unlimited on all plans. Storage limits apply:

Plan Storage
Free 100MB
Pro 2GB included, then $0.05/GB/month

Next Steps