What is Knowledge Base?
Knowledge Base — In the context of AI development, a knowledge base is a structured collection of documents, patterns, and rules that AI agents can search and reference during their work. Unlike general training data, a knowledge base contains curated, project-specific information — architecture decisions, coding patterns, API documentation — that keeps AI responses accurate and consistent with your codebase.
Knowledge Bases for AI Agents
Traditional knowledge bases store human-readable documentation. AI-native knowledge bases add:
- Vector embeddings for semantic search
- Structured metadata (tags, categories, confidence scores)
- Versioning to track how knowledge evolves
- Access control to scope what each agent can see
Quoth's Knowledge Base Architecture
Quoth provides two levels of knowledge management:
Cloud (SaaS): Shared knowledge base on Neon Postgres with pgvector. Supports multi-project, multi-organization access with role-based permissions. Documents are chunked and embedded for semantic search.
Local (Plugin): SQLite database with HNSW index storing learned patterns from development sessions. Patterns include confidence scores (Bayesian Beta distribution), usage counts, and source attribution.
The two levels sync through the promotion pipeline: high-confidence local patterns (>0.8 confidence, >10 uses) are automatically promoted to the cloud knowledge base for sharing across your team.