TypeGraph Documentation
TypeGraph is a unified RAG pipeline SDK that handles chunking, embedding, storing, and retrieving context across your business data. It combines semantic search, BM25, graph traversal, recency scoring, and agent memory into a single search interface with built-in tenant isolation and context-aware access control.
Core Concepts
The building blocks of a TypeGraph pipeline.
Buckets
Named collections that organize indexed documents and event artifacts with configurable chunking and extraction.
Documents
Durable long-form content with names, descriptions, metadata, chunks, and embeddings.
Events
Time-anchored business occurrences with participants and linked documents.
Search
Unified retrieval across semantic, BM25, graph, recency, and memory resources with rank fusion.
Memory
Persistent storage for facts, preferences, and thread history scoped through context.
Graph
Automatic entity and relationship extraction that powers graph-augmented retrieval.
Policy
Governance rules that control allowed SDK actions.
Popular Guides
Hands-on tutorials organized by use case.