How ruchita17697/graphrag works
TigerGraph GraphRAG combines a knowledge-graph builder with graph-powered Q&A. The user-facing flow is document upload or cloud ingestion, chunking and LLM-based entity/relationship extraction, then graph and embedding storage; questions enter through the UI or API and use schema-aware query generation and retrieval to produce answers. Separate capabilities include classic and agentic chat, MCP tools, administration, trace logs, and eventual-consistency rebuilds. Several subsystem links are documented in the README; where sampled source does not establish call ownership, links are omitted or marked unverified.
Main entry points: tigergraph_embedding_store.py TigerGraph graph and vector store · main.py API application · supportai_ingest.py Document ingestion · main.tsx Chat and Admin UI
Components 21
User interface
- Chat and Admin UI
graphrag-ui/src/main.tsx
API services
- API application
graphrag/app/main.py - API routes
graphrag/app/routers
Knowledge construction
- Document ingestion
graphrag/app/supportai/supportai_ingest.py - Chunking strategies
common/chunkers - KG extraction
common/extractors/LLMEntityRelationshipExtractor.py - Consistency and rebuild
ecc/app/main.py - Graph processing workers
ecc/app/graphrag/workers.py
Question answering
- Query orchestration
graphrag/app/routers/queryai.py - Support chat
graphrag/app/supportai/supportai.py - Graph retrievers
graphrag/app/supportai/retrievers - Query tools
graphrag/app/tools - Agentic chat
graphrag/app/agent - MCP integrations
graphrag/app/mcp_addons
Shared platform
- Runtime configuration
common/config.py - LLM providers
common/llm_services - Embedding services
common/embeddings/embedding_services.py - TigerGraph graph and vector store
common/embeddings/tigergraph_embedding_store.py - TigerGraph connections and schema
common/db/connections.py - Logging and metrics
common/logs/logwriter.py
Other components
- User
Connections 26
- User → Chat and Admin UI: uses
README.md - User → API application: calls API
README.md - Chat and Admin UI → API application: calls API
- API application → API routes: registers routes
graphrag/app/main.py - API routes → Query orchestration: routes questions
graphrag/app/routers/__init__.py - API routes → Support chat: routes support
graphrag/app/routers/__init__.py - Chat and Admin UI → Document ingestion: submits documents
README.md - Document ingestion → Chunking strategies: chunks documents
graphrag/app/supportai/supportai_ingest.py - Document ingestion → KG extraction: extracts graph
graphrag/app/supportai/supportai_ingest.py - KG extraction → LLM providers: calls model
common/extractors/LLMEntityRelationshipExtractor.py - Document ingestion → TigerGraph graph and vector store: writes graph data
README.md - Consistency and rebuild → Graph processing workers: starts processing
ecc/app/main.py - Graph processing workers → KG extraction: uses extractor
ecc/app/graphrag/workers.py - Graph processing workers → TigerGraph graph and vector store: updates embeddings
ecc/app/graphrag/workers.py - Query orchestration → Query tools: uses query tools
graphrag/app/routers/queryai.py - Support chat → Graph retrievers: retrieves context
README.md - Graph retrievers → TigerGraph graph and vector store: searches graph
README.md - Graph retrievers → Embedding services: embeds queries
- Query orchestration → Agentic chat: selects chat engine
- Agentic chat → MCP integrations: invokes tools
README.md - Runtime configuration → LLM providers: configures providers
common/config.py - Runtime configuration → TigerGraph graph and vector store: configures store
common/config.py - TigerGraph graph and vector store → Embedding services: generates vectors
common/embeddings/tigergraph_embedding_store.py - TigerGraph graph and vector store → TigerGraph connections and schema: uses connection
- API application → MCP integrations: initializes clients
graphrag/app/main.py - API application → Logging and metrics: records requests
graphrag/app/main.py