This project builds a Graph-RAG (Retrieval Augmented Generation) knowledge base from your documents.
It extracts a knowledge graph using an LLM, stores it in Neo4j, generates embeddings, and creates hybrid search indexes (vector + keyword) for powerful question-answering.

1. Document Pre-processing

   • Load PDFs (and optionally .txt / .md).
   • Clean and chunk text for downstream processing.

2. Knowledge Graph Creation

   • Convert chunks to entities and relationships using a Large Language Model (LLM).
   • Store nodes and relationships directly in Neo4j.

3. Vector & Keyword Indexing

   • Compute embeddings and store them as node properties.
   • Build vector and full-text indexes in Neo4j for similarity or hybrid search.

4. Graph-Aware QA (Optional)

   • Natural-language questions are translated into Cypher queries with an LLM.
   • Execute Cypher against the graph and return text answers.

.
├─ components/
│  ├─ embeddings.py      -- GetEmbeddings: returns a LangChain embeddings model (OpenAI, Cohere, etc.)
│  ├─ llms.py            -- GetLLM: returns a LangChain chat model (OpenAI, Anthropic, Google Gemini…)
│  ├─ graph_db.py     -- Neo4jStore: connect/clear/add docs/create hybrid indexes + similarity/hybrid search
│  └─ knowledge_graph.py    -- create_knowledge_graph() and cypher_qa() utilities (require an initialized LLM)
│
├─ options/
│  ├─ base_options.py    -- Base CLI arguments: API keys, Neo4j URI, DB name, etc.
│  └─ train_options.py   -- Training/ingestion options: chunk size, LLM provider/model, embedding model…
│
├─ utils/
│  └─ preprocessing.py   -- DocumentProcessor: loads & cleans files, splits into chunks with metadata
│
├─ train.py              -- Main entry point: runs the full pipeline
├─ requirements.txt      -- Python dependencies for the project
└─ README.md

This project is designed as a framework-agnostic starter kit.

You can easily extend it to:

• FastAPI Service: Expose endpoints (e.g., ingest documents, query Neo4j) to turn it into a web API.
• n8n Automation: Connect to Neo4j and run Graph RAG workflows as part of no-code automation.

This flexibility allows you to scale from a simple command-line script to a full API as a Service or automated pipeline.