How kamalesh346/graph_rag works
This repository is a legal-data preparation and GraphRAG retrieval system focused on Indian statutes and Supreme Court judgments. Its principal user workflow parses a legal query, expands statutory references through a knowledge graph, searches statute and judgment vectors, and reranks results. A separate statute pipeline validates, parses, enriches, chunks, and ingests statute data. The tree also documents judgment extraction and indexing capabilities; their internal wiring is only partly evidenced in the excerpts, so uncertain links are omitted.
Main entry points: hybrid_reranker.py Hybrid Search · main_statute_pipeline.py Statute Orchestrator · extract_pdfs.py PDF Extraction · statute_ingestion.py Statute Ingestion
Components 23
Judgment Data
- PDF Extraction
scripts/extract_pdfs.py - Judgment Parser
scripts/parse_legal_sections.py - Case Normalization
scripts/normalize_structured_cases.py - Judgment Chunking
scripts/chunk_normalized_cases.py - Case Vector Index
scripts/build_chroma_index.py - Reference Analysis
scripts/analyze_legal_references.py - Judgment PDFs
data/judgement_pdf - Extracted Cases
data/extracted
Statute Pipeline
- Statute Orchestrator
scripts/main_statute_pipeline.py - Statute Validation
scripts/statute_validator.py - Statute Parsing
scripts/statute_parser.py - Statute Chunking
scripts/statute_chunker.py - Statute Ingestion
scripts/statute_ingestion.py - Normalized Acts
data/legal/acts
Legal Retrieval
- Query Parsing
scripts/retrieval/query_parser.py - Hybrid Search
scripts/retrieval/hybrid_reranker.py - Graph Expansion
scripts/retrieval/graph_retriever.py - Retrieval Evaluation
scripts/retrieval/evaluate_retrieval.py
Search Stores
- Statute Vectors
scripts/statute_ingestion.py - Judgment Vectors
scripts/build_chroma_index.py - Statute Knowledge Graph
scripts/statute_ingestion.py
Other components
- Legal User
- Legal Researcher
Connections 18
- Judgment PDFs → PDF Extraction: reads PDFs
scripts/extract_pdfs.py - PDF Extraction → Judgment Parser: parses pages
scripts/extract_pdfs.py - PDF Extraction → Extracted Cases: writes JSON
scripts/extract_pdfs.py - Statute Orchestrator → Statute Validation: validates acts
scripts/main_statute_pipeline.py - Statute Orchestrator → Statute Parsing: parses acts
scripts/main_statute_pipeline.py - Statute Orchestrator → Statute Chunking: chunks sections
scripts/main_statute_pipeline.py - Statute Orchestrator → Statute Ingestion: ingests chunks
scripts/main_statute_pipeline.py - Normalized Acts → Statute Validation: reads acts
scripts/statute_validator.py - Statute Ingestion → Statute Vectors: writes chunks
scripts/statute_ingestion.py - Statute Ingestion → Statute Knowledge Graph: writes graph
scripts/statute_ingestion.py - Legal User → Hybrid Search: submits query
scripts/retrieval/hybrid_reranker.py - Hybrid Search → Query Parsing: parses intent
scripts/retrieval/hybrid_reranker.py - Hybrid Search → Graph Expansion: expands sections
scripts/retrieval/hybrid_reranker.py - Graph Expansion → Statute Knowledge Graph: queries graph
scripts/retrieval/graph_retriever.py - Hybrid Search → Statute Vectors: searches statutes
- Hybrid Search → Judgment Vectors: searches judgments
scripts/retrieval/hybrid_reranker.py - Retrieval Evaluation → Hybrid Search: benchmarks search
scripts/retrieval/evaluate_retrieval.py - Legal Researcher → Hybrid Search: requests research