How abhishek-aditya-bs/codegraph works
Project type and purpose
• CodeGraph is a full-stack “GraphRAG Codebase Analysis System.”
• It consists of a Streamlit-based frontend UI, a Python backend that handles ingestion, graph building, vector embedding, and a Neo4j graph database.
• It integrates with external services (OpenAI for embeddings & LLM, GitHub or local file system for code ingestion).
File structure
• Top-level:
• app/ – business logic and Streamlit UI
• docker/ – Dockerfile, Compose, prerequisite checks
• assets/ – demo gif
• tests/ – unit & integration test suites
• requirements.txt, pytest.ini, run_tests.py
• app/:
• main.py – Streamlit UI with workflow tabs
• config.py – env and CLI configuration
• database.py – Neo4j connection & session management
• ingestion.py – code parsing & chunking
• graph_builder.py – knowledge graph creation
• query_processor.py – GraphRAG hybrid query logic
• utilities/ – file, folder, git, LLM, Neo4j helper modules
README insights
• Describes a 3-layer architecture: Structural (knowledge graph), Semantic (vector embeddings), Bridge (integration).
• Workflow: Input → Ingestion → Knowledge Graph → Vector Index → Query Interface → Database Management.
• Dependencies: Streamlit, Neo4j, OpenAI API, Python 3.10+, Docker.
Detail
• How utilities (git_utils, file_utils, llm_utils) are invoked by ingestion or query services.
• Docker Compose network: CodeGraph ↔ Neo4j.