sri-mathi/Modular-RAG-System

Modular RAG System is an expert-grade pipeline to convert static documents into a searchable, structurally-aware knowledge base. Intended for ML engineers, RAG developers, and teams building document-QA or retrieval-augmented apps, it combines semantic vector search with graph-based relationship mapping so agents can retrieve facts and reason about entity connections. Key features: - Hybrid retrieval (GraphRAG): ChromaDB embeddings + NetworkX graph traversal. - Provider abstraction (BaseEmbedder/BaseLLM) for pluggable models (Ollama embeddings, Groq LLM). - LLM re-ranking (Groq cross-encoder) to trim candidates. - Multi-query expansion for technical queries. - Visualization (PCA embedding clusters, interactive knowledge graphs). - Automated evaluation (RAGAS) measuring faithfulness and relevance. - Tools to ingest PDFs, build graphs, query, rerank, and visualize.

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.