How rsrini7/support-buddy works
Project type and purpose
• This is a full-stack, containerized Retrieval-Augmented Generation (RAG) application called “Support Buddy.”
• Purpose: Ingest support data (Outlook MSG files, Jira tickets, Confluence pages, StackOverflow Q&A), index it in a vector store (ChromaDB or FAISS), and serve a unified semantic-search UI with LLM-powered summaries and recommendations.
File structure
• Top-level directories:
• backend/ – Python FastAPI service, ingestion pipelines, RAG and embedding logic
• frontend/ – React + Material-UI SPA
• confluence-config/ and jira-config/ – XML configs for local Confluence and Jira testing
• docker-compose.yml – orchestrates backend, ChromaDB server, frontend
• backend/app/services: individual modules for msg_parser, jira_service, confluence_service, stackoverflow_service, unified_rag_service, vector_service, etc.
• backend/app/utils: BM25, DSPy utils, deduplication, retrievers, rerankers
• Environment & build scripts: Dockerfile, start_frontend.sh, start_backend.sh, start_chromadb_server.sh/ps1, set_venv.sh
README for additional insights
• Core stacks: FastAPI, React, ChromaDB/FAISS, DSPy (for LLM orchestration), Docker.
• Detailed RAG pipeline: BM25 + vector hybrid, deduplication, reranking (CrossEncoder), LLM generation via DSPy/OpenRouter.
Detail
• Break the API container down to its main sub-components.
• DSPy integration explicitly in the retrieval pipeline.
• Hybrid retrieval (vector + BM25) branching.
• Runtime switch for ChromaDB vs FAISS using a decision diamond or note.