Vector Database Analysis (VDBA)

A Flask RAG app for comparing vector databases, embedding models, and LLMs

VDBA is a Flask web app for comparing vector database + embedding + LLM combinations on RAG tasks.

  • Supports multiple vector DBs: ChromaDB, PGVector (Postgres), Pinecone
  • Configurable embedding models (Sentence Transformers) and LLMs (local via Ollama, or OpenAI API)
  • Ingests custom datasets/documents, then runs similarity search and RAG-based Q&A/chat over them
  • Dashboard UI for choosing vector DB / embedding model / LLM, plus a chat interface with visualized reasoning steps
  • Built with Flask, LangChain, optional Docker for a PGVector container

Includes exploratory Jupyter notebooks alongside the Flask app for analysis/evaluation work.

References