MNK/portfolio
GenAI & agents

RAG LangChain Chroma

Hybrid search + dual-layer memory

Production-ready RAG application built with FastAPI and LangChain, featuring hybrid search (HNSW vector search + BM25), a dual-layer memory system that summarizes past sessions, and a clean single-page UI for grounded, cited answers over your own documents.

Code

Problem

Answer questions over your own documents with grounded, cited responses.

Approach

  1. Hybrid search combining HNSW vector search and BM25 in ChromaDB
  2. Dual-layer memory: in-chat session memory and LLM-summarized cross-chat memory
  3. FastAPI backend with a single-page UI and GPT-4o-mini
RAG LangChain Chroma workflow: 3 stages from input to result
Workflow

Screens

RAG LangChain Chroma screenshot 1 of 4RAG LangChain Chroma screenshot 2 of 4RAG LangChain Chroma screenshot 3 of 4RAG LangChain Chroma screenshot 4 of 4

Results

  • Cited answers grounded in uploaded documents

Stack

  • LangChain
  • ChromaDB
  • FastAPI
  • GPT-4o-mini

By Muzammil Nawaz Khan, AI Engineer.