---
title: MindMatch
emoji: 🚀
colorFrom: red
colorTo: red
sdk: docker
app_port: 8501
tags:
- streamlit
pinned: false
short_description: 'Hybrid dating recommendation system combining collaborative '
license: mit
---

# Welcome to Streamlit!

Edit `/src/streamlit_app.py` to customize this app to your heart's desire. :heart:

If you have any questions, checkout our [documentation](https://docs.streamlit.io) and [community
forums](https://discuss.streamlit.io).

# MatchMind: Dating Profile Recommendation System

## Overview
MatchMind is a hybrid dating recommendation system that combines collaborative filtering and content-based approaches to deliver high-quality match suggestions. The system generates text-based user profiles, encodes them into vector embeddings, and stores them in a Qdrant vector database. It retrieves potential matches using semantic search and applies a reranking step with a cross-encoder model for improved match quality. The system mirrors production-grade matching logic and supports benchmarking with LLMs.

## What the Project Does
- **Profile Generation:** Converts structured user data and Q&A into natural language profile templates.
- **Embedding:** Uses the `nomic-ai/nomic-embed-text-v1.5` model (SentenceTransformer) to generate 768-dimensional embeddings for each profile.
- **Vector Database:** Stores and searches embeddings in Qdrant (collections: `dating_M`, `dating_F`).
- **Semantic Search:** Retrieves top candidate matches using vector similarity (cosine distance).
- **Reranking:** Uses the cross-encoder model `cross-encoder/ms-marco-MiniLM-L-6-v2` to rerank the top candidates for final recommendations.
- **Multi-Signal Scoring:** Supports additional signals and benchmarking for match quality.
- **Web App:** Streamlit interface for user profile entry, Q&A, match discovery, and full profile viewing.
- **Benchmarking:** Optional LLM-based scoring for research and evaluation.

## Models and Technologies Used
- **Embedding Model:** `nomic-ai/nomic-embed-text-v1.5` (SentenceTransformer)
- **Reranking Model:** `cross-encoder/ms-marco-MiniLM-L-6-v2` (CrossEncoder)
- **Vector Database:** Qdrant (collections: `dating_M`, `dating_F`)
- **Backend:** Python, Flask, SQLAlchemy, MySQL
- **Frontend:** Streamlit
- **Benchmarking (optional):** LLMs via Ollama

## Installation
1. Clone the repository.
2. Install dependencies:
	 ```
	 pip install -r requirements.txt
	 ```
3. Set up environment variables in a `.env` file (see `config.py` for required keys).

## Try the Web Demo
You can try out the MatchMind web app here:
https://ultron3002-mindmatch.hf.space/

## Project Structure
- `app.py` — Streamlit web interface
- `main.py` — Core ranking and embedding logic
- `config.py` — Configuration and environment variables
- `db/` — Database connection and SQL utilities
- `utils/` — Embedding, ranking, and Qdrant utilities
- `page_modules/` — Streamlit page components
- `benchmarking/` — LLM benchmarking scripts
- `migration_notebooks/`, `templating_notebooks/` — Data and template notebooks

## Requirements
See `requirements.txt` for all dependencies.

## License
MIT License