README.md

metadata

title: Ai Restaurant Recommendation Chabot
emoji: 🐨
colorFrom: yellow
colorTo: pink
sdk: docker
pinned: false
license: mit
short_description: LLM based Restaurant Recommendation chat bot.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

AI Restaurant Recommendation Chatbot (Qwen2-1.5B-Instruct) 🍜

This project implements an intelligent restaurant recommendation system using OpenStreetMap, sentence-transformer embeddings, FAISS vector search, and an LLM assistant (Qwen2-1.5B-Instruct). It provides two main features:

1. Embedding-based restaurant recommendations

2. A conversational AI assistant that explains and selects the best restaurants based on the retrieved data.

The system is built with FastAPI, Sentence Transformers, FAISS, PyTorch, Transformers, and Docker.

1. System Architecture

Pipeline Overview

1. User provides a query + location (lat/lon).

2. API calls Overpass API to fetch real nearby restaurants from OSM.

3. Each restaurant is converted into a dense textual description containing:

   • Name

   • Cuisine

   • Opening hours

   • Coordinates

   • City

   • Street

   • Neighborhood

   • Amenity type

4. These descriptions are encoded into embeddings using: sentence-transformers/all-MiniLM-L6-v2

5. Embeddings are indexed using FAISS (Inner Product search with L2 normalization).

6. User query is embedded and compared to all restaurants.

7. Results are ranked using a blended score:

   final_score = 0.6 * semantic_similarity + 0.4 * distance_score
   

8. The top-k results are sent to the user.

9. For chat requests:

   • The selected restaurants are fed into an LLM prompt.

   • Qwen2-1.5B-Instruct generates a human-friendly explanation.

2. Technologies Used

Core

• FastAPI — API framework

• Uvicorn — ASGI server

• Sentence Transformers — embedding model

• FAISS — similarity search engine

• Transformers (HuggingFace) — LLM inference

• PyTorch — used by both LLM and embedding model

• OSM Overpass API — live restaurant data

Frontend

• Vanilla HTML, CSS, JavaScript

• Chat-style UI with:

  • User messages

  • AI messages

  • Loading animation

  • Location input + auto-detect

Deployment

• Docker — fully containerized backend

• Runs with a single command

3. Endpoints

GET /chat

Returns an AI-generated explanation from the LLM.

The endpoint:

1. Runs the same FAISS recommendation pipeline.

2. Creates a structured chat template.

3. Uses Qwen2-1.5B-Instruct.

4. Returns a natural-language text response recommending the top 4 best restaurant for the user's query.

4. Embeddings and Models

Embedding Model

• sentence-transformers/all-MiniLM-L6-v2

• 384-dimensional embeddings

Similarity Search

• FAISS IndexFlatIP with L2 normalized vectors

LLM Model

• Qwen/Qwen2-1.5B-Instruct

• Runs on CPU in this project

• Used only for final natural-language reasoning

5. How to Run Locally

Install dependencies

In a .venv environment, run:

pip install -r requirements.txt

Start the server

uvicorn app.main:app --reload

Then open:

http://localhost:7860

6. Docker Deployment

Build the Docker image

docker build -t restaurant-ai .

Run the container

docker run -p 7860:7860 restaurant-ai

Your API and frontend will be available at:

http://localhost:7860

7. Project Structure

app/
  ├── main.py                 # FastAPI app + LLM chat endpoint
  ├── osm_recommend.py        # OSM + embeddings + FAISS pipeline
  ├── static/
       ├── index.html         # UI
       ├── style.css
       └── script.js
requirements.txt
Dockerfile
README.md