File size: 2,802 Bytes
3d223e4
 
 
 
 
 
 
 
 
 
 
 
 
f2fd6b1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
---
title: GamerRadar RAG Chatbot
emoji: 📉
colorFrom: yellow
colorTo: purple
sdk: gradio
sdk_version: 6.19.0
python_version: '3.13'
app_file: app.py
pinned: false
---

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
# GamerRadar: Video Game Recommendation System using RAG

GamerRadar is an intelligent video game recommendation assistant built using a Retrieval-Augmented Generation (RAG) pipeline. The system allows users to input their gaming preferences or games they previously enjoyed, and it retrieves contextually relevant recommendations from a specialized dataset. 

This project was developed as part of the Generative AI Bootcamp to demonstrate practical implementation of document indexing, vector databases, LLM orchestration, and interactive user interfaces.

## Features

- Custom Dataset Integration: Uses a curated dataset of video games containing titles, genres, and detailed descriptions.
- Open-Source Embeddings: Leverages Hugging Face sentence-transformers for local text embedding generation without API costs.
- Vector Storage: Implements ChromaDB as a local vector database to store and perform similarity searches on game documents.
- LLM Pipeline: Utilizes LangChain Expression Language (LCEL) to create a robust RAG pipeline powered by Google Gemini 2.5 Flash.
- Interactive UI: Features a web-based chat interface deployed via Gradio for real-time user interaction.

## Tech Stack

- Framework: LangChain / LangChain Community Core
- Embedding Model: Hugging Face (all-MiniLM-L6-v2)
- Vector Store: ChromaDB
- Large Language Model: Google Gemini 2.5 Flash (via Google AI Studio)
- User Interface: Gradio
- Data Management: Pandas

## Project Structure and Pipeline

1. Data Preparation: A dataset of video games is initialized and saved to a local CSV file.
2. Document Loading: The CSV data is extracted and parsed into structured LangChain Document objects.
3. Embedding and Indexing: The documents are processed through the Hugging Face embedding model and indexed inside ChromaDB.
4. Retrieval and Prompting: A custom system prompt is configured to instruct the assistant. A retriever fetches the top 2 most similar games based on the user's input.
5. Generation: The retrieved game contexts and the user query are passed through an LCEL pipeline to Gemini 2.5 Flash to generate accurate, context-aware recommendations.
6. Interface Deployment: The entire workflow is wrapped into a Gradio ChatInterface, launching a local widget and a shareable public URL.

## Setup and Installation

To run this notebook in Google Colab or a local Python environment, install the required dependencies:

```bash
pip install langchain langchain-community langchain-chroma langchain-google-genai sentence-transformers gradio pandas