Steam Competitor Analysis Tool 🎮

This project is a full-stack research tool designed to help game developers analyze the competitive landscape on Steam. By inputting a game idea or description, developers can discover similar existing games, understand their market positioning, and gain insights into their strengths and weaknesses based on player reviews.

---

Features

• Competitor Identification: Uses natural language processing to find the top 5 most similar games to your concept.
• Uniqueness Score: Calculates a "uniqueness score" to gauge how crowded the market is for your game idea.
• AI-Powered Review Analysis: Fetches recent reviews for competitor games and uses the Gemini API to summarize common praises and criticisms.
• Interactive Frontend: A clean, modern web interface built with Next.js and React to easily input your ideas and visualize the results.
• RESTful API: A robust backend built with FastAPI that serves the analysis results to the frontend.

---

How It Works

The tool is comprised of a data pipeline, a backend API, and a web frontend.

1. Data Pipeline:

   • A Kaggle dataset of Steam games is downloaded and processed.
   • Game descriptions and tags are cleaned and combined.
   • The sentence-transformers library is used to generate vector embeddings for each game's text, creating a numerical representation of its content.
   • The processed data and embeddings are saved for the API to use.

2. Backend API:

   • The user submits their game description to the /analyze endpoint.
   • The backend generates an embedding for the user's description.
   • Cosine similarity is used to compare the user's embedding to all game embeddings in the database, identifying the most similar games.
   • A uniqueness score is calculated based on the similarity of the top match.
   • The API asynchronously fetches the latest reviews for the top 5 competitor games from Steam.
   • The fetched reviews are passed to the Gemini API with a specialized prompt to extract common challenges and praises.
   • The results are sent back to the frontend as a JSON object.

3. Frontend:

   • The user enters their game idea into a textarea.
   • The frontend calls the backend API and displays the results in a user-friendly format, including the uniqueness score, a list of similar games with links to their Steam pages, and the AI-generated review summary.

---

Tech Stack

This project demonstrates a range of modern software engineering and data science skills:

• Backend Development:
  • Built a high-performance, asynchronous REST API using FastAPI.
  • Utilized Pydantic for robust data validation.
  • Implemented asynchronous network requests with aiohttp to efficiently fetch external data from the Steam API.
• Natural Language Processing (NLP):
  • Applied sentence embeddings (all-MiniLM-L6-v2 model) to represent and compare the semantic content of game descriptions.
  • Used cosine similarity to perform a semantic search for competitor games.
  • Leveraged TF-IDF to predict relevant tags for a given game description.
• Generative AI:
  • Integrated the Google Gemini API for advanced text summarization, crafting a specific prompt to extract actionable insights from unstructured review data.
• Frontend Development:
  • Developed a responsive and interactive user interface with Next.js, React, and TypeScript.
  • Styled the application with TailwindCSS for a modern and clean aesthetic.
• Full-Stack Architecture:
  • Designed and built a complete full-stack application, from data preprocessing to a user-facing web interface.
  • Managed a data pipeline that includes fetching, cleaning, and processing data for use in a machine learning application.

---

USAGE

To run this project, you need to run the backend API and the frontend application separately.

Backend Setup

1. Navigate to the API directory:

   cd api

2. Install Python dependencies:

   pip install -r ../requirements.txt

3. Set up your environment variables:
   • Create a file named .env in the api/ directory.
   • Add your Google API key to the .env file:

     GOOGLE_API_KEY="YOUR_API_KEY_HERE"
     

4. Run the preprocessing script:
   • If updating the data, run the preprocesser first. Then, upload the data to GitHub

   python ../preprocess.py

5. Start the FastAPI server:

   uvicorn main:app --reload

Frontend Setup

1. Navigate to the web directory:

   cd web

2. Install Node.js dependencies:

   npm install
   
   npm run build

3. Start the Next.js development server:

   npm run start

---

Tech Stack

• Backend: Python, FastAPI, sentence-transformers, scikit-learn, pandas, numpy, Google Gemini API
• Frontend: Next.js, React, TypeScript, TailwindCSS
• Data: Kaggle, SQLite
• DevOps: update_data.sh script for data pipeline management