Steam Reviews Sentiment Analysis API

API to analyze sentiment of Steam game reviews using Large Language Models.

How it works:

• We use the Steam API to obtain game reviews
• We process reviews with Llama 3.1-70B via HuggingFace
• We return a concise sentiment analysis summary

API Endpoints

Health Check

GET /

Returns system status to verify the server is running.

Response:

{
  "status": "system ok",
  "message": "Steam Reviews API is running"
}

Sentiment Analysis

GET /sentiments?game_id={game_id}&limit={limit}

Analyzes sentiment for Steam game reviews.

Query Parameters:

• game_id (optional): Steam game ID. If not provided, uses default game (2592160)
• limit (optional): Maximum number of reviews to analyze. Default is 200

Examples:

# Default game with default limit
GET /sentiments

# Specific game (CS:GO) with default limit
GET /sentiments?game_id=730

# Specific game with custom limit
GET /sentiments?game_id=730&limit=100

Response:

{
  "game_id": 730,
  "sentiment": "The reviews for this game are generally positive...",
  "reviews_count": 100,
  "success": true
}

Setup

Prerequisites

• Python 3.11+
• Docker Compose
• HuggingFace API key

Installation

pip install -r requirements.txt

Environment Variables

export HUGGINGFACE_API_KEY="your_huggingface_api_key_here"

Running the Server

uvicorn main:app --reload --host 0.0.0.0 --port 8000

Running with Docker Compose

# Run the application with Docker Compose with no logs, remove the -d for logs
docker-compose up -d --build

# Stop the application (keeps containers for quick restart)
docker-compose stop

# For complete removal (stops and removes containers)
docker-compose down

Note: Ensure your docker-compose.yaml includes the service definition and environment variables (e.g., via env_file: .env or environment).

Documentation

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

Testing-TBD

Example Usage

# Health check
curl http://localhost:8000/

# Analyze CS:GO reviews
curl "http://localhost:8000/sentiments?game_id=730&limit=50"