Sentiment Analysis and Data Generation API

A FastAPI-based service for sentiment analysis and synthetic data generation

Features

• Sentiment Analysis with metadata and confidence scores
• Multiple data generation approaches (full metadata, simple sentiment, basic reviews)
• Cost tracking for OpenAI API usage
• Configurable batch processing
• API key authentication
• Rate limit handling
• Extensive error handling and validation

I didnt fully grasp what exactly the input would be either generate or receive data , cleanse and then return in a nice format, so I simply did both variants

API Endpoints

Sentiment Analysis

• POST /analyze

  • Analyzes sentiment of multiple texts with detailed metadata
  • Includes confidence scores and domain-specific insights
  • Batch processing support

• POST /clean-and-analyze

  • Processes and analyzes raw text data
  • Provides sentiment distribution statistics
  • Returns detailed metadata and confidence scores

Data Generation

• POST /generate-data

  • Generates synthetic domain-specific data with full metadata
  • Configurable sentiment distribution
  • Supports verbose mode for detailed metadata
  • Example request:

  {
    "domain": "Restaurants",
    "count": 50,
    "sentiment_distribution": {
      "positive": 0.33,
      "neutral": 0.33,
      "negative": 0.34
    },
    "verbose": true
  }
  

• POST /generate-simple

  • Generates basic review data without sentiment analysis
  • Optimized for high-volume generation
  • Example request:

  {
    "domain": "Shoes",
    "count": 100
  }
  

• POST /bad-generate-data

  • Demonstration endpoint for basic generation
  • Shows impact of prompt engineering on output quality

Utility

• GET /health
  • Health check endpoint
  • Returns service status

Technical Details

Models

• Uses gpt-4o-mini for optimal performance
• Configurable token limits and batch sizes
• Automatic retry mechanisms for API failures

Authentication

All endpoints (except /health) require an API key: