sentimetric 1.0.2

A modern sentiment analysis library with optional LLM support

🎭 Sentimetric - Modern Sentiment Analysis

Sentimetric is a modern, fast, and accurate sentiment analysis library with optional LLM support for complex emotions, sarcasm, and nuanced context.

✨ Features

• 🚀 Fast Rule-Based Analysis
• 🧠 Multi-LLM Support (OpenAI, Google Gemini, Anthropic Claude, Cohere, Hugging Face)
• 💰 Cost-Aware Model Selection with automatic fallback to cheaper models
• 📊 Batch Processing with parallel execution
• 🎯 High Accuracy with modern slang & emojis
• 🔧 Simple API: from sentimetric import analyze

🚀 Quick Start

Installation

pip install sentimetric

Basic Usage

from sentimetric import analyze

# Quick analysis
result = analyze("This is amazing!")
print(result)
# Example output: SentimentResult(polarity=+0.90, category='positive', confidence=0.85)

Multi-LLM Usage

Sentimetric now supports multiple LLM providers! Choose from OpenAI, Google Gemini, Anthropic Claude, Cohere, or Hugging Face.

Installation with LLM Support

# Install with all LLM providers
pip install "sentimetric[all]"

# Or install specific providers
pip install "sentimetric[openai]"      # For OpenAI
pip install "sentimetric[google]"      # For Google Gemini
pip install "sentimetric[anthropic]"   # For Anthropic Claude
pip install "sentimetric[cohere]"      # For Cohere
pip install "sentimetric[huggingface]" # For Hugging Face

Basic LLM Usage

from sentimetric import LLMAnalyzer

# Auto-selects the best available provider
analyzer = LLMAnalyzer()  # Automatically detects available API keys

# Or specify a provider
analyzer = LLMAnalyzer(provider="openai", model="gpt-3.5-turbo")
# analyzer = LLMAnalyzer(provider="google", model="gemini-1.5-flash")
# analyzer = LLMAnalyzer(provider="anthropic", model="claude-3-haiku-20240307")
# analyzer = LLMAnalyzer(provider="cohere", model="command")
# analyzer = LLMAnalyzer(provider="huggingface", model="mistralai/Mixtral-8x7B-Instruct-v0.1")

result = analyzer.analyze("Oh great, another bug 🙄")
print(result.category)  # 'negative' (catches sarcasm)
print(result.reasoning) # Explanation from the LLM

Environment Variables

Set your API keys as environment variables:

# OpenAI
export OPENAI_API_KEY="your-openai-key"

# Google Gemini
export GOOGLE_API_KEY="your-google-key"

# Anthropic Claude
export ANTHROPIC_API_KEY="your-anthropic-key"

# Cohere
export COHERE_API_KEY="your-cohere-key"

# Hugging Face
export HUGGINGFACE_API_KEY="your-hf-key"

Cost-Aware Features

# Auto-select cheapest model
analyzer = LLMAnalyzer(provider="openai", model="auto")  # Uses gpt-3.5-turbo

# Fallback to cheaper models on failure
analyzer = LLMAnalyzer(
    provider="openai",
    model="gpt-4",
    fallback_to_cheaper=True  # Falls back to gpt-3.5-turbo if gpt-4 fails
)

📚 Examples

See examples.py for comprehensive usage examples. Use python examples.py to run them locally.

🛠️ API Reference

Core Functions

• analyze(text, method='auto') - Quick sentiment analysis
• analyze_batch(texts, method='rule') - Batch sentiment analysis
• compare_methods(text, api_key=None) - Compare rule-based vs LLM analysis

Classes

• Analyzer - Fast rule-based sentiment analyzer
• LLMAnalyzer - Multi-provider LLM analyzer (OpenAI, Google, Anthropic, Cohere, Hugging Face)
• SentimentResult - Result container with polarity, category, confidence, reasoning, emotions, tone
• Benchmark - Accuracy testing and comparison utilities

LLMAnalyzer Constructor Parameters

• provider - LLM provider ('openai', 'google', 'anthropic', 'cohere', 'huggingface', or 'auto')
• model - Model name or 'auto' for cheapest available
• api_key - API key (optional, uses environment variables)
• fallback_to_cheaper - Whether to fall back to cheaper models if requested model fails (default: True)

📞 Support

• Author: Abel Peter
• Email: [e-mail]
• Issues: https://github.com/peter-abel/sentimetric/issues

---

Made with ❤️ by Abel Peter