📦 txtemo

txtemo is a lightweight, production-ready Python package for sentiment/emotion detection using a quantized RoBERTa model exported to ONNX.
It runs fast on CPU, making it suitable for local apps, servers, and even lightweight devices.

🚀 Features

• ✅ Quantized RoBERTa ONNX model for speed & efficiency
• ✅ Runs fully on CPU (no GPU required)
• ✅ Easy-to-use Python API
• ✅ Hugging Face Hub integration (auto-downloads model + tokenizer)
• ✅ Returns both labels (Positive/Negative/Neutral) and confidence score

📥 Installation

pip install txtemo

📝 Usage

from txtemo import predict

print(predict("I love this AI model!"))  
# ('Positive 😃', 0.92)

print(predict("This is the worst thing ever."))  
# ('Negative 😡', 0.89)

print(predict("Pranesh"))  
# ('Neutral 😐', 0.75)

🖥️ Command Line Interface (CLI)

You can also use txtemo directly from the command line:

txtemo "This library is amazing!"
# Output: Positive 😃 (0.92)

📊 Labels

• Negative 😡
• Neutral 😐
• Positive 😃

⚡ Performance

• Model: RoBERTa-base (quantized, ONNX)
• Average inference speed: ~3x faster than PyTorch version
• Memory footprint: Reduced by 50%+

🌍 Use Cases

• Chatbots 🤖
• Customer feedback analysis 📢
• Social media monitoring 📱
• Product reviews sentiment 🛒

🔗 Model Source

Hosted on Hugging Face Hub:
PraneshJs/Emotion-detection-Text

📌 Author

Pranesh S
📧 Contact: [praneshmadhan646@gmail.com]