Smart Prompt Routing System

An intelligent multi-task NLP system that automatically detects user intent and routes prompts to specialized language models for translation, summarization, sentiment analysis, question-answering, and conversational chat.

Examples

Overview

This project implements a smart routing system that:

1. Classifies user intent using a trained machine learning classifier
2. Routes prompts to the appropriate NLP model
3. Returns intelligent responses based on the detected task

Supported Tasks

• Translation (EN → FR): Translate English text to French
• Summarization: Generate concise summaries of longer texts
• Sentiment Analysis: Analyze emotional tone (positive/negative)
• Question Answering: Answer factual questions
• Conversational Chat: Engage in casual dialogue

Models Used

Task	Model	Size	Description
Intent Classification	Logistic Regression + TF-IDF	~1 MB	Custom-trained classifier to detect user intent
Translation	Helsinki-NLP/opus-mt-en-fr	~300 MB	English to French translation
Summarization	t5-small	~240 MB	Text summarization with repetition prevention
Sentiment Analysis	distilbert-base-uncased-finetuned-sst-2-english	~250 MB	Binary sentiment classification
Question Answering	google/flan-t5-small	~300 MB	Factual question answering
Chat	microsoft/DialoGPT-small	~350 MB	Conversational dialogue generation

Directory Layout

SMART-PROMPT-ROUTING-SYSTEM/
├── app/
│   ├── config/
│   ├── model/
│   ├── routers/
│   ├── util/
│   ├── __init__.py
│   └── main.py
├── data/
├── notebooks/
├── .env.example
├── .gitattributes
└── README.md

Directory Descriptions

/app

The main application directory containing all core application code.

• config/ - Configuration files and settings for the application
• model/ - Machine learning models and model-related code
• routers/ - API route definitions and endpoint handlers
• util/ - Utility functions and helper modules
• main.py - Application entry point and main execution file

/data

Storage directory for data files, datasets, and data-related resources.

/notebooks

Jupyter notebooks for experimentation, analysis, and documentation.

Root Files

• .env.example - Template for environment variables configuration
• .gitattributes - Git attributes configuration for repository
• README.md - Project documentation and overview

Installation

1. Clone the repository:

git clone
cd smart-prompt-routing-system

2. Install dependencies:

pip install -r requirements.txt

3. Ensure you have the trained classifier models in the app/model directory:
   • intent_classifier.pkl
   • label_encoder.pkl

Usage

Command Line Interface

Run the interactive CLI:

python main.py

Acknowledgments

• Made by ABADLI Badreddine, MAHAMMED Abdallah Slimane and BOUHAMZA Abderrahmane
• Hugging Face for the Transformers library
• Model creators: Helsinki-NLP, Google, Microsoft, Hugging Face team
• scikit-learn for machine learning utilities