Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

14 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Smart Support AI

Smart Support AI is an innovative, AI-driven customer support solution built to revolutionize query handling. By combining the power of DistilBERT, Google Gemini, and a sophisticated multi-agent system orchestrated through LangGraph, this project automates the entire support process—from categorizing queries to generating personalized responses—all in real time.

With a sleek, user-friendly web interface and a robust FastAPI-powered backend, Smart Support AI is designed for scalability, performance, and exceptional customer experiences.


🚀 Project Objectives

  • Automate Support Tasks: Minimize manual effort using AI agents for query categorization, sentiment analysis, prioritization, answer retrieval, and response generation.
  • Boost Customer Satisfaction: Deliver fast, accurate, and personalized replies to enhance user experience and reduce resolution times.
  • Scale Seamlessly: Handle high query volumes with asynchronous processing and optimized workflows.
  • Ensure Dependability: Incorporate robust error handling, logging, and fallback mechanisms for uninterrupted service.

🔑 Key Features

🧠 Multi-Agent System

A team of specialized AI agents collaborates to process incoming queries:

  • Categorizer Agent: Classifies queries (e.g., Billing, Order) using a fine-tuned DistilBERT model.
  • Sentiment Analyzer Agent: Detects emotional tone (positive, negative, neutral).
  • Priority Agent: Assigns urgency based on sentiment and category.
  • Knowledge Base Search Agent: Retrieves relevant answers from a JSON-based knowledge base.
  • Response Agent: Crafts context-aware replies using Google Gemini.
  • Escalation Agent: Flags queries that need human attention.
  • Ticket Agent: Generates unique support tickets for escalated cases.

🌐 Real-Time Web Interface

  • Tailwind-CSS-styled front-end for submitting queries and viewing responses instantly.

⚙️ Asynchronous Processing

  • Efficiently handles multiple queries using async workflows.

📜 Robust Logging

  • Daily-rotated logs and detailed error handling for high reliability.

🆔 Unique Ticket IDs

  • Secure, collision-free ticket generation using UUIDs.

🛠️ Technologies Used

  • Python — Core language
  • FastAPI — High-performance backend API
  • LangGraph — Agent orchestration framework
  • DistilBERT (Hugging Face) — Query classification
  • Google Gemini — Response generation
  • TensorFlow — Model fine-tuning
  • aiofiles — Asynchronous file handling
  • UUID — Unique ticket generation
  • Tailwind CSS — Frontend styling

⚙️ Setup Instructions

✅ Prerequisites

  • Python 3.12
  • Google Cloud Account (API enabled for Gemini)

📦 Installation Steps

  1. Clone the Repository

    git clone https://github.com/Aswin-Cheerngodan/Smart-Support-AI.git
    cd Smart-Support-AI
  2. Create .env file in the root

    GOOGLE_API_KEY=your_google_api_key

    Run with Docker(Recommended)

    Build the Docker image

    docker build -t smart-support-ai .

    Run the Docker container

    docker run --env-file .env -p 8000:8000 smart-support-ai

    Access the app Open your browser and go to: http://localhost:8000

    Run Locally(without docker)

  3. Set Up a Virtual Environment

    python -m venv venv
    source venv/bin/activate  # On Windows: venv\Scripts\activate
  4. Install Dependencies

    pip install -r requirements.txt
  5. Run the Server

    python src/api/main.py

    Access the Interface Open your browser and go to: http://localhost:8000

▶️ Usage

  • Submit a Query

    Example: "Why hasn’t my order shipped?"

  • Monitor Activity

    Logs: Check logs/ Tickets: View data/tickets.csv

📬 Contact

Questions or suggestions? Open an issue or contact me at aachu8966@gmail.com.

About

An AI-powered customer support system leveraging LLMs, LangGraph, and FastAPI. Automates query categorization, sentiment analysis, and response generation with fine-tuned DistilBERT models, achieving 85%+ accuracy. Features real-time processing, modular escalation workflows, and robust logging with daily rotation.

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Used by

Contributors

Languages