RESOLVEAI - AI Complaint Classifier and Responder

RESOLVEAI is an AI-powered customer complaint management system built using Streamlit. It intelligently classifies complaints by category and urgency, generates AI-based responses, and provides separate interfaces for customers and administrators to manage and resolve issues effectively.

---

🚀 Features

🧑‍💻 Customer Interface

• Submit new complaints with auto-tagging of category and urgency
• AI-generated response using Groq LLM (Gemma2 9B IT)
• View history of past complaints
• Submit follow-up messages

🛠️ Admin Interface

• Secure login for admins
• View, search, and filter all customer complaints
• Update complaint status and resolution messages
• Track urgency and categorize complaints

🤖 Under the Hood

• Machine Learning model for category prediction
• Sentiment model for urgency prediction
• Urgency detection using a custom classifier
• Vector Search with FAISS for semantic understanding
• LLM response generation using HuggingFace embeddings + Groq’s Gemma 2B

---

🗂️ Project Structure

Final AI CCR/
│
├──main/
│   ├── pages/
│      ├── admin_dashboard.py         # Admin dashboard and functionality
│      ├── chatbot.py                 # Standalone chatbot interface
│      └── customer_form.py           # Customer UI for complaint submission & history
│   ├── app.py                      # Entry point (Login & routing logic)
│   ├── complaint_classifier.pkl      # Classification model
│   ├── classifier.py              # Complaint classification, FAISS vector search, LLM response
│   └── improved_complaint_classifier_training.ipynb  # ML training model
│
├── data/
│   └── realistic_complaints_dataset_9990_cleaned.csv
│
├── faiss_index/
│   ├── index.faiss
│   └── index.pkl              # FAISS vector store for semantic search
│
├── .gitignore
├── complaint_responder.sql    # SQL schema and initialization
├── data.txt
├── venv/                      # Virtual environment
└── improved_complaint_classifier_training.ipynb

Note: REQUIRES MYSQL CONNECTION

Database Schema

The system uses a MySQL database with the following structure:

-- Create the complaints table (if it doesn't exist)
CREATE TABLE IF NOT EXISTS complaints (
    Id INT NOT NULL AUTO_INCREMENT PRIMARY KEY,
    C_id INT NOT NULL,
    Complaint_text TEXT,
    Predicted_category VARCHAR(100),
    Auto_response TEXT,
    Timestamp DATETIME DEFAULT CURRENT_TIMESTAMP,
    Complaint_status VARCHAR(50) DEFAULT 'None',
    Urgency VARCHAR(10),
    Followup TEXT,
    Admin_note TEXT

);

-- Simplified Customers Table
CREATE TABLE IF NOT EXISTS customers (
    customer_id INT NOT NULL AUTO_INCREMENT PRIMARY KEY,
    name VARCHAR(100) NOT NULL,
    email VARCHAR(100) NOT NULL UNIQUE
);

-- Simplified Admins Table
CREATE TABLE IF NOT EXISTS admins (
    admin_id INT NOT NULL AUTO_INCREMENT PRIMARY KEY,
    name VARCHAR(100) NOT NULL,
    email VARCHAR(100) NOT NULL UNIQUE
);

---

🧠 Tech Stack

• Frontend: Streamlit
• Backend: Python, Hugging Face Transformers, Groq API
• ML Models: BERT classifier (sklearn), Urgency classifier
• Vector Store: FAISS
• Database: MySQL (via complaint_responder.sql)
• Embeddings: Hugging Face sentence transformer

---

🛠️ Setup Instructions

1. Clone the repository

   git clone https://github.com/yourusername/RESOLVEAI.git
   cd RESOLVEAI

2. Create and activate a virtual environment

   python -m venv venv
   source venv/bin/activate  # on Unix
   .\venv\Scripts\activate   # on Windows

3. Install dependencies

   pip install -r requirements.txt

4. Configure Groq API key
   In classifier.py, set your API key:

   os.environ["GROQ_API_KEY"] = "your-api-key"

5. Run the app

   streamlit run app.py

---

📄 License

MIT License. Feel free to use and modify for educational and commercial use.

🙋‍♂️ Author

Prakshi
AI & Python Developer
LinkedIn | GitHub