Sally: AI Customer Service Chatbot 🤖💬

Sally is an AI-powered customer service chatbot built with Streamlit and Qwen 2 1.5B from Hugging Face Transformers. Sally provides instant, friendly, and helpful support to users, answering questions, resolving issues, and generating daily analytics reports for your business.

---

🚀 Features

• Conversational AI (Qwen 2 1.5B): Uses Alibaba’s advanced open-source Qwen 2 1.5B model to understand and respond to customer queries in real time with balanced performance and efficiency.
• Customer Service Focus: Empathetic, clear, and concise responses tailored for support scenarios.
• Conversation Logging: Every interaction is logged for quality assurance and analytics.
• Daily Analytics Reports: Generates a daily summary of conversations, top questions, and sample Q&A for team review.
• Easy Deployment: Launchable on Streamlit Community Cloud or your own server.
• Customizable: Easily adapt Sally for your brand, FAQ, or support workflow.

---

🛠️ Tech Stack

• Python 3.9+
• Streamlit for interactive web UI
• Hugging Face Transformers with Qwen 2 1.5B model for natural language understanding
• PyTorch (for local inference)
• pandas for analytics and reporting
• CSV for lightweight conversation logs

---

📦 Project Structure

sally-chatbot/
├── deploy.py                # Main Streamlit app
├── requirements.txt         # Python dependencies
├── conversation_logs.csv    # Conversation log (auto-created)
├── daily_report.txt         # Daily analytics report (auto-created)
├── qwen2-faq-finetuned/     # Local model files for Qwen 2 1.5B
└── README.md                # This file

---

🚦 Deployment Guide

1. Clone the Repository

git clone https://github.com/yourusername/sally-chatbot.git
cd sally-chatbot

2. Install Dependencies

pip install -r requirements.txt

3. Add Qwen 2 1.5B Model Files

Place your Qwen 2 1.5B Hugging Face model files in ./qwen2-faq-finetuned/ (must include config.json, pytorch_model.bin or model.safetensors, and tokenizer files).

4. Run Locally

streamlit run deploy.py

5. Deploy to Streamlit Cloud

• Push your code to GitHub.
• Go to Streamlit Cloud, connect your repo, and deploy.

Note: Due to resource limits on Streamlit Cloud, large models like Qwen 2 1.5B may not work efficiently. For production, use a hosted inference API or deploy on your own infrastructure with GPU support.

---

📊 Analytics & Reporting

• All conversations are logged in conversation_logs.csv.
• Type report in the chat to generate a daily analytics report (daily_report.txt), including:
  • Total conversations
  • Top customer questions
  • Sample Q&A

---

⚠️ Limitations

• Model files are not included in the repository due to size constraints.
• Streamlit Cloud does not support large models well. For optimal performance, use an API-based inference or a server with sufficient GPU resources.
• No live agent handoff (but Sally can suggest escalation).

---

📜 License & Credits

• Model: Qwen 2 1.5B by Alibaba (open-source with respective licensing terms).
• Frameworks: Streamlit, PyTorch, Hugging Face Transformers.

---