🤖 Sales Meeting Chatbot

An AI-powered chatbot that helps sales representatives query information from previous client meetings using RAG (Retrieval-Augmented Generation).
It processes sales meeting transcripts and summaries, chunks content, embeds it into Qdrant, and serves a chatbot API with FastAPI + LangChain + OpenAI.

---

✨ Features

• 📁 Manual Data Upload from Google Drive (transcripts & summaries)
• 📝 Text Processing of meeting transcripts into structured format
• 🧩 Smart Chunking (token-based with overlap for context retention)
• 📦 Vector Storage in Qdrant (Cloud or Local)
• 🧠 RAG Chatbot API powered by LangChain & OpenAI
• 🌐 Simple Chat Widget for sales team
• 🔄 Manual Refresh Workflow (process → chunk → ingest → deploy)
• ☁️ Deployable on Render / Railway / AWS / Docker

---

📂 Project Structure

sales_chatbot/
├── api/                     # FastAPI app
│   └── main.py
├── data/
│   ├── raw/                 # Raw meeting transcripts/summaries
│   └── clean/               # Processed text + chunks
├── frontend/                # Simple HTML/JS chatbot widget
├── process_meetings.py      # Processes meeting transcripts
├── chunking.py              # Splits text into chunks for embeddings
├── ingest_qdrant.py         # Ingests chunks into Qdrant
├── ensure_indexes.py        # Ensures indexes exist in Qdrant
├── run_pipeline.sh          # End-to-end pipeline runner
├── requirements.txt         # Python dependencies
├── Dockerfile               # For containerized deployment
└── .github/workflows/       # GitHub Actions automation

---

⚙️ Setup

1. Clone the repo

git clone https://github.com/your-username/sales_chatbot.git
cd sales_chatbot

2. Create virtual environment

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

3. Environment variables

Create a .env file:

OPENAI_API_KEY=your_openai_api_key
QDRANT_URL=https://your-qdrant-instance.qdrant.io:6333
QDRANT_API_KEY=your_qdrant_api_key
COLLECTION_NAME=sales_meetings

---

🛠️ Usage

1. Add meeting data

Place your meeting transcripts and summaries in data/raw/ folder.

2. Run the pipeline

./run_pipeline.sh

This will:

• Process meeting transcripts and summaries
• Chunk content
• Ingest into Qdrant
• Ensure indexes

3. Run chatbot locally

uvicorn api.main:app --reload --port 8000

Test endpoint:

curl -s -X POST http://127.0.0.1:8000/chat \
  -H "Content-Type: application/json" \
  -d '{"session_id":"sales_rep_1","message":"What pricing concerns did clients have?"}'

---

🌐 Deployment

Docker

docker build -t sales-chatbot .
docker run -p 8000:8000 sales-chatbot

Render (recommended)

1. Push repo to GitHub
2. Create a Web Service on Render
3. Point to this repo
4. Expose port 8000

Your bot will be live at https://<your-app>.onrender.com.

---

💬 Embedding Chat Widget

Add this snippet to your sales dashboard:

<div id="sales-chatbot"></div>
<link rel="stylesheet" href="https://sales-bot.onrender.com/style.css">
<script src="https://sales-bot.onrender.com/app.js"></script>
<script>
  SalesChatbotWidget.init({
    endpoint: "https://sales-bot.onrender.com/chat",
    title: "Sales Assistant 🤖",
    subtitle: "Query previous client meetings"
  });
</script>

---

🧪 Example Queries

• What pricing objections did clients mention?
• Which features were most requested by enterprise clients?
• What concerns did clients have about implementation?
• Show me feedback about our competitor comparisons
• What questions do clients ask about security?
• Which industries showed most interest in our product?

---

📜 License

MIT License © 2025