Multi Document RAG Customer Support Telegram Chatbot : OpenAI, Langchain, Pinecone

A Customer Support chatbot for dentistry built using OpenAI, Langchain, and Pinecone.The chatbot is powered by generative AI, enabling it to provide assistance and responses on medical topics and clinical info(prices, schedule and policy). It's designed to offer interactive and intelligent support for users seeking Dental-related information.

Technology Stack Overview

• FastAPI(Web framework): Web framework for building APIs with Python based on standard Python type hints. In order to deploy on Vercel(or any other hosting platforms) you should move the bot logic to an HTTP handler (e.g., using FastAPI or Flask)
• Telegram(Frontend): Telegram serves as a frontend, a user interface, for chatbots, allowing users to interact with automated programs through a familiar messaging interface
• OpenAI(LLM): Provides AI-powered responses based on user input.
• Langchain(LLM framework): Facilitates the integration of language models for better response generation.
• Pinecone(Vector Store): Manages and stores vectors to enable semantic search for medical information.
• Medical resource(Knowledge base):The GALE ENCYCLOPEDIA of MEDICINE plus to the clinic's private info .
• Deployment : Deploy locally, once everything works locally, you can deploy to a server (e.g., Railway, Render, Vercel). Follow the instruction in VERCEL.md for deploying to the Vercel platform.

🚀 Setup Guide

Prerequisites

• Python 3.8+
• FastAPI
• python-telegram-bot
• OpenAI LLM model
• Langchain
• Pinecone

Prerequisites before cloning the repo :

• create telegram bot (discussed below)
• create a vector index in Pinecone, upload your own document(use store_index.py)

✅ 1. Create a Telegram Bot

1. Open Telegram and search for @BotFather.

2. Start a chat and send the command:

   /start
   

3. Create a new bot by typing:

   /newbot
   

4. Provide:

   • Bot Name (e.g., "MySupportBot")
   • Username (must end with bot, e.g., "my_support_bot")

5. BotFather will send you a Bot Token like:

   123456789:ABCdefGhijkLmnoPQRstuVwxyZ
   

6. Save this token — you’ll need it for .env.

---

✅ 2. Clone the Repository

git clone https://github.com/Sarach-git/dentist_assistant

---

✅ 3. Create and Update the .env File

Create a .env file in the project root and add your values; To enable retrieval-augmented generation (RAG), you need to create a vector index in Pinecone and upload your own documents.

TELEGRAM_BOT_TOKEN=your_telegram_bot_token_here
PINECONE_API_KEY=your_openai_api_key_here

---

✅ 4. Install Dependencies

it is recommended to create virtual env

pip install -r requirements.txt

---

✅ 5. Run the Chatbot Locally

• multi doc rag :

python multi.py

• single doc rag :

python app.py

---

✅ 6. Test the Bot

1. Open Telegram
2. Find your bot by its username (e.g., @my_support_bot)
3. Send a test message — your chatbot should respond!

---

✅ 7. Run the Chatbot on hosting Platform; Vercel

Follow the instruction in VERCEL.md for deploying to the Vercel platform.

---

Contributing

Feel free to fork the repository, make changes, and submit a pull request if you want to contribute to the project.

License

This project is licensed under the MIT License.