Chatbot Dashboard with Streamlit

This repository contains a Chatbot that uses Streamlit for the frontend, LangChain for creating the retrieval-augmented generation (RAG) chain, MongoDB for storing conversation history, and LangSmith to trace the logs for Dashboard analytics.

Table of Contents

• Introduction
• Features
• Technologies Used
• Workflow
• Environment Variables (.env) Configuration
• How to Set Up the Project
• How to Run the Project Locally
• Upload PDF/TXT Files for Knowledge Base

---

Introduction

This project is a Chatbot that lets users interact with an AI-based assistant using RAG (Retrieval-Augmented Generation). The chatbot uses LangChain to manage conversation history, and MongoDB is used to store user conversations. A detailed analytics dashboard is included to visualize token usage and sentiment analysis for each conversation.

Features

• Used ChromaDB: For RAG (Retrieval-Augmented Generation) chromadb vector database.
• Chatbot Conversation: Users can start new conversations, continue existing conversations, and get answers to their questions based on a retrieval-augmented generation system.
• LangSmith To trace the logs and display dashboard feature that shows the total token usage per conversation.
• MongoDB Integration: Conversation history is stored and retrieved from MongoDB.
• Custom Knowledge Base: Upload your own PDF or TXT files to create a custom knowledge base for the chatbot to use.
• File Management: Easily delete uploaded files anytime from the interface.
• Multiple File Uploads Supported: Users can upload multiple files for the knowledge base at once.

---

Technologies Used

• Streamlit: Frontend UI framework.
• LangChain: Manages the RAG chain and message history.
• LangSmith: Trace logs for dashboard analytics.
• MongoDB: Stores conversation histories.
• ChromaDB: Retrieval-Augmented Generation.
• Python: The language used for developing the application.

---

Workflow

1. Conversation Flow:

   • Users interact with the chatbot via the Streamlit app.
   • The question is passed to the RAG chain, which retrieves relevant documents and generates a concise answer.
   • The conversation is stored in MongoDB with the user's session ID.

2. Dashboard Flow:

   • Token usage is tracked for each conversation.
   • Sentiment analysis is done using VADER, and results are displayed in bar charts.
   • Both token usage and sentiment are visualized in a Streamlit dashboard.

3. RAG Chain:

   • A question is passed to LangChain’s retrieval-augmented generation (RAG) chain.
   • Context is retrieved from a vector store (Chroma).
   • LangChain processes the retrieved context and generates an answer using GPT-4.

4. MongoDB Storage:

   • Conversations are stored in MongoDB.
   • On starting a new conversation, a new session ID is generated.
   • Conversations can be retrieved by their session ID and displayed on the dashboard.

---

Environment Variables (.env) Configuration

The .env file is essential for configuring your environment. You should create this file in the root directory of the project. Below is a sample of what it should contain:

MONGODB_URI=mongodb+srv://:@/?retryWrites=true&w=majority

OPENAI_API_KEY=<your_openai_api_key>

LANGCHAIN_PROJECT=<your_langchain_project_name>

How to Set Up the Project

Clone the repository:

git clone https://github.com/zainzia0341/Chatbot_Dashboard.git
cd Chatbot_Dashboard

Install the dependencies:

Ensure you have Python installed, preferably version 3.9.x, and then install the required dependencies using pip:

pip install -r requirements.txt

Set up MongoDB:

1. Create a MongoDB Atlas account if you don’t have one.
2. Set up a database named chatbot_db and a collection named conversation.
3. Make sure to replace the MONGODB_URI in your .env file with your own MongoDB connection string.

---

How to Run the Project Locally

Set up environment variables:

Create a .env file in the project root directory with the values as shown in the Environment Variables (.env) Configuration section.

Use the following command to start the Streamlit app:

streamlit run app.py