ChatBot with Streamlit, LangChain, and Firebase

Overview

This project is a conversational AI chatbot application built with Streamlit for the frontend, LangChain for AI capabilities, Firebase for user data storage, and Pinecone for vector-based long-term memory. The chatbot maintains conversation history and can recall relevant facts from previous interactions to provide context-aware responses.

Features

• Interactive Chat Interface: User-friendly web interface built with Streamlit.
• GPT-4o-mini Integration.
• Dynamic LLM Tuning: Users can customize LLM parameters in real-time for personalized responses.
• Conversation Memory: Stores and retrieves past conversations for each user using Firebase.
• Long-term Context:
  • Fact Extraction: LLM extracts user key facts from a conversation.
  • Pinecone Vector Store: Maintains relevant facts about the user in vector store for tailored responses.
• Authentication
  • Secure Login: Email/password via Firebase REST API
  • User Management:
    • Backend: Firebase Admin SDK (create/update users)
    • Metadata: Firestore stores last_login, preferences, and LLM settings

Technology Stack

• Frontend: Streamlit
• AI Framework: LangChain
• AI model: gpt-4o-mini
• Database:
  • Firebase Firestore (user data and conversations)
  • Firebase (vector store for long-term memory)
• Authentication: Firebase Authentication

Screenshots

Installation

To run the app locally, follow these steps:

1. Clone the repository:

   git clone git@github.com:TuringCollegeSubmissions/azhurb-AE.2.5.git
   cd cd azhurb-AE.2.5/

2. Install Poetry (if you don't have it installed yet):

   curl -sSL https://install.python-poetry.org | python3 -

3. Install the required dependencies:

   poetry install
   

4. Activate the virtual environment:

   poetry env activate

5. Set up Firebase:

   • Create a Firebase project at Firebase Console
   • Download your service account JSON file, paste its contents as FIREBASE_CREDENTIALS into .streamlit/secrets.toml file
   • Enable Firestore Database and Authentication in your Firebase console

6. Set up environment variables

   OPENAI_API_KEY=your-api-key
   FIREBASE_PROJECT_ID=your_firebase_project_id
   PINECONE_API_KEY=your_api-key

7. Run the Streamlit app:

   streamlit run app.py