GENERIC BOT

Overview

Generic Bot is a lightweight, modular chatbot framework designed to help researchers deploy and experiment with text-based conversational agents efficiently. Built with a Django backend, a React frontend, the Kani Framework for LLM integration, and a MariaDB database, it is fully containerized using Docker for seamless deployment.

Researchers can easily customize in config.json file:

• Language model selection
• Anthropomorphism settings
• Bot-specific prompts
• And more to come

The project streamlines research workflow with integration with tools like Qualtrics, and allows efficient data collection and analysis. For detailed design specifications, see the link

Directory Structure

---

HUMANLIKE-CHATBOT/
├── generic_chatbot/
    ├── generic_chatbot/
        ├── settings.py
        ├── urls.py 
    ├── chatbot/
        ├── urls.py
        ├── views.py
        ├── models.py
    ├── server/
        ├── engine.py 
    ├── config.json # Custom Chatbot configuration 
    ├── Dockerfile
    ├── manage.py
    ├── Pipfile
    ├── Pipfile.lock
    ├── wait-for-db.sh
├── generic_chatbot_frontend/
    ├── public/
        ├── index.html/
    ├── src/
        ├── utils/
            ├── api.js
        ├── App.js
    ├── Dockerfile.local   
├── docker-compose.yml
├── .env # Replace .env_template with .env and add API keys
├── init.sql
├── README.md

---

Getting Started

Prerequisites

• Docker
• Docker Compose

Setup

1. Branch Information:

   • main branch is the stable production branch
   • staging branch is the development environment branch
   • Other branches are feature branches

2. Create .env file: Copy example.env and add required API for your chosen LLM model:

   cp example.env .env

3. Build and run the containers for the bot:

   make start

4. Access the Application:

   • Run a quick session at http://localhost:3000/
   • Access admin interface at http://localhost:8000/api/admin/ for defining bots and tracking conversations
   • For admin interface credential setup, see the Admin Interface section at the end of this file

5. Available Commands:

   • make start - Start the containers (builds if needed)
   • make stop - Stop the containers
   • make stop-clean - Stop and remove volumes (clean slate)
   • make test - Run all django app backend tests (requires containers to be running)

---

Project Status

Date:26/12/2024

Current Functionality

• Language model selection and bot/prompt management via config.json
• Chat interaction logging (stored in MariaDB).
• Dockerized for easy setup and deployment.

---

Known Bugs

• N/A

---

V1 Next Steps

1. Build and integrate the React-based frontend.
2. Implement user ID tracking and storage (consider user-flow for chat page entry).
3. Create a script to convert SQL conversation data to ConvoKit-formatted JSON.
4. Host the project on AWS EC2.
5. Develop comprehensive API documentation for integration.

---

V2 Future Enhancements

1. Add voice interaction support.
2. Create adaptable UX/UI for various research case

---

Challenges

• HuggingFace Support: Compatibility issues with PyTorch libraries and GPU support in Docker need to be resolved to integrate HuggingFace models.

---

To Be Specced & Open Questions

• How can user authentication and tracking be best integrated with Qualtrics?
• The name of the project Humanlike chatbot does not intuitively describe what this project does do I need to keep this repository name?

Admin Interface

1. Access your Django instance (run inside the backend container):

   docker exec -it humanlike-chatbot-backend-1 bash
   cd /app

2. Create a superuser

   python manage.py createsuperuser
   # follow the prompts to set username, email, and password

3. Log in to the admin panel
   Open your browser and go to:

   https://<your-domain>/api/admin/
   

   (or http://localhost:8000/api/admin/ when running locally)
   Enter the superuser credentials you just created.