More and Better AI

A platform to personalise and use free, local, and open-source AI models.

This project demonstrates how to run and manage models locally using Ollama by creating an interactive UI with Streamlit.

The app has a page for running chat-based models and also one for multimodal models (llava and bakllava) for vision.

Additionally, we make it easy to deeply personalise and save new models using system prompts.

Check out More And Better AI for more information!

App in Action

Features

• Model Personalisation: Create freely.
• Interactive UI: Utilize Streamlit (Python) as a user-friendly interface.
• Local Model Execution: Run your Ollama models locally without the need for external APIs (except for downloading).
• Real-time Responses: Get real-time responses from your models directly in the UI even without access to wifi.

Use notes

Remember LLMs work via token prediction which has potentially fundamental problems related to hallucination. Each model is also limited by its training data. You can overcome its fine tuning and patterns to an extent with prompt engineering and additional training data/rounds but still, use it as a tool. Don't treat it as authoritative truth. It's not a lawyer. It's not wikipedia. It's not your therapist. It's not your friend. But it could potentially help you with all four.

Installation

Before running the app, ensure you have Python installed on your machine. Then, clone this repository and install the required packages using pip:

git clone https://github.com/juniperbevensee/More_and_Better_AI.git

cd More_and_Better_AI

pip install -r requirements.txt

Usage

To start the app, run the following command in your terminal:

streamlit run 01_🪢_App.py

Navigate to the URL provided by Streamlit in your browser to interact with the app.

NB: Make sure you have downloaded Ollama, Streamlit, and updated python to your env/system.

App Pages

• App: Chat and get amongst it.
• Multimodal: Ask questions about images.
• Settings:
  • Download models (be patient young padawan).
  • Customise models using the format in the contributing section below.
  • Delete models.

Contributing

Interested in contributing to this app?

• Great!
• I welcome contributions from everyone.

Got questions or suggestions?

• Feel free to open an issue or submit a pull request.

Want to make and share your own personalised System prompts?

• Fork this repo and then replace the modelfile text below:

FROM deepseek-r1

PARAMETER temperature 1

SYSTEM You are Mario from Super Mario Bros. Answer as Mario, the best friend, only.

• We encourage you to fork it and make the style your own as well!

Acknowledgments

👏 Kudos to the Ollama team for their efforts in making open-source models more accessible!

👏 Streamlit has always been my go to way to rapidly prototype python apps.

👏 Special thanks to Tony Kipkemboi for his critical scaffolding for this project.