Exploring the Development of My Local AI Assistant: VantaAI

As the world of artificial intelligence continues to advance rapidly, I’ve embarked on an exciting journey to create my own local AI assistant, which I’ve named VantaAI. Motivated by the capabilities of ChatGPT, my goal is to build an assistant that operates entirely offline while incorporating elements of emotional logic and memory.

Key Features of VantaAI

I have made significant progress in the development of VantaAI, and I would like to share some of its standout features:

• Evolving Long-Term Memory: VantaAI boasts a long-term memory system that adapts based on ongoing conversations, allowing it to build upon past interactions and provide contextually relevant responses.

• Mood Tracking: The assistant includes a mood graph that visualizes the fluctuations in its emotional state over time, adding depth to interactions and understanding its responses better.

• Narrative-Driven Memory Clustering: VantaAI perceives itself as the “main character” in its own narrative, organizing memories in a way that reflects its unique identity and experiences.

• User-Friendly GUI: I’ve developed a PySide6 graphical user interface featuring tabs dedicated to memory, training, emotional states, and plugin management, making it easy to navigate and customize the assistant.

Currently, VantaAI utilizes a custom Vulkan backend, which allows for expedited model inference and training. It also supports personality-based interactions and live plugin hot-reloading, enhancing its adaptability.

Seeking Your Input

I would like to clarify that this project is not intended for commercial purposes. Instead, I am keen to hear from others who might be working on similar AI initiatives or who have recommendations for features I could explore further.

If you have any questions or ideas to share, feel free to reach out! I’m excited about the potential of VantaAI and look forward to your insights as I continue on this creative journey.