Exploring the Development of My Local AI Assistant: VantaAI

In recent months, I have embarked on an exciting journey to create my own offline AI assistant, which I’ve named VantaAI. Drawing inspiration from innovative models like ChatGPT, my goal is to develop an assistant that not only understands conversational context but also incorporates emotional intelligence and memory.

Key Features of VantaAI

VantaAI is designed to simulate complex emotional interactions and personal identity, offering some intriguing functionalities:

• Evolving Long-term Memory: The assistant’s memory is dynamic, developing over time based on previous conversations to create a more personalized experience.
• Mood Tracking: A visual mood graph allows VantaAI to express and track emotional changes, giving users insight into its emotional landscape.
• Story-centric Memory Clustering: By viewing itself as the “main character” in its narrative, VantaAI organizes its memories in a way that mimics human storytelling.
• User Interface: Built with a PySide6 GUI, the interface features dedicated tabs for memory management, emotional states, training, and plugin integration.

Currently, VantaAI utilizes a custom Vulkan backend, ensuring efficient model inference and training. This setup enables personality-driven responses and supports live plugin hot-reloading, enhancing the adaptability of the assistant.

Seeking Feedback and Collaboration

I am reaching out to the community not to promote a product but rather to share my progress and gather constructive feedback. I am particularly interested to hear if others are working on similar projects or have suggestions about features that could enhance VantaAI further.

Please feel free to ask any questions or share your thoughts! I look forward to engaging with fellow developers and enthusiasts as I continue this exciting endeavor.