Symbio (github.com)

Symbio has announced its powerful, customizable AI assistant that learns interactively through user conversations. This machine learning tool allows users to communicate via a local command line interface (CLI), read, write, and manage markdown notes, and even run sandboxed shell commands and Python snippets. Unique features include automatic detection of user corrections, which are saved as training data, allowing the model to fine-tune itself with LoRA (Low-Rank Adaptation) in real-time. Users can persist conversations in JSONL format and an SQLite database, making Symbio an adaptable and efficient tool for both personal and professional use. This development is significant for the AI/ML community as it showcases an innovative approach to machine learning where the assistant evolves based on user interactions without requiring constant manual input for corrections. Symbio's architecture is built to adaptively store mistakes and improve itself based on user input thresholds, providing a practical application of active learning in everyday use. Developers are encouraged to contribute to enhancements such as CUDA support for broader hardware compatibility and optimizations for LoRA training efficiency, reinforcing the collaborative nature of AI advancements.