NodeOne — Changelog

Formalized the split between coding, strategy, and utility agents. Messages can now be routed based on task type, reducing cognitive overhead in sessions.

Laid out the initial dashboard and routing interface in the NodeOne GUI. Included placeholder views for agent routing, coding panel, and memory monitor.

Introduced an ethics file guiding behaviour boundaries. This keeps NodeOne aligned with intended use cases and ensures safer local experimentation.

Created a code-writer agent that can draft scripts or code blocks and write them directly into local files, enabling rapid prototyping workflows.

Gave NodeOne the ability to pause, critique its own outputs, and adjust follow-up actions before finalizing results — an early “self-improvement” loop.

Implemented a dual memory approach: short-term scratchpad for immediate context and long-term SQLite store for continuity across sessions.

Initial integration of NodeOne with the locally hosted LLaMA-3 model. Confirmed the backend could accept prompts and return completions within the GUI shell.