Show HN: LYRN Context Management Dashboard (github.com)

LYRN (Live-reasoning & Structured Memory) is a modular, professional-grade local LLM GUI that prioritizes persistent, structured context over repeated prompt injection. Built with CustomTkinter and an asynchronous UI, it runs GGUF-format models locally and exposes full model controls (temperature, top_p, top_k, stop generation, presets). Its architecture is file-based: personality, episodic chats, topic indexes, and memory tables live as structured text files; a background topic-indexer and a “Quote to Context” feature turn chat fragments into searchable, long-term memory. LYRN also includes live system gauges (CPU/RAM/VRAM/Disk), a dynamic two-panel system-prompt builder (component toggling, reordering, wrappers), and automation tooling—job manager, calendar scheduler, cycle manager—and support for gpt-oss “harmony” tool definitions. For the AI/ML community this matters because LYRN emphasizes cognitive continuity and reproducibility on consumer hardware, offering an alternative path to improving agent behavior without chasing larger models. Its structured-memory-first approach enables persistent identity, deterministic context injection, and easier auditing/debugging of agent state—useful for research into long-term agents, tool chaining, and on-device deployment. It’s extensible (file-driven tools, theme builder) and easy to run (pip install -r dependencies/requirements.txt, place GGUF models in a models/ folder, run lyrn_sad_v4.2.6.pyw), but comes under a custom source-available license that allows non-commercial use only unless commercial permission is granted.