a-llama

A fully-embedded Ollama replacement with continual learning: a local LLM daemon speaking an Ollama-compatible HTTP API, with an in-process AstraeaDB graph+vector store and Eunomia semantic cache as its memory subsystem.