A generalised AI orchestration framework built on LangGraph. Ollama-first — uses local LLMs by default, escalates to cloud APIs only when needed. Thirteen domain skills, composable pipelines, persistent memory, and a knowledge palace system.
OVERMIND coordinates multiple AI capabilities into a single orchestration layer. Individual domain skills handle specific problem areas — audio analysis, security research, hardware debugging, code review — and can be chained into pipelines, gated on confidence thresholds, and run with shared memory context.
The framework is Ollama-first: local LLMs handle the majority of work. When a task genuinely needs more capability, escalation to Claude or other cloud APIs is automatic and transparent.