cmmchsvc-dev/Galactic-AI

Built around a **Big Brain/Builder two-model architecture**, Galactic AI uses a high-tier Planner model in its own ReAct loop to autonomously scan your codebase and generate step-by-step plans before a leaner Builder model executes them, reducing costs while maintaining intelligence. Memory is truly persistent—learned information writes to disk and is loaded once per session, while optional ChromaDB vector indexing enables semantic search across your entire conversation history and local workspace with automatic context compaction when conversations grow too large. The system integrates with 14+ AI providers (including fully local Ollama), features continuous workspace monitoring for code-aware RAG, includes self-healing test-driven code execution, and exposes everything via a real-time web Control Deck with tabs for chat, live agent tracing, memory editing, and tool/skill management.