Core: AI coding with immutable constitution and human quorum (open-source) (github.com)

CORE is an open‑source, production‑grade system for governed AI coding that pairs autonomous agents with an immutable “constitution” to enforce architecture, policy and safety. It converts architecture into a machine‑readable Single Source of Truth (a PostgreSQL-backed Knowledge Graph), wires strict dependency injection and a centralized Service Registry, and runs a stable self‑governance loop: introspection (code → symbolic graph), ConstitutionalAuditor validation, and automated self‑healing (docs, formatting, structural fixes). Agents plan, write and review code but cannot act without pre‑validation against constitutional rules, enabling traceability, auditable change history and reduced architectural drift. For AI/ML teams this matters because CORE tackles long‑standing problems—design rot, undocumented dependencies and unsafe autonomous code changes—by making governance deterministic and automated. Key technical features: database-backed SSOT, deterministic tools for auditing, filesystem and git control, strict lifecycle management, and integration with LLMs (OpenAI/Anthropic/Ollama). It’s runnable today (Python 3.12+, Poetry) from the GitHub repo, with commands for vector syncing, full audits, autonomous repair and conversational coding tasks (e.g., core-admin check audit, core-admin fix all, core-admin run develop). CORE’s roadmap toward “A2” governed code generation and further autonomy frames it as a practical substrate for safe, auditable AI-driven software delivery under constitutional governance (MIT license).