Show HN: Coding agent that compiles intent into deterministic DAG before running (github.com)

Rigorix, a new deterministic coding agent, introduces a structured approach to AI-assisted software engineering by compiling natural language development tasks into executable Directed Acyclic Graphs (DAGs). This innovative method separates planning from execution, allowing for repeatable and auditable processes that align with modern CI/CD environments. Rigorix can run independently, enforcing policies and budget constraints while generating code, making it a powerful tool for teams seeking to maintain governance without the constant oversight typically required by traditional models. Significantly, Rigorix addresses the unpredictability of current coding agents that rely on free-form interactions with language models, which can lead to uncontrolled outcomes in automated contexts. By translating intent into a defined DAG, Rigorix ensures that every action is governed by pre-set templates and risk guidelines, providing both structure and audit trails for compliance. It incorporates execution plans validated by proof scripts and delivers a framework that focuses on the 'what' (task structure) while leveraging LLM capabilities for the 'how' (code generation). With support for languages like TypeScript and Python and a focus on automated verification, Rigorix presents a compelling solution for organizations aiming to streamline their development workflows while maintaining strict governance.