Q00/ouroboros

Project overview

By introducing a mathematical ambiguity gate that blocks seed generation below a clarity threshold and a 3-stage evaluation pipeline, ouroboros addresses architecture drift and manual QA at the input stage rather than after code is produced.

Project type

AI Agent

Deployment

Refer to project documentation

License

MIT

Best for

• Developers and AI engineers who want to enforce specification clarity before AI code generation begins
• Teams experiencing architecture drift and manual QA overhead from non-deterministic AI coding agents

Key capabilities

• Implements an interview, crystallize, execute, evaluate, and evolve loop.
• Forces clarity before code by exposing hidden assumptions in user intent.
• Blocks code generation until the weighted clarity ambiguity score is 0.2 or lower.
• Verifies output mechanically, semantically, and via multi-model consensus.
• Iterates on generations until ontology similarity is 0.95 or higher.
• Provides on-demand agents including Socratic Interviewer, Ontologist, Hacker, and Simplifier.
• Abstracts execution across various AI coding CLIs and runtimes.
• Records actions as ledger events using SQLAlchemy and aiosqlite.

Limitations and risks

• Requires Python >= 3.12.
• Reliance on external LLMs incurs variable inference costs for operations such as Multi-Model Consensus. A paid external LLM service is required.

Getting started

• Run the curl install script, then type 'ooo' in an AI coding agent. Requires Python >= 3.12. The installer auto-detects available runtimes.

Evidence and sources

Open original GitHub project