irrevon 0.1.0

Evidence-first reconciliation for irreversible AI-agent actions

Irrevon

Evidence-first reconciliation for irreversible AI-agent actions.

Irrevon records business intent before dispatch, preserves ambiguous outcomes, and queries destination evidence before any policy-controlled redispatch. It is designed for the failure where an external system commits an action but the response is lost, so a blind retry could duplicate the real-world effect.

Version 0.1.0 is the initial Alpha release. The Alpha label describes software maturity: development evidence is synthetic, the preregistration is not frozen, no confirmatory result exists, and the draft provider adapters have never been live-called. It does not claim universal exactly-once execution, provider qualification, production adoption, or production readiness.

Documentation, quickstart, security policy, limitations, and source:

• Documentation
• Project overview and quickstart
• Current project status
• Security policy
• Changelog
• License

Irrevon is licensed under Apache-2.0.