Skip to content

Introduction

Diffowl is a self-hostable pull-request review system for small, AI-accelerated product teams. Review OWL runs an evidence-backed first pass so a person can focus on product judgment and material findings.

For every eligible pull request, Diffowl:

  1. classifies the execution context before risky work begins;
  2. loads review policy from the trusted base revision;
  3. runs reviewer, challenger, and verifier roles;
  4. normalizes supported claims into material findings or advisory suggestions;
  5. returns a typed outcome that preserves incomplete, limited, and failed states;
  6. publishes actionable findings through GitHub when the trust class permits it.

A material finding claims that a change causes a meaningful problem. It includes a location, impact, evidence, lifecycle state, and verification state. Material findings affect review readiness.

An advisory suggestion is a non-blocking improvement. Advice remains separate so style preferences and optional polish do not masquerade as defects.

Diffowl does not provide:

  • a correctness guarantee;
  • autonomous merge authority;
  • default automated approval;
  • a hosted dashboard or required hosted service;
  • broad coding-agent behavior or automatic patch application;
  • first-class support for GitLab, Bitbucket, or Azure DevOps in V1.