Automate AI Pull Request Review and Decision Memory

Introduction

Speed up code reviews by sending patch changes to an AI, then keep the AI’s decisions stored per PR for easy follow‑up. ⚡

Trigger Event

A webhook receives a pull request patch (the diff) when a PR is opened or updated.

Actions

• The patch is parsed to separate additions from deletions, ignoring file header lines, so the AI sees a clean, structured view of what changed.
• The separated changes are sent to a selected AI review agent, which replies with comments, risks, and suggested actions.
• The AI’s response is stored in workspace memory under a “decisions” list, keyed by the PR’s unique ID, so every update appends and nothing is lost.
• A success response is returned to confirm the review was captured.

Benefits

• ✔️ Faster reviews: AI summarizes risks and recommendations from raw diffs

• ✔️ Cleaner context: Only meaningful additions/deletions are analyzed, reducing noise

• ✔️ Team memory: Centralized “decisions” list per PR for later auditing and coordination

• ✔️ Fewer manual steps: Automatic trigger from webhook; results saved instantly