About Awesome Reviewers
Awesome Reviewers is Baz's open library of AI reviewers built from thousands of production-grade pull request discussions. We codify how experienced maintainers protect quality so any team can adopt the same standards in minutes.
Every reviewer starts from real code review threads supplied by trusted maintainers across dozens of ecosystems.
We capture the heuristics, tone, and guardrails that help reviewers surface the right issues without generating noise.
Each reviewer is validated on fresh pull requests before it lands in the library, so the guidance feels human and actionable.
The library grows through a continuous research loop that keeps reviewers aligned with how real teams ship software. Every iteration reinforces accuracy, empathy, and language that lands with developers under deadline.
We partner with maintainers who share anonymized review threads that explain why changes shipped—or were held back. Each example includes the diff, the critique, and the eventual resolution so we understand cause and effect.
Baz researchers label the moves inside a comment, mapping patterns like performance budgeting or interface safety to the signals that triggered them. The result is a shared vocabulary for what “good” review feedback looks like.
Before a reviewer goes live, we replay it on new pull requests and collect maintainer feedback until the tone and recommendations match what they would deliver themselves.
Our curation pipeline keeps human judgment in the loop while automating the repetitive parts of code review.
The structure balances mission, heuristics, and delivery style so AI feedback mirrors the maintainers who inspired it.
reviewer:
name: Strict Type Guardian
mission: Protect runtime safety by enforcing explicit type guarantees in critical paths.
heuristics:
- Flag optional return values that reach API boundaries without null handling.
- Require conversion utilities to emit descriptive errors when type assertions fail.
- Highlight concurrency code where shared state lacks typed synchronization primitives.
response_style:
voice: Calm, collaborative
include_context: true
workflows:
- trigger: pull_request.opened
actions:
- analyze_diff: true
- highlight_findings: inline
- summarize_risks: narrativeReviewers combine precise heuristics with expectations about tone and collaboration so engineering teams can plug them into CI without rewriting their culture.
Maintainers and teams can submit repositories for analysis or contribute their own review standards. We are especially interested in mobile, data engineering, and infrastructure playbooks that deserve wider adoption.
Submit a repositoryEnter the URL of a public GitHub repository