Hey Product Hunt! 👋

I'm thrilled to share Prism Reviewer AI with the community today! 🌈

### 💡 Why We Built This

As developers maintaining fast-moving repositories, we noticed major flaws with first-generation AI code review tools:

1. Single LLM prompts get overwhelmed by large PR diffs, resulting in shallow, surface-level nitpicks.

2. AI agents frequently hallucinate line numbers on files that weren't even modified in the PR.

3. Re-running full PR scans on every single commit push burns through API token budgets fast.

We asked ourselves: What if we treated code review like optics?

When white light passes through a prism, it disperses into distinct wavelengths. Prism Reviewer AI does the exact same thing to git diffs:

- 👮 Warden focuses exclusively on Security, Credentials & Compliance.

- 📐 Architect focuses on System Design, Performance & Architecture.

- 🔍 Inspector focuses on Code Hygiene, Logic Bugs & Edge Cases.

### 🛡️ Built for Developers & CI Workflows

- 🎯 Zero Hallucinations: Every single line reference is strictly validated against git diff index coordinates.

- ⚡ Smart Incremental Review: Automatically switches to delta review on push updates, reducing token costs by up to 90% while preserving full PR context.

- 🤖 Multi-Model Flexibility: Powered by LiteLLM — run with Gemini 3.1 Flash, Claude 3.5 Sonnet, GPT-4o, or OpenRouter models.

- 📦 100% Open Source: MIT licensed and ready to drop into GitHub Actions with just 2 lines of YAML.

🔍 Live Demo PR: Want to see it in action? Check out real Prism Reviewer comments on our [Live Demo Pull Request #35](https://github.com/aravinthan-n/savourly-recipes/pull/35).

### 🎁 We'd Love Your Feedback!

Prism Reviewer AI is built for the developer community, and we'd love to hear your thoughts:

- What features or language ASTs would you like to see next?

- How do you currently handle automated PR reviews in your team?

Drop your questions and feedback below — we'll be here all day responding! 🚀