🤖 Multi-Agent Sequential Code Reviewer (n8n)

A high-fidelity GitHub Pull Request reviewer built on n8n that utilizes a Smart Sequential Refinement Loop. This project is designed specifically for Open Source Maintainers and Student Contributors to scale mentorship and reduce manual review bottlenecks.

---

🚀 The "Smart Pipeline" Architecture

This workflow implements a three-tier routing logic to handle PRs based on their complexity and size:

🚦 Dynamic Routing Logic (If1 & If2 Nodes)

• Massive PRs (> 200 Changes): Automatically flagged for human intervention with a waitingForHuman label to prevent AI hallucination on overly large diffs.
• Standard PRs (21–200 Changes): Trigger the full 4-Stage Sequential Loop for deep technical auditing.
• Small PRs (≤ 20 Changes): Routed to the Fast-Track Auditor (AI Agent 4) for a supportive, one-sentence verification.

🧠 The Sequential Refinement Loop

1. Stage 1: Initial Auditor (AI Agent 1): Performs the first pass using Step-3.5-Flash to identify logic errors and security risks.
2. Stage 2: Critical Reviewer (AI Agent): Analyzes the first agent's output using Nvidia Nemotron-3 to identify gaps or technical debt.
3. Stage 3: Synthesizer (AI Agent 2): Consolidates findings from previous stages, ensuring feedback is comprehensive without being redundant.
4. Stage 4: Polishing Agent (AI Agent 3): Uses Nvidia Nemotron-12b to humanize the tone and strictly format the output into a concise 3-4 line summary.

---

🌟 Open Source & Educational Impact

• Reducing Maintainer Burnout: Acts as a "Teaching Assistant" by pre-digesting diffs into actionable summaries, allowing maintainers to focus on architectural decisions.
• Instant Feedback Loop: Provides student contributors with immediate technical feedback, keeping the momentum of open-source contributions alive.
• AI-Powered Mentorship: The sequential model explains the why behind a fix, helping students learn better coding patterns through automated peer review.

---

🛡️ Reliability & Security

• Human-in-the-Loop: Before any comment is posted to GitHub, the workflow sends the finalized suggestion via Gmail for manual "Yes/No" approval.
• Universal Compatibility: Uses dynamic expressions to pull repository and owner metadata, allowing the workflow to function across any repository it is connected to.
• Safe-String Processing: A custom JavaScript node cleans code diffs, escapes formatting-breakers (like triple backticks), and removes binary data to ensure prompt stability.

---

🛠️ Setup Instructions

1. Import: Download the github-ai-reviewer.json from this repo and import it into your n8n instance.
2. Credentials: Configure credentials for GitHub (OAuth2), Gmail (OAuth2), and OpenRouter.
3. Webhook: Configure the GitHub Trigger node to listen for pull_request events on your target repository.
4. Deploy: Set the workflow to Active to start receiving automated, high-quality code reviews.