I built an AI code reviewer solo for 3 months. Here's what I learned and why I'm open sourcing it.

Source: Dev.to

Problem

A PR gets opened, sits for a couple of days, someone glances at it, types “LGTM 👍” and merges it. A few days later production is on fire because of a SQL injection vulnerability that a proper review would have caught in seconds. This happens everywhere—every team, every company—and everyone knows it’s happening.

What I built

DevPulse connects to your GitHub or GitLab repository, selects any PR, and runs a full AI review in about 30 seconds. It provides concrete, actionable feedback such as:

• “Line 47 — SQL injection vulnerability, here’s the snippet, here’s how to fix it.”
• “You’re making a database call inside a loop — this will kill your server under load.”
• “This API key is hard‑coded — you think you’ll fix it later, you won’t.”

Every developer on your team receives a quality score and grade (A+, A, B, C) based on real issues found in their PRs over time, letting you see who consistently writes clean code and who slips past LGTM approvals.

You can click any developer to view their full profile—live commits from GitHub/GitLab, PR history, contributed repos, categories of issues they introduce, and severity breakdown.

A full repo scanner also traverses your entire codebase (not just PRs) to find vulnerabilities, outdated dependencies, and security holes that have been sitting there for months.

The tech stack

• Backend: Python, Django, Django REST Framework
• Frontend: React + Vite, custom design system built from scratch
• AI: Google Gemini API
• Integrations: GitHub API & GitLab API (both fully supported)

What I learned building this solo

Building alone is fast at the start but brutal at the end. The first month I shipped features daily; by month three I spent more time context‑switching between backend, frontend, AI prompts, and API integrations than actually building.

The hardest part wasn’t the code—it was making product decisions alone. Every feature, design choice, and API design had no one to bounce ideas off, leading to constant second‑guessing.

It taught me to ship. When you’re solo there’s no “let me check with the team.” You decide and build.

Why I’m open‑sourcing it

One person can only take something so far. The product works, the idea is real, and the problem is real. What it needs now is more people.

I’m looking for contributors who want to help shape this into something truly useful—not just fix bugs but define its future. Backend, frontend, DevOps, design, AI/ML—any help is welcome. If you’re a developer tired of LGTM culture, clone it, run it, break it, and tell me what’s wrong.

The code

• Backend →
• Frontend →