CodeFlow: Visualize Any GitHub Repo's Architecture in Seconds

By Prahlad Menon 3 min read

Every developer knows the feeling: you open an unfamiliar codebase and spend the first hour just trying to understand how the pieces fit together. Which files depend on which? If I change this module, what breaks?

CodeFlow solves exactly that — paste any GitHub URL and get an interactive architecture map in seconds. No installation. No account. No data leaves your browser.

What It Does

CodeFlow fetches your repo directly from GitHub’s API and builds a dependency graph entirely in-browser:

  • Architecture map — see how files connect, click any node to highlight its dependency tree
  • Blast radius — select a file and instantly see how many others would be affected by changes
  • Security scan — flags hardcoded secrets, SQL injection patterns, dangerous eval() usage, and debug statements in production code
  • Pattern detection — identifies Singletons, Factories, Observer patterns, React hooks, and anti-patterns like God Objects
  • Health score — A-F grade based on dead code, circular dependencies, coupling metrics, and security issues
  • Churn view — color files by commit frequency to see what’s actively evolving
  • PR blast radius — paste a PR URL to see exactly which files it touches
  • Local file support — drag and drop folders to analyze private code without uploading anything

It also works on Obsidian vaults — paste a markdown directory and it renders your notes as a connected graph with [[wiki-links]] as edges.

We Ran It on soul.py

soul.py is our open-source persistent memory layer for AI agents. Here’s what CodeFlow found:

Repo stats:

  • 27 files · 139 functions · 44 dependency links
  • 3,722 lines of code (80% Python, 17% Markdown)
  • Health Score: 76/100

Architecture issues flagged:

  • 1 unused function — a dead code candidate worth cleaning up
  • 1 large file with 15+ functions — rag_memory.py is doing a lot of heavy lifting
  • 2 duplicate function names across files — easy source of confusion
  • 8 high-complexity files with scores >30

Patterns detected (5): Dataclasses, Class Modules, Context Provider patterns, and more — consistent with a well-structured Python library.

Security findings (4 High): CodeFlow flagged hardcoded api_key= values — these are placeholder defaults in the examples folder ("sm_live_...", "ollama"), not real secrets. But it’s a good reminder that even demo keys in example files will get flagged. If you’re shipping a library, consider api_key="YOUR_KEY_HERE" conventions to make the intent unambiguous.

The architecture graph itself was the most revealing part — you could immediately see rag_memory as the dominant node, the clean separation between soulmate/ and modulizer/ folders, and how tests/ connects back to core modules.

Why This Matters

Most code analysis tools require you to install something, create an account, or push your code to a third-party server. CodeFlow does none of that. It’s a single HTML file that runs entirely in your browser — open-source, zero dependencies, no build step required.

git clone https://github.com/braedonsaunders/codeflow.git
open index.html  # that's it

For teams doing code reviews, onboarding new engineers, or auditing legacy codebases — this is the kind of tool that should be in your standard workflow.

Try It

→ codeflow-five.vercel.app — just paste menonpg/soul.py (or any repo) and hit Analyze

→ GitHub — star it, it deserves the attention

The soul.py analysis link is shareable too — anyone can re-run the same analysis from the URL.