CodeFlow AI — VS Code Extension Starter

CodeFlow scans a workspace and creates an interactive, clickable project map. It currently uses lightweight static heuristics; it does not require an AI API.

Features in this starter

• Detects likely entry points such as main.tsx, index.tsx, app.py, manage.py, urls.py, wsgi.py, asgi.py, Program.cs, and Startup.cs.
• Scans JavaScript/TypeScript, Python, C#, Java, Go, Ruby, and PHP source files.
• Finds many static imports, common frontend HTTP calls, common backend route declarations, function declarations, and likely database-access code.
• Displays clickable nodes and connections in a VS Code webview.
• Supports search/filter and rescanning.
• Skips common generated and dependency folders.

Important limitations

This is a usable starter, not a complete compiler-grade call graph. Static heuristics cannot reliably resolve all dynamic imports, dependency injection, reflection, framework-generated routes, runtime-selected code paths, or API URLs assembled from variables. A future version should add language-specific parsers (TypeScript compiler API, Python AST, Roslyn for C#), framework adapters, and an actual graph layout with directed edges.

Run locally

Requirements: Node.js, npm, and VS Code.

1. Extract this project.
2. Open its folder in VS Code.
3. Run npm install.
4. Run npm run compile.
5. Press F5 to open an Extension Development Host window.
6. Open a sample project folder in that new window.
7. Run CodeFlow: Generate Project Flow from the Command Palette (Cmd+Shift+P on Mac, Ctrl+Shift+P on Windows/Linux).

Package as VSIX

Install dependencies first, then run:

npm run compile
npx vsce package

This creates a .vsix package in the project folder. In VS Code, use Extensions → ... → Install from VSIX... to install it.

Commands

• CodeFlow: Generate Project Flow
• CodeFlow: Rescan Project

Shortcut: Cmd+Option+F on macOS, Ctrl+Alt+F on Windows/Linux.

Privacy

The scanner processes workspace source files locally. This starter does not upload code to a cloud service. Do not add cloud AI analysis without making that behavior explicit to users.