CodeLens — Offline API Documentation Generator

Automatically scan your codebase and generate API documentation, architecture docs, and runnable test suites — entirely offline.

CodeLens is a VS Code extension that performs deterministic AST analysis on your API endpoints, enriches them with data-flow analysis, and optionally uses a locally-running LLM to generate human-readable documentation and test cases. No data ever leaves your machine.

✨ Features

🏗️ Supported Frameworks

🚀 Quick Start

1. Install the Extension

Install from the VS Code marketplace, or build from source:

git clone <repository>
cd codelens
npm install
npm run build

Then press F5 in VS Code to launch the Extension Development Host.

2. Scan Your Workspace

• Open a project containing API endpoints
• Press Ctrl+Shift+L S (or Cmd+Shift+L S on Mac)
• Or use the Command Palette: CodeLens: Scan Workspace

3. View Results

• Open the CodeLens panel in the Activity Bar to browse endpoints
• Run CodeLens: Generate Documentation (Ctrl+Shift+L D) to generate docs
• Run CodeLens: Generate Tests (Ctrl+Shift+L T) to generate test suites

4. (Optional) Enable LLM Enrichment

For richer documentation with summaries, descriptions, and curl examples:

# Install Ollama (https://ollama.ai)
ollama pull deepseek-coder-v2:16b

The extension will automatically detect Ollama at localhost:11434.

📁 Output Structure

All generated files are placed in the .codelens/ directory:

.codelens/
├── pipeline.json          # Raw scan data (useful for debugging)
├── openapi.yaml           # OpenAPI 3.1 specification
├── README.md              # Overview with endpoint index
├── endpoints/
│   ├── GET_users.md
│   ├── POST_users.md
│   └── GET_users_{id}.md
├── schemas/
│   ├── UserResponse.md
│   └── CreateUserDto.md
└── pytest/                # (or jest/ for TypeScript projects)
    ├── conftest.py
    └── test_app.py

⚙️ Configuration

All settings are available under Settings → Extensions → CodeLens:

⌨️ Keyboard Shortcuts

🔒 Security

CodeLens is designed with data security as a first principle:

• Zero network calls — the extension never phones home, sends telemetry, or makes any outbound requests
• No raw source code to LLM — Tier 3 only sends structured JSON metadata to the local Ollama instance, never actual source code
• Local-only LLM — Ollama runs on localhost with no external API calls
• Deterministic fallback — if Ollama is unavailable, Tiers 1+2 still produce complete, useful output

🏛️ Architecture

The pipeline has three tiers of analysis:

┌─────────────────────────────────────────────────────────┐
│                    VS Code Extension Host                │
│  ┌──────────┐  ┌───────────┐  ┌──────────────────────┐  │
│  │ Commands  │  │ TreeView  │  │   Status Bar         │  │
│  └─────┬────┘  └───────────┘  └──────────────────────┘  │
│        │                                                 │
│        ▼         IPC (JSON messages)                     │
│  ┌──────────────────────────────────────────────────┐   │
│  │              Child Process (Worker)                │   │
│  │  ┌───────┐  ┌────────┐  ┌────────┐  ┌────────┐  │   │
│  │  │Tier 1 │→ │Tier 2  │→ │Tier 3  │→ │Output  │  │   │
│  │  │AST    │  │DataFlow│  │LLM     │  │Gen     │  │   │
│  │  └───────┘  └────────┘  └────────┘  └────────┘  │   │
│  └──────────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────────┘

• Tier 1 (Deterministic): Pure AST parsing extracts endpoint definitions
• Tier 2 (Deterministic): Inter-procedural data-flow analysis resolves types and detects side effects
• Tier 3 (Optional): Local LLM generates summaries, descriptions, and test cases from structured JSON

🛠️ Development

Prerequisites

• Node.js 18+
• npm 9+
• VS Code 1.85+

Building

npm install
npm run build        # Production build
npm run dev          # Watch mode for development

Testing

npm test             # Run unit tests
npm run test:integration  # Run integration tests

Packaging

npm run package      # Create .vsix package

📜 License

MIT