Genie Assistant

Genie Assistant is a lightweight Visual Studio Code extension that helps generate documentation for codebases using a language model. It scans files in the open workspace, sends contextual prompts to a model, and writes the model-produced documentation into workspace output files.

Primary goals:

• Make it easy to generate per-file documentation from natural-language prompts.
• Provide a clear demo of integrating VS Code chat participants and language models.
• Keep the implementation concise and easy to adapt for demos or forks.

Highlights

• Workspace scanning and file discovery: utils/fs_scanner.js.
• Model invocation and streaming capture: utils/model_pipeline.js.
• Lightweight logging: utils/scribe.js (append-only timestamped log).
• VS Code command and chat participant: genie.processWorkspace and genie.assistant.

Requirements

• Node.js (v16+)
• Visual Studio Code (recent stable release)

Installation

npm install

Usage (quickstart)

1. Open the repository folder in VS Code.
2. Run the command palette action: Process Workspace for Documentation (genie.processWorkspace).
3. Enter a prompt describing the documentation you want (e.g., "Create a short developer-facing README and usage examples").
4. Generated documentation files are written to the genie_outputs/ folder at the workspace root. Workflow events and errors are appended to genie-log.txt.

Notes on behavior

• The extension keeps chat history in memory and builds a small prompt that includes the target file's content before sending it to the configured language model.
• Outputs are versioned: if genie_outputs/<filename>.md already exists, the code will create _<n> suffixes to avoid overwriting.

File structure (important files)

genie-assistant/
├── extension.js                # VS Code extension entry (chat participant + command)
├── utils/
│   ├── fs_scanner.js          # Recursively find workspace files
│   ├── model_pipeline.js      # Send file contents to model and save outputs
│   └── scribe.js              # Timestamped logging helper
├── images/
│   └── genie_banner.svg       # Banner for social/marketing
├── scripts/
│   ├── GENIE_BANNER_EXPORT_INSTRUCTIONS.txt
│   └── merge_banner.js        # (optional) script to composite images using sharp
├── test/
│   └── extension.test.js      # Simple, local-run tests
├── package.json
├── README.md
└── LICENSE

Development

• Run lint:

npm run lint

• Run the local test script (no VS Code integration required):

npm test

Exporting the banner image

If you need PNG/JPEG versions of the banner, see scripts/GENIE_BANNER_EXPORT_INSTRUCTIONS.txt. Common commands use ImageMagick, Inkscape, or Node + sharp. Example (ImageMagick):

# convert SVG to PNG (1200x627)
magick convert images\genie_banner.svg -background none -resize 1200x627 images\genie_banner.png

Security & privacy

• The extension reads files from the workspace and sends content to a configured language model. Do not use it on sensitive or private repositories without reviewing the model provider's data handling policy.

Contributing

• Open issues or PRs with focused changes. Keep changes small and explain your motivation.
• If you add features that call external services, document required environment variables and auth steps.

License

This project is available under the MIT License — see the LICENSE file.

Contact / attribution

If you'd like me to prepare a short LinkedIn post and packaging (zip + images) for you to share, I can produce that next.

Example prompts

Here are a few sample prompts to try with Genie Assistant:

• "Create a short README for this file explaining purpose, usage, and example code snippets."
• "Summarize the responsibilities of the functions in this file and list potential edge-cases."
• "Generate unit test ideas for the following code and include one example test."
• "Write a high-level architecture summary for this project suitable for new contributors."
• "Produce a short changelog entry describing recent refactors and why they were done."

Troubleshooting

• "No models found" or model selection errors:

  • Ensure you have configured a language model provider in your VS Code environment. Some VS Code language model APIs require vendor-specific setup.

• Network or authentication errors when calling the model:

  • Check network access and any API keys required by your model provider; avoid exposing sensitive keys in the workspace.

• Permission denied when writing to genie_outputs/:

  • Ensure your workspace is writeable and that no external process is locking files. On Windows, run VS Code as the same user.

• Large repositories are slow to scan:

  • Exclude large directories (node_modules, build folders) from your workspace or adapt utils/fs_scanner.js to skip additional paths.

• Tests fail locally:

  • Run npm run lint first to catch code-style issues. The included test is a minimal sanity check; more robust integration tests require a configured VS Code test environment.