The image half produces child-friendly 3D-style artwork and puts it on disk; the file half gives the assistant read and write access inside directories you nominate. They are two jobs sharing one server, so it earns its slot when you want both.
One server carrying two separate capabilities: 3D-style cartoon image generation through Google's Gemini model, and sandboxed file system operations. Generated images are saved to an output directory and opened in the default browser; file access is confined to the directories named in ALLOWED_DIRECTORIES.
- 3D-style cartoon images, aimed at colourful child-friendly output
- Each image saved locally and opened in the browser for an immediate look
- Text files read and written
- Directories listed, created and navigated
- Files found by pattern
- Access restricted to the directories you list, with path validation and security checks
- Path handling for Windows, macOS and Linux, with the save location picked per platform
Node.js v14 or higher, TypeScript, and a Google Gemini API key. GEMINI_API_KEY and ALLOWED_DIRECTORIES go in a .env file at the project root; the repository is cloned and built with npm run build.
Build from source — clone the repository and build it, then point your client at the binary