Skip to main content

Tabs & Persistence

Multi-tab workspace​

CodefyUI supports multiple independent canvases as tabs. Each tab has its own:

  • nodes, edges, and layout
  • execution context and logs
  • recorded outputs and persisted weights (see Teaching Inspector)
  • undo/redo history (up to 50 steps)

This lets you keep several experiments side by side — for example a training graph in one tab and an inference graph in another — without their state interfering.

Automatic saving​

All tabs are auto-saved to your browser's localStorage, so your work is restored when you reload the page. This is local to the browser; it is not synced to the server.

Import / export​

You can export any graph to a JSON file and import it back later (or share it):

  • Export writes the current tab's graph (nodes, edges, parameters, and segment markers) to a .json file.
  • Import loads a .json graph into a new tab.
  • Export as Python writes a single-file headless runner that embeds the executable graph. Run it with the Python environment from a compatible CodefyUI installation; it reuses the same nodes and graph engine without starting the web server. Use --help for device, GraphInput JSON, timeout, and project-asset options.

The same JSON format is what the backend's example graphs use, so an exported graph can also be run headless with the CLI Graph Runner.

tip

Because graphs are plain JSON, they diff and version-control cleanly. Commit a graph alongside your code to capture an exact, reproducible pipeline.