Prompt Manager Browser

This is the front door of the pack: a full prompt-library web UI, rendered inside a node on your canvas. Most prompt managers hide their library behind a separate window or a browser tab; this one gives you a browsable, searchable folder tree of prompt cards that stays in the workflow with you. It's a UI-only node - no inputs, no outputs, doesn't participate in execution - so think of it as a docked panel rather than a step in the graph.

What you get in the panel

Prompts live as a normal folder tree of JSON cards (which can hold prompts plus LoRA/model/sampler settings) and TXT wildcard files, backed by real folders on disk. That's the point: because the library is just files, you back it up by copying a folder and share it by sending a folder. From the browser you can:

• browse, search, and edit cards and wildcards;
• import/export the library, including a bundle format (pm4a-prompt-bundle) for batch import;
• generate preview images for cards - the pack ships a built-in API workflow for this, and you can drop in your own api.json to replace it;
• align card model names to local files by hash after a library refresh;
• drop an image onto a new card and load generation settings from it (with an overwrite confirm), or push a card's models/LoRA/parameters into the canvas in one click.

The panel remembers your last folder and sidebar expansion across reloads, which is the kind of small quality-of-life thing that tells you the author actually uses it.

Where the scheduler meets the library

The browser is the input side of the loop; the Prompt Scheduler is the output side. In the scheduler you reference library items with wildcard syntax (__folder__ / __path/file__), and the scheduler expands them at queue time - including, if you enable it, auto-applying the card's LoRAs and generation settings to matching nodes before the run and restoring the canvas after. So the typical rhythm is: browse and maintain prompts in this panel, then drive batches from the scheduler without leaving the canvas.

Installing it

Ships with the pack:

cd ComfyUI/custom_nodes
git clone https://github.com/tsukino4a/ComfyUI-4A-Prompt-Manager.git
cd ComfyUI-4A-Prompt-Manager
python install.py   # or: pip install -r requirements.txt

Restart ComfyUI after - the restart matters here, because the browser is a web app served by the pack's own HTTP routes (under /pm4a), and those only register at startup. ComfyUI Manager users: search 4A Prompt Manager. No models to download; the only Python extra is piexif, and that's for Image Saver, not the browser. Sample cards ship under wildcards/examples/ so there's something to look at on first launch.

Troubleshooting

• Blank/empty panel after install. You didn't restart ComfyUI, so the /pm4a routes aren't registered. Restart.
• "Browser interface files missing" error. The pack's web/ assets aren't where the server expects them - usually an incomplete clone or a manual install that skipped files. Re-clone cleanly.
• You tried to wire it to something. Don't. There are no sockets; it's a viewer. Anything you want to pull out of the library goes through the Prompt Scheduler instead.
• Preview generation picks the wrong model. It uses the bundled workflows/default_api.json; replace that file with your own API workflow if you want previews built your way.