Prompt Vault — Favorites

Browse and Build both have a save_to_favorites toggle. This is the node that lets you get those saved prompts back. It's the smallest of the three in the pack - one action dropdown, one optional index, two string outputs - but it's the glue that makes the "vault" feel like a real library instead of a scratchpad.

The whole thing reads and writes a single prompt_vault_favorites.json file in your ComfyUI root (the folder next to output). Everything persists there across restarts, so a prompt you starred in March is still there in June. Favorites are deduplicated by a hash of the prompt text, so saving the same thing twice doesn't pile up duplicates.

How it works

The action dropdown has three options:

• list - returns an index listing of every saved prompt ([0] (browse) ..., [1] (build) ...). Useful for seeing what's stored and figuring out which index you want.
• load_latest - returns the actual prompt text, ready to feed into a sampler. favorite_index picks which one, and 0 means the most recent save.
• clear_all - deletes the favorites file entirely. There's no undo and no trash; the node will happily delete every prompt you've saved. Do not click this one reflexively.

Outputs are prompt and info. And here's the trap: only with load_latest is prompt something you'd wire into a CLIP Text Encode. With list, the prompt output is a human-readable index listing, not a usable prompt. If you run list and forget to change it before rendering, you'll generate a picture of your favorites menu. The info output is always diagnostic text - the count of stored favorites, or "Loaded favorite #2 (build)". Treat it as status, not prompt material.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/jeremieLouvaert/ComfyUI-Prompt-Vault.git

Restart ComfyUI and it appears under AKURATE/Prompt Vault alongside Browse and Build. Pure Python, no dependencies, nothing to download - the README's "no dependencies" claim is accurate here, unlike some packs that say it and then drag in PyAV.

Common issues

• Feeding list output into CLIP - the classic. Check your action before rendering.
• Favorites saved elsewhere? The prompt_vault_favorites.json lives in your ComfyUI root, and each ComfyUI install has its own. Back it up if you reinstall; the node won't.
• Empty vault returns "No favorites saved yet." and a hint to use save_to_favorites on the other two nodes - which is the actual workflow: star it in Browse or Build, grab it here.