Nodes/Prompt DB/Prompt DB Manager
ComfyUI Node

Prompt DB Manager

Your prompt library finally lives inside the graph

By HalfikCodeΒ·Created 5 months agoΒ·Updated 5 months agoΒ· 0
Prompt DB Manager

      The Prompt DB Manager looks like a scam when you first drop it in: no inputs, no outputs, nothing wires anywhere. It's not a processing node - it's a doorbell. Click the "πŸ“‚ Open Manager" button and your browser opens the web UI for the whole Prompt DB pack, which is where you build the library that the rest of the nodes read from. If you've been living with a notepad of prompts or re-typing the same character description into a text box for the hundredth time, this is the node that fixes it. It's the front door: nothing in this pack works until you've populated a database through it.

      The workflow it feeds is genuinely nice once you've seen it. You create categories (the pack ships with Location, Character, Style, Mood ready to go) and save prompts into them, attaching LoRAs with per-LoRA strengths right inside the UI. Then Prompt Selector nodes each pick one prompt from a category, a Prompt Combiner joins them, and Apply Lora Stack lays the combined LoRA stack onto the model. Your whole style system becomes a set of dropdowns instead of a wall of text.

      How it works

      Technically it's a stub of a node: PromptDBManager is an output node that takes a hidden unique_id and returns nothing. The real work happens in the pack's API layer. On import, the pack registers routes on ComfyUI's built-in PromptServer - /prompt_db/manager serves the HTML page, and a set of REST endpoints (/prompt_db/prompts, /prompt_db/categories, /prompt_db/loras, and POST/PUT/DELETE routes) do the CRUD. The web UI talks to those, and everything lands in a single prompts.json in the pack directory. No database, no server to configure, no dependencies.

      The UI itself punches above the pack's weight: categories, a slide-out editor, search across names and text, drag-and-drop reordering, grid/list views, and keyboard shortcuts (Ctrl+N new, / search, Esc close). Two things are worth calling out:

      • LoRA auto-discovery. The editor scans your ComfyUI loras folder, so attaching a LoRA to a prompt is a dropdown pick, not a filename you type from memory.
      • Export/Import. You can dump the whole database to JSON and import it back with Merge (append new, skip duplicates) or Overwrite. This is your backup button, and you should treat it as a habit.

      Installing

      Search "Prompt Database" in ComfyUI Manager and hit install, or clone it manually:

      cd ComfyUI/custom_nodes/
      git clone https://github.com/HalfikCode/ComfyUI-Prompt-DB.git
      

      Restart ComfyUI and add the node from the Prompt Database category. No requirements.txt, no model downloads - it's all stdlib JSON plus ComfyUI's own server. One of the least fussy installs in the ecosystem.

      Common issues

      • Your prompts live in the pack's folder. prompts.json is stored inside custom_nodes/ComfyUI-Prompt-DB/. A git pull to update the pack won't delete it, but removing the folder to reinstall will wipe everything. Export a backup before you touch the pack.
      • Corrupted file = reset to defaults. The code handles a malformed prompts.json by quietly writing a fresh default database (the four starter categories, no prompts). So if you hand-edit the file and botch the JSON, you don't get an error - you get a reset. Back up first, and use the UI's Export rather than editing the file by hand.
      • It's an output node with no wires. It has zero outputs by design, so if ComfyUI ever complains it's unconnected, ignore it - that's expected. Just place it somewhere out of the way in your template workflow and let it ride.
      CategoryPrompt Database

      Inputs (0)

      No inputs

      Outputs (0)

      No outputs