[Book Tools] Prompt Selector

BTPromptSelector is the picker half of the Book Tools prompt-batch pair. Its sibling, BTPromptSchedule, turns a list of page descriptions into a dictionary; this node pulls one entry out of that dictionary and hands it to your sampler as a plain string. If you're rendering a book one page at a time - cover, title page, chapter one, chapter two - that's the whole workflow in two nodes, and this one is where you choose which page you're on.

How it works

The node takes two inputs: dictionary (the DICTIONARY output from BTPromptSchedule) and selected_indexes. It looks up one key and returns whatever string lives there. That's it - no prompting, no fancy merging, just an indexed read from a dict.

And here's the trap that will cost you twenty minutes: the default value for selected_indexes is literally "1,2,3". That reads like a range, and the docstring even talks about "indices and ranges," but the actual code does a single lookup: dictionary.get(str(selected_indexes)). So with the default untouched, it goes looking for a key called "1,2,3", finds nothing, and returns an empty string. Your sampler runs with a blank prompt and you wonder what broke. Set it to a single number - 1, 2, 3 - and it works exactly as expected.

The one output that matters

The STRING output is what you wire into a CLIPTextEncode's text input (positive or negative, whichever side you scheduled). Because it's a plain string, you can also pipe it into anything else that eats a prompt - a LoRA stack's prompt side, a text concat node, whatever.

One structural note: DICTIONARY is a custom type, not a core ComfyUI type. That means BTPromptSchedule's output only plugs into inputs that also expect DICTIONARY - and in practice, that's BTPromptSelector's dictionary input. They're designed as a matched pair, which is convenient and also means the dictionary is useless to other node packs.

Installing

ComfyUI Manager is the easy route: Manage → Install Custom Nodes → search "ComfyUI-Book-Tools", install, restart. Or by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/Big-Idea-Technology/ComfyUI-Book-Tools
# then restart ComfyUI

The whole pack is pure Python - Pillow, torch, numpy, all already in your ComfyUI environment. No model files, no weights to download, nothing heavy. (The README mentions a requirements.txt; the repo doesn't actually ship one, and it doesn't need to.)

Common issues

The empty-output-with-default problem above is the big one, and it's easy to confuse with a broken prompt. Beyond that: if the dictionary is empty or you ask for an index that doesn't exist, you get an empty string back rather than an error - so validate by previewing the output with a Show Text node before you burn a batch. For a real book workflow, this node pairs naturally with the BTLoop iteration counter as your page index, so each loop pass renders the next page's prompt.