Symbiotica Prompts Split

LLMs like to hand back one list. Edit nodes like ten sockets. Symbiotica PromptsSplit is the adapter between the two: it takes an enhanced-prompts list and splits it into desc_1..desc_10 STRING outputs, in region order.

This is a pure utility node, and it earns its place because of a specific failure mode the KB documents well: an LLM asked for structured output often returns a preamble, markdown scaffolding, or the wrong shape entirely. PromptsSplit is strict on purpose - it accepts a JSON array of one prompt per region, or numbered lines - so a messy reply fails loudly instead of silently mis-wiring.

How it works

One input, prompts, which is the LLM output: a strict JSON array of prompts in region order, or numbered lines. It parses and distributes to the ten desc_N outputs. Empty slots come out empty, and downstream ERPK nodes treat an empty desc_N as "leave the region's original description untouched." So you don't need all ten regions populated - a four-region template just uses the first four outputs and ignores the rest.

The inputs and outputs

• prompts (required) - the text to split. JSON array or numbered lines, region order.
• desc_1..desc_10 - one per region. Wire them into ERPK's desc_N sockets, or wherever the per-region text goes.

That's the whole node. No optional inputs, no config.

Install

The whole pack installs together:

cd ComfyUI/custom_nodes
git clone https://github.com/symbiotica-ai/comfyui-nodes.git symbiotica
pip install -r symbiotica/requirements.txt

Restart after - or use ComfyUI Manager and search Symbiotica. No key, no model download.

Gotchas

If your LLM replies with Here are the enhanced prompts: before the JSON, this node will refuse it. That's not a bug - it's the guard against chat-scaffolding bleeding into a prompt. Strip the preamble upstream (a prompt that says "output ONLY the JSON array, nothing else" goes a long way), or feed it the numbered-lines format, which is more forgiving of LLM habits. And remember the ordering contract: position in the list is region order. If your template has nine regions and the LLM returns nine prompts, desc_9 is the ninth region - get the order from the template right, and the sockets stay honest.