BigPlayer LoRA State

BigPlayer LoRA State is the LoRA half of the pack's state-indication trio. It exists to tell the LLM which LoRAs are in play so the prompt it writes actually targets them. It does not apply LoRAs - that's still your Lora Loader's job, usually via ComfyUI-Lora-Manager - and it doesn't merge anything. It turns what you tell it into context the root can hand to the model.

The inputs that matter

• lora_syntax - the main field. Multiline, and it expects LoRA Manager syntax: <lora:name:0.6> or <lora:name:0.6:0.4> (model strength, then optional clip strength). That's the same tag syntax Lora Manager uses, so if you already speak it, you're home.
• lora_syntax_also (optional) - takes a string or a list of strings from another node and concatenates it onto lora_syntax. Handy when the list comes from a wildcard or a text input elsewhere in the graph.
• lora_stack (optional) - link a LORA_STACK from ComfyUI-Lora-Manager directly, and the node reads the entries (name, model strength, clip strength) from the stack instead of you retyping them.

The output is a single preset_config, chaining into other state nodes and finally into the root's optional preset_config.

The mechanism, and where it's strict

The syntax parsing is a hard regex: it accepts only LoRA Manager tags, and anything else - stray parentheses, bare filenames, random text - raises a clear "invalid LoRA syntax" error rather than silently dropping it. That's deliberate: the README's reliability goals are "fail clearly and predictably," and this node is one of the strictest about it. Entries are deduplicated case-insensitively, and when two sources list the same LoRA, the linked one wins.

When the root renders the preset, the LLM sees something like:

- LoRAs:
  - <lora:furry_fox_v2:0.6>

and the prompt fragments tell it to craft output "tuned for the chosen setup," including LoRA-specific considerations in its comments. That's the difference between an LLM vaguely mentioning a style and an LLM aiming at the style your loaded LoRA actually produces.

The honest position

This is context-only, and it's meant to be. The pack's own roadmap puts LoRA recommendation and application nodes in Phase 2 - this node is the phase-one "let the LLM know what's loaded" step. Don't come here expecting it to stack LoRAs; come here expecting your prompts to stop ignoring them.

Install and troubleshooting

Pack install: ComfyUI Manager, search "BigPlayer Prompting", or git clone https://github.com/josh-ent/comfyui-bigplayer-llm-nodes into ComfyUI/custom_nodes, then restart. Deps are httpx and pydantic; no model downloads. The most common stumble is typing LoRA names with commas or bare filenames instead of <lora:name:strength> tags - the node will tell you exactly that.