BigPlayer Basic Prompt

BigPlayer Basic Prompt is the plain-vanilla consumer module of the pack: it takes the shared session from BigPlayer Natural Language Root and reads out three strings - positive_prompt, negative_prompt, and comments. Nothing more. If you want the simple version of "LLM turns my prose into a usable prompt," this is it.

The comments output is the part people overlook, and it's genuinely the best feature here. The LLM is told to explain its thinking on the prompt construction - model-specific quirks, LoRA-specific considerations - and to add a "Consider" section for anything missing from your setup. So instead of getting an opaque prompt you have to trust, you get a prompt plus a note that says "I wrote this for the checkpoint you're using and here's what I'd add if you had a LoRA for fur texture." That's a rare transparency property for an LLM node, and it's very much in this pack's "no fake magic" spirit.

How it works

The root performs one provider call that covers every attached module. Basic Prompt resolves only its own validated slice (basic_prompt) from the shared session, through a strict pydantic schema - the provider has to return exactly positive_prompt, negative_prompt, and comments, or local validation rejects the whole result. The comments field is even required to be non-empty, so you always get the reasoning.

Wire positive_prompt and negative_prompt into a CLIP Text Encode node ahead of your KSampler, and you're done.

One honest caveat

The negative prompt is only as useful as the model you're wiring it into. On SDXL-lineage checkpoints - Illustrious, Pony, NoobAI and friends - it works, and the LLM is told to craft it for the checkpoint's quirks. On guidance-distilled models that run at CFG 1 (most modern speed/turbo checkpoints), the negative prompt does nothing, because at CFG 1 there's no unconditional pass for it to attach to. The LLM can write you a lovely negative; if your model ignores it, that's an architecture fact, not a node bug. Know which family your checkpoint is on.

Install and troubleshooting

Install is the 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. Dependencies are just httpx and pydantic; no model downloads beyond needing a Grok API key.

If the node errors, the usual causes are session-level, not this node's fault: the root wasn't given a working provider config, or the session you're reading doesn't include a basic_prompt capability. Both error clearly. Attach this node directly to the root's session output and it registers.