GROQ Document Analyzer

The one honest sentence up front: this node does nothing. GROQ Document Analyzer looks complete - five analysis modes, a structured output, the works - but its analyze_document() method is an empty pass in the source code. There is no implementation. It's the most tantalizing shell in the ComfyUI_ComfyGroq pack (downlifted/ComfyUI_GROQ-PromptWizard) because the design is actually good, and none of it is wired up.

What it's for

Documents in a ComfyUI workflow are rare, but not unheard of: an LLM's captioned output, a model card's README, a licensing text, a transcript you want condensed. The node's plan is simple - paste text in, pick an analysis_type, get useful output. The five modes are summarize, extract_key_points, qa, sentiment, and entities. That's a well-chosen set: summary, bullet extraction, question-answering, tone, and named-entity spotting covers most of what anyone actually wants from a document.

What the schema says

Inputs: api_key (or GROQ_API_KEY), document_text (a multiline field - paste or wire your text in), analysis_type, temperature (0–1, default 0.3, stepped at 0.05), and max_tokens (up to 4096). Outputs are the interesting part: analysis is a STRING with the human-readable result, and extracted_data is a LIST - the structured payload, presumably the entities or key points as machine-readable data you could pipe into other nodes. That LIST output is the feature that would make this node worth having; most LLM text nodes only spit out a string.

Install

Standard for this pack: ComfyUI Manager, search "GROQ" or "ComfyGroq", or clone https://github.com/downlifted/ComfyUI_GROQ-PromptWizard into custom_nodes. Dependency is one line - pip install groq - and you need a GROQ API key from console.groq.com. Nothing to download; it's a cloud API.

The reality check

Three of this pack's nodes - this one, the Audio Processor, and the Code Assistant - are stubs with pass bodies, and the whole repo sits at a single commit with zero community usage behind it. If you need document analysis now, any LLM chat or API call does this, though you'd be hand-rolling the structured output yourself. Watch the pack's repo for a commit that actually implements analyze_document - the design is sound enough that a finished version could slot neatly into a pipeline that processes text alongside images. Until then, this is a spec sheet, not a tool.