Analyze Product (Mock)

The name is the pitch: this is the no-cost version of AFS_AnalyzeProduct, and it turns your product text into a structured specification without calling any API. No key, no model, no GPU, no surprises. Same contract as the paid node, just cheaper and dumber.

Here's the mechanism, because it's worth knowing how dumb "dumb" is. It takes the product_bundle from Load Product Assets, reads the product_name and product_description, and runs a keyword scan. A few lists live in the code: categories (jacket, dress, shirt, pants, skirt, top), colors (black, white, red, ...), fits (cropped, oversized, slim, ...), sleeve lengths. If your text says "six buttons" it counts them and adds them to must_preserve; "logo" adds logo placement; anything unrecognized falls back to garment and unspecified. It's deterministic - same text, same spec, every time.

Which is exactly the point. The mock spec isn't a product of understanding, it's a contract shaped by your words. If your product description is vague, the spec is vague, and the mock try-on and evaluation stages downstream inherit that vagueness. So there's a genuinely useful habit here: write the description the way you'd write a spec sheet, because that's what this node parses.

Inputs and outputs are minimal. In: one product_bundle. Out: product_specification (the AFS_PRODUCT_SPEC object) and specification_json (the same thing as pretty JSON, which is the honest way to see what keywords were actually caught). Wire the spec into Virtual Try-On Candidates (Mock) and into the live Evaluate Image Candidates if you're mixing mock analysis with a real QA judge - the spec is the thing the judge checks against.

Where does it fit? In the pack's mock vertical slice it replaces the analysis stage so the whole graph runs free. If you're building a real workflow on this pack, it's also a sane staging step: build and validate the plumbing with the mock analyzer, then swap in AFS_AnalyzeProduct (the live one) once you're ready to spend on the real thing.

Installation is the pack's usual story - Manager search "AI Fashion Studio", or clone into custom_nodes - and this node specifically needs nothing extra. It's part of the "Mock MVP" category, sitting under AI Fashion Studio / Mock MVP.

One honest caveat: the keyword list is short, and a modern fashion catalog is full of words it doesn't know. A "puffer vest" becomes category garment, and that's fine for plumbing tests - but don't let a mock spec's neat-looking JSON convince you the real analysis will be equally precise. The live node actually looks at the image. This one only reads your text. Use it to test the graph, not to judge what the pipeline will produce.