Productivity

Kōbō PLM Fashion Tools

by Gaia Tek Limited (Kōbō)

Overview

Kōbō PLM Fashion Tools helps fashion founders and designers get a garment from idea to factory without knowing the jargon first. Tech pack builder: describe a garment or upload a sketch and get a factory-ready tech pack with a graded measurement spec across your sizes (UK, US, EU, letter or waist-inch sizing, in cm or inches), tolerances and how to measure each point, materials with an amount per garment, construction, colourways, artwork and label placement, care and packing. A plain-English checklist shows what the factory still needs. Edit any measurement in the panel, add photos of fabrics, trims and artwork, and download a PDF or spec CSV. Revisions: ask for a change ("add size XXL", "chest 1 cm bigger for the PP sample") and the tech pack moves to the next revision with a dated change log and the sample stage on every page. Technical flats: no sketch? Ask for front and back flats to be drawn, with numbered callouts matching the construction notes. Sample fit review: tell ChatGPT what your sample measured. Each point is checked against the spec and its tolerance, with the biggest miss, what to tell the factory and which sample to make next. Works on its own or against a tech pack. Download the fit report as PDF or CSV. HS code lookup: describe a garment or accessory and get the likely 6-digit customs code with the reasons, close alternatives, and links to the US, UK and EU tariffs. Landed cost: from a factory price to landed cost per destination market, with freight, duty, other costs, wholesale and retail margins, and suggested prices. No account is needed. Images and tech packs are kept for up to 7 days so you can keep working on them, then deleted.

Tools

apply_fit_review

ChatGPT

Writes the fit review's New values into the tech pack as the user's own measurements, as the next revision at the next sample stage. Called by the fit review panel's "Apply to tech pack" button.

attach_image

ChatGPT

Stores an image the user added from the tech pack panel (a flat, a material photo, artwork or a close-up) and returns the updated tech pack. Called by the tech pack panel.

build_tech_pack

ChatGPT

Builds a garment tech pack for a factory: header, graded measurement spec (points of measure per size), bill of materials, construction, colourways, care and packing, with a checklist of what is still missing. The user sees it as an interactive panel and can download a PDF. BEFORE CALLING: you must know what the garment is and who it is for (womens, mens, unisex or kids) from what the user has said in this conversation or shown in an uploaded sketch. If they have not said, for example they only wrote "make me a tech pack", do NOT call this tool and do NOT pick a garment from memory or earlier chats. Instead ask, in one short message: what the garment is and its key details, who it is for, the main fabric, the size range and unit (cm or inches), colourways, and whether they have a sketch or measurements to share. If they then want you to go ahead with gaps, call it; the checklist shows what is missing. FILLING IT IN: use what the user gave you and what is clearly visible in their sketch or document, and write it the way a production tech pack would. Never invent supplier names, supplier references or Pantone codes. You may propose what is standard for the garment, and say in your reply that these are suggestions to confirm: (1) construction per area with stitch type and ISO class, stitches per inch/cm, seam allowance and finish (e.g. "Side seams: 4-thread overlock ISO 504, 1 cm SA, 12 SPI, pressed to back"); (2) a full BOM: shell, lining/pocketing, every trim the design implies (buttons with ligne size, zips with length and type, elastic with width, interfacing), thread, main/size/care labels, and packaging (polybag, hangtag, swing tag string, size sticker); (3) an estimated consumption per garment for each BOM line; (4) care and packing. MEASUREMENTS: if the user gives measurements, pass them in poms exactly as given (use values for a full size table, or base plus grade). If they give none, omit poms and Kōbō's starter block for the garment is graded across their sizes. Each result lists where every row's numbers come from. On a rebuild (with pack_id) the previous measurements are kept, so re-pass poms ONLY for rows the user supplied or edited; leave out Kōbō starter rows so they regrade from the block (for example across a new size) and stay marked as starter, and never copy starter or estimated numbers into poms as if they were the user's. Use add_poms for points of measure specific to this design (pocket opening and position, slit or vent length, shirring panel, cuff or hood details, strap length) and drop_poms for block points that do not apply (e.g. sleeve measurements on a sleeveless dress). For design details with well-known typical sizes (pocket dimensions, placket width, strap length, rib heights) give a typical value with base and grade (0 if it does not grade) and set estimate: true; never set estimate on the user's own measurements. Do not add a point that is already in the block under another name. IMAGES: put every image the user uploads in this conversation in uploads (each one once), then point at it by its position with {"upload": n} where it belongs: images.front and images.back (flat drawings or photos of the whole garment), images.detail (close-ups, with a short caption), bom[].image (a photo of that material: fabric swatch, button, zip, label), or artwork[].image (a print, embroidery, patch or label design; fill name, kind, placement, size and technique, and for labels the text, including "Made in …" if the user gave it). Work out which image is which from what the user said and what the image shows; ask only if you really cannot tell (e.g. two near-identical flats with no front or back cue). An upload you do not place is kept as a close-up photo. KEEPING IMAGES: images are stored by Kōbō and have an image_id (listed in each result). When you call this tool again, pass every existing image back as {"image_id": "…"} in the same place. Never drop an image on a rebuild, and never re-send an old upload instead of its image_id. CHANGES: when the user asks for any …

calculate_landed_cost

ChatGPT

Calculates the landed cost of an imported garment or accessory per destination market (factory/FOB price + freight + insurance + import duty + other costs, in the market currency), the wholesale and retail margins at the user's prices, and suggested prices for target margins. The user sees an interactive breakdown they can edit. BEFORE CALLING: you need the factory/FOB price and its currency, the country of manufacture, and at least one destination market. If any of these is missing, ask for it first (one short message) instead of guessing. Prices and target margins are optional. DUTY: you must supply duty_percent for each market. Work it out from the HS code (infer the 6-digit code from the product and fibre content if the user does not know it), the destination tariff and the origin, including additional tariffs on that origin (for example extra US tariffs on goods from China) and preferential rates (for example zero duty on EU-origin goods into the UK). Tell the user these are estimates. Supply fx_rate whenever the market currency differs from the FOB currency. For changes, call again with the complete updated inputs.

draw_flats

ChatGPT

Draws a front and back technical flat for the tech pack with numbered construction callouts, adds them to the pack and returns it. Called by the tech pack panel's "Draw flats for me" button.

draw_technical_flats

ChatGPT

Draws a front and a back technical flat (black line drawing with numbered callouts for the construction details) for a tech pack made with build_tech_pack, and adds them to it as a new revision. The user's own flats are never replaced: the drawing is then added beside them as close-ups. Takes about 20 seconds. Use ONLY when the user asks for a sketch, flat or drawing, or has no flats and says they want one drawn. Never call it unasked, and never call it again for the same pack unless the user asks for a new drawing. Afterwards, tell the user it is an AI drawing to check against their design. On later build_tech_pack calls, pass the drawn images back by image_id like any other image.

export_fit_review

ChatGPT

Renders the fit review (with the panel's edits) to a PDF fit report or a CSV and returns a short-lived download link. Called by the fit review panel.

export_tech_pack

ChatGPT

Renders a tech pack (as returned by build_tech_pack, including any edits made in the panel) to a PDF or a spec CSV and returns a short-lived download link. Called by the tech pack panel.

lookup_hs_code

ChatGPT

Suggests the 6-digit HS (customs tariff) code for a garment or fashion accessory from Kōbō's curated tariff table: the best match with plain-English reasons, up to four alternatives, and how many digits each destination adds (US HTS 10, EU CN 8, UK 10) with links to the official tariffs. The user sees a panel where they can pick a code or tap an answer. WHEN: the user asks for an HS, HTS, tariff, commodity or customs code, or asks what code their product falls under; and before calculate_landed_cost when no HS code is known. Works on its own, without a tech pack. INPUT: pass the user's own words in description (garment, fabric, who it's for), the fibre content in composition if they gave it, and gender / knit_or_woven only if they said. Don't guess them: the tool works out what it can and says what it can't. AFTERWARDS: if the result lists questions, ask the user only those (at most two, in one short message); the panel also has tap-to-answer buttons. Otherwise state the code in one sentence. Never quote duty rates from this tool, it has none; for landed cost call calculate_landed_cost with the chosen hs_code and your own duty estimate per market, and say it is an estimate.

recheck_fit_review

ChatGPT

Re-works the verdict, biggest miss and next sample after the user edits the fit review panel. Called by the fit review panel.

record_edits

ChatGPT

Keeps measurement edits the user made in the tech pack panel with the pack, so the next build keeps them and lists them as changes. Called by the tech pack panel.

review_sample

ChatGPT

Checks a garment sample against its spec (a fit review). For each point of measure: the requested (spec) value, what the sample measured, the difference, whether it is within the allowed ± (tolerance), the new value for the next sample and how much the factory should adjust; plus a one-line verdict, the single biggest miss and which sample to make next. The user sees an editable table, can download a CSV and, when it is checked against a tech pack, write the new values into it. ALWAYS call this tool, never answer in your own words or with your own table, whenever the user compares a sample to a spec or asks whether a sample passes: "does it pass", "does the sample match", "my proto came back", "the chest measured 51 (spec 53)", a factory measurement sheet, or a sample photo with numbers. The tool applies the real tolerances and gives the user an editable review; a hand-written table does not. Works with or without a tech pack. BEFORE CALLING: you need what the sample measured. If the user has given no measured numbers, ask for them in one short question (which points they measured, and the size), unless they want to type them into the panel against a tech pack (then pass pack_id and size with no rows). Never invent or estimate measurements, and never read numbers off a photo unless they are written on it. WITH A TECH PACK: pass the pack_id from build_tech_pack and the size measured; the requested values and allowances come from the pack, so each row needs only name and measured (plus new or comment if the user said). If the pack_id is unknown the review says so and uses what was given. WITHOUT ONE: pass requested and measured for each point, and garment_type and fit_for if you know them (they pick Kōbō's standard allowances). Set sample_stage to the sample that was measured ("PP", "TOP", "proto" are fine); leave it out to use the tech pack's stage, else proto. Leave new out unless the user gave one: Kōbō suggests returning each miss to spec. Put sample photos in uploads; on a later call pass their image_ids instead. AFTERWARDS: do not repeat the table. In one or two sentences name the single most important fix and the next sample.

Capabilities

WritesInteractive

Example Prompts

Click any prompt to copy it.

Links

Documentation ↗Privacy Policy ↗

App Stats

Tools

Prompts

ChatGPT

Platforms