Carnet Manifest Merger

Combine vendor manifests into a Boomerang Carnet General List

International touring runs on Carnets, and Carnets live or die by the General List attached to them. The Carnet Manifest Merger is a free tool from Truck Packer that takes the pull sheets you already have (Clair Global XLSX, Spectrum Sound PDF, Pisgah AVL container manifest, Christie Lites, ATOMIC, anything the vendors send your way) and folds them into one Boomerang-ready XLSX with every column the customs office wants: item number, trade description, quantity, kilograms, US dollar value, country of origin, serial number. Drop the files into the upload zone, let the merge run, review the combined rows in the editable table, then download the General List. No retyping, no spreadsheet copy-paste, no missed rows.

How the extraction works

Spreadsheets go through a deterministic column-matching pass that handles the usual suspects: Qty, Part Desc, Weight, Value, Origin, etc. Lb-to-kg conversion is automatic, country names normalise to ISO-3166-1 alpha-2 (USA → US, Deutschland → DE, china → CN), and rows with no real description are dropped so the General List doesn't end up padded with garbage. Spreadsheets the heuristic already handles cleanly skip the AI step entirely, which keeps the merge fast and free of token cost.

PDFs are sent to a recent OpenAI model as native file inputs, so the model parses the PDF the same way a human would: it sees the rendered text and the visual table structure together. That handles the common touring-vendor failure modes that naive text extraction misses: scanned or image-based PDFs, multi-column layouts, tables with embedded icons, vendor templates with custom fonts. Every row is produced under strict JSON Schema mode, so every cell of every row is grammar-constrained to a valid Carnet field; the model cannot return prose, markdown, or hallucinated columns. Rows then pass through a Zod validator before they reach the on-screen review table.

Frequently asked questions

What is an ATA Carnet General List?

The General List is the master inventory of every item travelling on an ATA Carnet. The Carnet itself is a one-page customs document that lets you temporarily import equipment into a foreign country without paying duties. The list must include each item's description, quantity, country of origin, net weight, and fair value. Boomerang Carnet, Roanoke, and other carnet providers all accept the same column layout, which is what this tool produces.

Why combine multiple vendor manifests into one Carnet?

A single touring production usually rents gear from several vendors (Clair Global for audio, Christie Lites for lighting, ATOMIC for staging, etc.). Each vendor sends a pull sheet in their own format, whether PDF, Excel, or sometimes Word. Customs only sees one document at the border. The Carnet Manifest Merger consolidates every vendor's manifest into one Boomerang-ready General List so you don't have to retype hundreds of rows.

Will the AI hallucinate values for items?

Only when you explicitly opt in via the "Fill in missing weights and values" checkbox. With the box checked, the AI estimates fair used-market replacement values for items whose source manifest didn't state a value, and marks every estimate clearly in the on-screen review table with a dotted amber underline. Editing any estimated cell clears the flag automatically. The exported XLSX matches Boomerang's six-column General List template exactly, so estimate flags don't appear in the downloaded file (you reviewed them on screen before exporting). Uncheck the box if your customs broker requires only documented values.

How accurate is the column extraction from PDFs?

Spreadsheets go through a deterministic header-matching heuristic that produces clean rows in under a second. PDFs are read by a recent OpenAI model using strict JSON Schema mode, which forces the model to emit only valid Carnet rows: never free-form text, never hallucinated columns, never prompt-injectable. Every row also lands in an editable preview, so you can correct anything before downloading.

Is my manifest data shared with anyone?

OpenAI receives the PDF files you upload (as native file inputs through the Responses API) and the extracted cell text from any spreadsheets that need AI fallback, for the duration of the normalization call. OpenAI does not train on API submissions. The tool can optionally archive completed merges to a Vercel Blob bucket the operator controls; that's off by default until BLOB_READ_WRITE_TOKEN is set. No data is sent to any third party other than OpenAI.