COLA Cloud Sample Pack v2
=========================

1,000 recent approved COLA (Certificate of Label Approval) records
from the TTB's federal registry of US alcohol product labels, with associated
label-image metadata, image storage references, and extracted barcodes.

Generated: 2026-08-20T08:47:30Z
Source: https://colacloud.us
Category taxonomy: 2.0.2
Category taxonomy SHA-256: 481244008e2523f2babece0e2e8bc879d33cfeb630d0446ce9f973e489d73d02
Classification release: taxonomy-v2-prompt22-full-wave-2015plus-2026-08-03-v1


FILES
-----

cola.csv (1,000 rows)
  Recent approved product-label records. LLM_CATEGORY and LLM_CATEGORY_PATH use
  the released COLA Cloud Taxonomy 2.0.2. A defensible category may
  stop at a non-leaf node when the available evidence does not support a deeper
  placement.

cola_image.csv (1,873 rows)
  Label-image metadata and OCR text for the selected COLAs.

cola_image_barcode.csv (526 rows)
  Barcodes and QR codes extracted from the selected label images.

manifest.json
  Exact taxonomy/release identity plus SHA-256, byte count, and row count for
  every data file in this archive.


RELATIONSHIPS
-------------

cola.TTB_ID -> cola_image.TTB_ID (one-to-many)
cola_image.TTB_IMAGE_ID -> cola_image_barcode.TTB_IMAGE_ID (one-to-many)
cola.TTB_ID -> cola_image_barcode.TTB_ID (one-to-many)


NOTES
-----

- LLM_CATEGORY and LLM_CATEGORY_PATH are the new multimodal Taxonomy v2
  classifications. They combine label imagery, bounded OCR, and relevant TTB
  metadata; accepted preservation/fallback outcomes remain evidence-bound.
- Other LLM_-prefixed enrichment fields may come from earlier task-specific
  image enrichment and should not be interpreted as Taxonomy v2 outputs.
- OCR fields (OCR_ABV, OCR_VOLUME, OCR_TEXT) are extracted via Google Vision AI.
- BARCODE_COLA_OCCURENCES retains the public field's historical spelling and
  counts how many distinct products share a barcode value.
- S3_KEY and MAIN_IMAGE_S3_KEY are storage references, not public image URLs.
  Authorized label viewing uses authenticated, file-specific URLs that expire.
  Do not construct a CloudFront URL from a TTB_IMAGE_ID or an S3 key; use the
  COLA Cloud web app or API to obtain current authorized access.


FULL DATASET
------------

The complete COLA Cloud dataset is updated daily and is available through the
web app, REST API, and licensed bulk delivery.

  Docs:       https://docs.colacloud.us
  API:        https://docs.colacloud.us/api-reference
  Dashboard:  https://app.colacloud.us
  Contact:    help@colacloud.us
