Dogwood vendor product lookup

Fill blank product or service cells in the Dogwood vendor list from vendors’ public web pages. The source file stays untouched; the script writes a new review file with the source and confidence for every guess.

Review columns

Column	What it records
Products	Extracted or inferred product/service description
Product Source URL	Public page used for the description
Product Confidence	medium, low, or none
Scrape Status	Whether the value came from the web, the company name, or was skipped
Scrape Notes	Fetch errors or the reason a row was not filled

Existing product values are left alone unless --overwrite is supplied. Treat every new value as a draft: the scraper uses text heuristics, not a vendor catalog API.

Setup

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt

Run the Dogwood list

Keep the source PDF outside this repository because it contains private contact details.

python product_scraper.py \
  "/Users/ezrakahn/Downloads/Master Product List.xlsx - Vendor Contact Numbers.pdf" \
  --output output/dogwood_enriched.csv \
  --xlsx-output output/dogwood_enriched.xlsx

Inputs may be PDF tables, .xlsx/.xlsm workbooks, or CSV files. CSV output is always written; --xlsx-output adds a formatted workbook.

Common passes

Check five missing rows before running the whole file:

python product_scraper.py path/to/vendors.pdf --limit 5

Avoid all network requests and infer only from company/contact names:

python product_scraper.py path/to/vendors.pdf --no-web

Name the columns when automatic header detection picks the wrong fields:

python product_scraper.py path/to/vendors.xlsx \
  --company-column "Vendor" \
  --product-column "Product"

Replace existing product values instead of filling blanks only:

python product_scraper.py path/to/vendors.xlsx --overwrite

HTTP results are cached in .scrape_cache.json. Pass --cache '' for a run that should neither read nor write the cache.

Tests

The test suite uses local HTML and CSV fixtures; it does not make web requests.

python -m unittest discover -s tests -v