Cracking PDF Catalogs: AI to Airtable for SKUs and Prices

Описание и применение

Автоматизация извлечения SKU, описаний и цен из PDF‑каталогов поставщиков с помощью LLM и Airtable сократила время обработки с половины дня до 30 секунд.

Задача

Manual extraction of SKUs, product descriptions, and prices from vendor PDF catalogs was time‑consuming, error‑prone, and required constant updates.

Результат

Processing time reduced from half‑day to ~30 seconds per catalog; 90% accuracy on first pass; human review needed for remaining errors.

Рабочий процесс

• Extract raw text from PDF with pypdf
• Prompt self‑hosted Llama‑3‑70b‑instruct via Ollama with clear instructions and JSON output format
• Parse LLM JSON output
• Use pyairtable to insert records into Airtable, setting Review Status to 'Needs Review'

Доказательства

Each one was a unique snowflake of inconsistent formatting, and my team's job was to manually pull out SKUs, product descriptions, and prices into our Airtable database.

Источник

I fed this raw, often unstructured, text into an LLM. I'm using a self-hosted llama-3-70b-instruct instance via Ollama – saved me a ton on API costs.

Источник

From there, it was straightforward. A simple Python script takes the JSON output and uses the pyairtable library to push the extracted products into our Airtable base.

Источник