BARC Digital Asset Management

A Python pipeline for Bhutan Animal Rescue & Care that turns their shared Google Drive photo archive into a searchable Airtable catalog.

Stack: Python, Claude Haiku 4.5, Pillow + pillow-heif, pyairtable, Google Drive

What It Does

This turned a Google Drive without metadata into an Airtable-backed highly searchable and browsable interface.

How it works

Helped BARC, an animal sanctuary in Bhutan turn its collection of images and documents on Google Drive into an AI-tagged and auto-updating AirTable base.

Built with

Model: Claude Haiku 4.5 via the Anthropic Python SDK, called with structured output bound to a Pydantic v2 schema.
Image handling: Pillow with pillow-heif for native HEIC (iPhone) support, resized to a 1568 px long edge per Anthropic's vision cost/quality guidance.
EXIF: Pillow's EXIF reader for DateTimeOriginal → year folder. A separate inspection utility surfaces GPS, camera, and date coverage stats before scanning.
Storage: Photos canonical on a shared Google Drive folder (read from a locally-synced copy so EXIF and bytes are available). Airtable is the structured metadata index.
Airtable: pyairtable against the REST API for record I/O and the Meta API for schema bootstrap. Batch-10 writes with typecast so the controlled vocabulary grows organically.
Output format: JSONL, append-only, jq-friendly, pandas-friendly, trivial to resume from.