Dive Site Data Enricher

This project is a resilient data enrichment pipeline that processes dive sites one row at a time, enriching them with AI-generated content and geospatial metadata.

The system is designed for safety and reliability:

• Reads an input CSV (new_table.csv) and appends results row-by-row to an output CSV (new_table_updated.csv).

• Supports resume-on-crash by skipping rows already written.

• Uses OpenAI Chat Completions (JSON mode) to generate two fields:

  • description (short location overview)

  • dive_description (detailed dive conditions, access, visibility, highlights)

• Applies geocoding via Nominatim (OpenStreetMap) to fill missing country/region from latitude/longitude.

• Normalizes and rounds up max depth values for consistency.

• Flushes data after every row so progress is never lost.

This project demonstrates how to build a fault-tolerant ETL pipeline with AI integration, combining structured data cleaning, geospatial enrichment, and automated content generation into one cohesive process.