bigquery-seeds

What problem does it solve?

BigQuery-seeds helps you turn publicly queryable BigQuery tables into labeled training seeds without manual dataset wrangling.

Core Features & Use Cases

• Seed generation from BigQuery: Run SQL against BigQuery to retrieve text plus a date field, then package rows as seeds for downstream labeling.
• Label-in-SQL pattern: Embed ground-truth fields (like scores or accepted answers) inside the seed text so a question/label generator can extract both from the same result.
• Schema inspection & safe dataset selection: Validate table schemas and restrict sourcing to publicly queryable datasets (no private tables, no paid commercial datasets).

Use case: generate hundreds of forecasting question candidates from Hacker News stories by selecting title, score, and text while attaching timestamps for time-aware labeling.

Quick Start

Ask an AI to generate BigQuery seed rows from a public table by providing a query, the text column, and the date column.