Introducing a SQL-based metrics layer powered by
summary
Loads a GitHub commits parquet file into a DuckDB-powered local dev environment, auto-generates a metrics model of about ten measures (total commits, merge commits, avg additions) via an OpenAI call and builds a dashboard that immediately surfaces a 7,000-commit spike on one day. Claims over 80% of dashboards are first-pass usable when generated with GPT 40, argues metrics-first SQL definitions shrink data 10–100x with OLAP cubes, notes single-node scale limits for large enterprise customers.
AI-written summary · about
chapters (9)
- 0:00SQL-based metrics layer intro
- 1:25Metrics core to analytics
- 3:12DuckDB dashboard demo
- 5:22Metrics as building blocks
- 7:27Metrics defined in SQL
- 10:17Legacy metric stack challenges
- 11:40DuckDB metric stack architecture
- 13:22AI accelerating metrics
- 14:47Summary
related talks
07mAerodynamic data models: Flying fast at scale with DuckDB
01h 02mAnalytics for not-so-big data with DuckDB
28mStop querying tables: give AI governed metrics instead
22mData Exploration Made Easy with Open Source
45mAnalytics for not so big data with DuckDB
40mMetrics and Semantics: An evolution in data consumption
27mThe Duck and the DataFrame: A Data Engineer’s Journey with DuckDB
06mA quack at building scalable data pipelines with DuckDB