data-warehouse

PostHog 决定将数据仓库从 DuckDB/DuckLake 迁出，原因是单节点引擎和 DuckLake 的目录设计难以支撑云端大规模、复杂的工作负载。文章详述了容量预估、S3 异步抓取、目录统计膨胀以及 Postgres 连接扩展等实际瓶颈。

The article describes deploying an AI text-to-sql system for a bank, highlighting that the model was less important than verification mechanisms, evaluation sets, and governance rules for production success.

Jitsu is an open-source event data pipeline that collects events from web, apps, and other sources, routing them to data warehouses like ClickHouse, Snowflake, and BigQuery with JavaScript functions for flexibility and a free tier for limited events.

A pattern for giving AI agents reliable access to data warehouses by using a curated semantic layer (Databricks Genie) instead of raw text-to-SQL, improving accuracy and governance. The agent calls Genie's Conversation API as a tool, receiving both natural-language responses and exact SQL.