Blog
Agentic Workloads on Airflow: Observable, Retryable, and Auditable by Design
How Dynamic Task Mapping and the common.ai provider turn a multi-dimensional research question into a fan-out/fan-in pipeline where every LLM call is a named, logged, independently retryable task, not a hidden step inside a reasoning loop.
Ask Your Survey Anything: Building AI Analysis Pipelines with Airflow 3
A walkthrough of two natural language analysis pipelines over the 2025 Airflow Community Survey, covering an interactive human-in-the-loop version and a fully automated scheduled version, using operators from the common.ai and common.sql providers.