Schedule
Bind compute to a flow with one decorator. Different hardware for different steps of the same pipeline.
Read on belowPer-node retries, parallel execution, and state-aware skipping for dbt.
Read on belowDebugging, automations, config plugins, result storage, assets, rate limits, and more.
Read on belowDay 03 · Headliner
Last month, we orchestrated two millennia of compute. We’ve spent the quarter turning the patterns we saw from high-scale teams into first-class features: infrastructure debugging in the UI, managed automations, config plugins, default result storage, the Assets Catalog, and more.
01
Memory pressure, pod logs, and evictions in the run page. Stop opening CloudWatch.
Read the docsDay 02
dbt repeats work that hasn’t changed and spins up a Kubernetes pod per task. The Prefect dbt Orchestrator runs your graph node by node, with state-aware caching that skips unchanged work, per-model retries, and concurrency that uses the rest of your stack.
Day 01 · Headliner
Pipelines that need different machines for different steps used to require three flows, three deployments, and a fourth flow to coordinate them. Now you can route a function to a work pool with a decorator, and Prefect handles the rest.
Social
Posts on LinkedIn and X each day. We’d love for you to help spread the news!
From your terminal
Run the command. See what unfolds.
uvx --from git+https://github.com/PrefectHQ/prefect-launch.git prefect-launchLaunch Week Swag
We made a small batch of hats for launch week. Drop your details and we’ll send one your way.
Claim a hat