Honestly didn't expect to get this deep into Kestra but here we are. I built a price monitoring system for my uncle's mobile shop — it checks prices on Google Shopping every morning and WhatsApps him the report before he opens. Built another one that summarizes GitHub issues for beginners using AI. Both run on YAML, no backend needed. The thing that got me was the Gantt view — saw my AI task was eating 47 seconds, fixed the prompt, got it to 11 seconds. Never would've caught that otherwise. Retries, secrets, plugin ecosystem — all solid.

The ForEach tasks, cron triggers, and secrets management made it easy for me to automate my Smart Restaurant IoT project end-to-end.

I use Kestra to orchestrate an AI research pipeline, and the biggest benefit for me is how it handles errors at the orchestration layer instead of forcing me to write complicated retry logic in Python. I rely on runIf conditions to create multi-tier fallbacks, so if my primary LLM fails, it automatically switches to another one. I also appreciate how clearly it separates configuration from code. I keep my configuration as YAML files in Kestra Namespace Files, and adding a new workload is as simple as uploading a new file. That way, I don’t have to touch the underlying flow logic. The observability setup has been a huge time saver as well. I use an ExecutionNamespaceCondition trigger to manage alerts globally: rather than bloating every workflow with repetitive error handling, I have one central flow that listens to a namespace prefix and pages me in Slack if something breaks. I can also run tasks in parallel with allowFailure: true, so if one API goes down, the rest of the pipeline can keep running.