About Flow Doctor

Kestra's core bet — declarative, language-agnostic, everything-as-code orchestration — is working. The product's strongest moment is authoring a flow. Its weakest moment is the first failed execution: that's where duplicated logs, empty loop Gantts, silent YAML traps, and worker-memory surprises show up, in exactly the debugging loop that's hardest.

Flow Doctor moves that discovery left — from after the failure in production to before you ship. Paste a flow, see where it's fragile, get a one-line fix with the Kestra issue it traces back to. The idea is to turn hard-won, scattered failure-mode knowledge into a proactive, teachable signal.

The roadmap this implies

The web tool is the wedge. The same engine wants to live where flows already live:

• Public tool (this) — top-of-funnel, "learn to write reliable flows."
• kestra flow doctor CLI — fits the everything-as-code workflow, runs locally.
• CI gate — block a PR when a flow's reliability score drops. Reliability-as-code, reviewed in Git alongside the flow.
• In-editor lint — surface the warnings inline at authoring time, before the first failure ever happens.

Who built it

Roman Martins — a product manager and builder who spent a year hand-rolling a worse version of Kestra (launchd + cron + retries + observability, by hand) before discovering the real thing. That's the lens here: someone who knows exactly where home-grown orchestration breaks.

This is an independent, unofficial tool — not affiliated with or endorsed by Kestra. It's a heuristic built from public sources, offered free. The methodology is fully open; corrections welcome.