The first attempt at a working GitHub Actions deploy pipeline for BotWhy took a long, visible trail of small failed commits over about a week — SSH connection failures, Docker build errors, `git pull` conflicts, React build issues — each fixed one at a time until the pipeline finally ran clean.
There wasn't one single root cause — this was the ordinary cost of standing up a real CI/CD pipeline for the first time: SSH key setup, known-hosts trust, Docker Compose install on the runner, image tagging syntax, and the interaction between all of them, each surfacing only once the previous one was fixed.
Worked through it commit by commit: `test SSH connection` → `debug ssh connection` → `fix ssh` → `add know host` → `fix exact path`, then a second wave on the build side: `fix react build` → `fix react force-recreate` → `fix npm run` → `fix rm -rf dist/*` → `try npm run clean` → `fix git pull origin main`. No shortcuts — each fix was small and specific to what had just failed.
This pipeline has run without needing this kind of rescue since. The pattern that emerged — and that shows up again in every app onboarded after this one — was copying a known-working `deploy.yml` from an already-working repo rather than building each new one from scratch, which is exactly why BotWhy, PopRoom, Apercu, and x all share the same workflow shape today.