CD: auto-deploy main to fly.io on push #2
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally filed by @coilysiren on 2026-05-13T14:40:43Z - https://github.com/coilysiren/cli-mcp-go-ecosystem/issues/8
Today fly.io deploys are manual (
flyctl deployfrom a workstation). Surfaced during testing of #4/#5/#7: even after fixing bugs and pushing to main, the hosted endpoint at https://cli-mcp-go-ecosystem.fly.dev/mcp keeps serving the previous binary until somebody remembers to ssh into a laptop and runflyctl deploy.The repo already has the build artifacts in place:
Dockerfile- reproducible build targetfly.toml- app config (sjc, port 8080, auto-stop)Missing: the GitHub Actions glue. Want a
.github/workflows/deploy.ymlthat on push tomain:flyctl deploy --remote-onlyagainst the configured appFLY_API_TOKENrepo secretOpen questions:
deploytag is pushed? Repo norm acrosscoilysiren/*is commit-to-main with no PRs, so deploy-on-main matches. Optional: skip deploy when only*.md/docs//.github/changed (paths-ignore filter).FLY_API_TOKENneeds to land in repo secrets.flyctl tokens create deploy -a cli-mcp-go-ecosystemproduces a scoped deploy-only token. Store under repo secrets, not org secrets, to keep blast radius tight.flyctl releases/flyctl deploy --image <prev>in the deploy.yml comments so future-Kai knows where to look when an autodeployed commit breaks prod./health(if implemented) or/mcpinitialize, fail the workflow if the hosted endpoint doesn't respond. Saves "deploy succeeded, app crashed" silence.Lower-priority adjacent: same pattern would be useful for any future cli-mcp-* consumer that wants a hosted demo, so the workflow is reusable. Maybe a snippet in cli-mcp-go-ecosystem's README's "Local dev" section once #5 is closed.