Skip to content

MARKDOWN SOURCE

Raw
# Vertical Slices: ino VPS server

1. **Server tracer (local validation):** build the Actix server image locally and tag it as `silenloc/ino:<version>`, add idempotent local `ino once server setup`, and run `ino once serve` inside its Once-managed container. Serve `/up` and a bearer-protected stub `GET /api/v1/apps` through the existing local Once proxy; verify with host `curl` that health and authentication work end to end. A registry push and VPS deployment are deferred; the list response is deliberately stubbed in this slice.
2. **Read real apps:** connect the authenticated list route to the restored Once namespace, refresh before reading, and return only redacted `AppSummary` values. Validate with existing Once-shaped apps and prove sensitive settings never appear in the response.
3. **Deploy remotely:** wire `POST /api/v1/apps` to the existing Once deploy path, accept registry-backed image/settings input, and return a structured summary after health verification. Verify a real app becomes reachable through the proxy and that a failed deployment leaves existing data/resources safe.
4. **Update apps and the server:** wire `PATCH /api/v1/apps/{host}` to Once update semantics, preserving omitted settings and old working containers on failure. Preserve the marked server app's socket mount, key, and port across its own replacement; verify health/route handoff and read-after-write confirmation when the original response is interrupted.
5. **Release-triggered update:** parse the explicit `ino.deploy.server`/`ino.deploy.host` Dockerfile labels, read the local `INO_API_KEY`, and have the normal release flow request an update to the published version. Cover the HTTP request and self-update read-back with a fake server; no-label releases retain existing behavior. Do not perform Git pushes, registry pushes, or VPS deployments during local validation.
6. **Lifecycle and safe removal:** add authenticated start, stop, restart, and remove routes. Keep persistent data on removal by default and delete it only when `remove_data=true`; verify both paths against an isolated Once namespace.
7. **Exec and logs:** add non-TTY command execution with separate output/exit status and bounded log retrieval. Verify byte/tail limits, truncation reporting, and that no interactive stream or unbounded response is exposed.
8. **Release readiness:** complete stable/sanitized HTTP errors, constant-time authentication checks, request/response limits, serialized operations and refresh behavior, server self-removal protection, Docker-backed compatibility checks, quick-start/API/security docs, and a successful local production-image build. Defer public VPS/registry release validation until credentials and a target are supplied.

Vertical Slices: ino VPS server

  1. Server tracer (local validation): build the Actix server image locally and tag it as silenloc/ino:<version>, add idempotent local ino once server setup, and run ino once serve inside its Once-managed container. Serve /up and a bearer-protected stub GET /api/v1/apps through the existing local Once proxy; verify with host curl that health and authentication work end to end. A registry push and VPS deployment are deferred; the list response is deliberately stubbed in this slice.
  2. Read real apps: connect the authenticated list route to the restored Once namespace, refresh before reading, and return only redacted AppSummary values. Validate with existing Once-shaped apps and prove sensitive settings never appear in the response.
  3. Deploy remotely: wire POST /api/v1/apps to the existing Once deploy path, accept registry-backed image/settings input, and return a structured summary after health verification. Verify a real app becomes reachable through the proxy and that a failed deployment leaves existing data/resources safe.
  4. Update apps and the server: wire PATCH /api/v1/apps/{host} to Once update semantics, preserving omitted settings and old working containers on failure. Preserve the marked server app's socket mount, key, and port across its own replacement; verify health/route handoff and read-after-write confirmation when the original response is interrupted.
  5. Release-triggered update: parse the explicit ino.deploy.server/ino.deploy.host Dockerfile labels, read the local INO_API_KEY, and have the normal release flow request an update to the published version. Cover the HTTP request and self-update read-back with a fake server; no-label releases retain existing behavior. Do not perform Git pushes, registry pushes, or VPS deployments during local validation.
  6. Lifecycle and safe removal: add authenticated start, stop, restart, and remove routes. Keep persistent data on removal by default and delete it only when remove_data=true; verify both paths against an isolated Once namespace.
  7. Exec and logs: add non-TTY command execution with separate output/exit status and bounded log retrieval. Verify byte/tail limits, truncation reporting, and that no interactive stream or unbounded response is exposed.
  8. Release readiness: complete stable/sanitized HTTP errors, constant-time authentication checks, request/response limits, serialized operations and refresh behavior, server self-removal protection, Docker-backed compatibility checks, quick-start/API/security docs, and a successful local production-image build. Defer public VPS/registry release validation until credentials and a target are supplied.