story: service install is not reproducible across dev machines #216

New issue

Open

opened 2026-05-06 17:07:15 +00:00 by mik-tf · 0 comments

mik-tf commented

2026-05-06 17:07:15 +00:00

Owner

Overview

service_<name> install --update --release produces different results on different developer machines. Onboarding a new dev requires hours of "tribal" debugging because no two machines have the same baseline. Today's coping strategy is "everyone works on the shared dev VM" (home#121) — that's a workaround, not a fix.

Why

Meeting 2026-05-06: "cant install / not reproducible / team is out of sync / devs work on same machine"
Concrete examples:
- hero_skills#187 — install_core missing unzip; install_bun fails on fresh VMs
- hero_skills#185 — service_livekit hardcodes redis port 6379, depends on host having claude CLI
- hero_skills#191 — load_init_sh does not follow source directives — env.sh exports go missing

Acceptance

Document a known-clean baseline (Ubuntu 24.04 LTS + listed apt deps) for hero_skills install scripts
Add a CI/scheduled job that runs service_install_all on a fresh container and asserts green
Resolve the three child bugs above
New-developer onboarding doc lists exactly the apt packages, env vars, and tokens needed (no "ask Slack")

Parent: home#121 — shared dev VM workaround
Parent: home#213 — repo validation skills

Source: meeting notes 2026-05-06.

## Overview `service_<name> install --update --release` produces different results on different developer machines. Onboarding a new dev requires hours of "tribal" debugging because no two machines have the same baseline. Today's coping strategy is "everyone works on the shared dev VM" ([home#121](https://forge.ourworld.tf/lhumina_code/home/issues/121)) — that's a workaround, not a fix. ## Why - Meeting 2026-05-06: "cant install / not reproducible / team is out of sync / devs work on same machine" - Concrete examples: - [hero_skills#187](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/187) — `install_core` missing unzip; install_bun fails on fresh VMs - [hero_skills#185](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/185) — `service_livekit` hardcodes redis port 6379, depends on host having `claude` CLI - [hero_skills#191](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/191) — `load_init_sh` does not follow `source` directives — env.sh exports go missing ## Acceptance - [ ] Document a known-clean baseline (Ubuntu 24.04 LTS + listed apt deps) for hero_skills install scripts - [ ] Add a CI/scheduled job that runs `service_install_all` on a fresh container and asserts green - [ ] Resolve the three child bugs above - [ ] New-developer onboarding doc lists exactly the apt packages, env vars, and tokens needed (no "ask Slack") ## Related - Parent: [home#121](https://forge.ourworld.tf/lhumina_code/home/issues/121) — shared dev VM workaround - Parent: [home#213](https://forge.ourworld.tf/lhumina_code/home/issues/213) — repo validation skills Source: meeting notes 2026-05-06.