service slides start --all should include mycelium between db and slides; --clear on proc wipes prior mycelium reg #243

New issue

Closed

opened 2026-05-10 02:41:08 +00:00 by mik-tf · 1 comment

mik-tf commented

2026-05-10 02:41:08 +00:00

Owner

Context

From hero_slides#49 session 86 close-out — repeatedly hit during stack-up.

Two coupled problems

1. `--all` doesn't include mycelium

service_slides.nu start --all runs (per current source):

service_proc start --reset --update --clear
service_router start --reset --update
service_db start --reset --update
# … then hero_aibroker prereq, then hero_slides

But hero_aibroker requires mycelium to be reachable for external access (per service_slides.nu's own svx_require_hero_aibroker block). On a from-nothing workstation --all errors out at the aibroker prereq because mycelium isn't up.

Workaround used in session 85/86: bring mycelium up manually first (service mycelium start --download), then run service slides start without --all.

2. `--clear` on proc wipes mycelium even when it was already up

If the operator brings mycelium up first then runs service slides start --all, the --all's service_proc start --reset --update --clear empties the hero_proc registry — including the mycelium registration. The chain then fails because mycelium is no longer registered.

Proposal

Add mycelium between db and aibroker in --all's sequence so the dependency graph is honoured:

service_proc start --reset --update --clear
service_router start --reset --update
service_db start --reset --update
service_mycelium start --download   # NEW — adds the missing dependency edge
# … aibroker auto-bringup, then slides

Matches the canonical service_complete chain in packages.nu.

Alternatively: drop --clear from --all's proc start (so existing mycelium registration survives) and document the trade-off.

Acceptance

On a clean workstation, service slides start --all succeeds end-to-end without manual service mycelium start first.
On a workstation that already has mycelium running, --all doesn't kill it.
DEPLOYMENT.md / service_slides.nu docstrings updated to reflect the new behaviour.

Refs

lhumina_code/hero_slides#49
lhumina_code/hero_skills/nutools/modules/services/service_slides.nu lines ~239-253 (--all block)
lhumina_code/hero_skills/nutools/modules/services/packages.nu (service_complete chain — canonical order)

## Context From [hero_slides#49](https://forge.ourworld.tf/lhumina_code/hero_slides/issues/49) session 86 close-out — repeatedly hit during stack-up. ## Two coupled problems ### 1. `--all` doesn't include mycelium `service_slides.nu start --all` runs (per current source): ``` service_proc start --reset --update --clear service_router start --reset --update service_db start --reset --update # … then hero_aibroker prereq, then hero_slides ``` But hero_aibroker requires mycelium to be reachable for external access (per `service_slides.nu`'s own `svx_require_hero_aibroker` block). On a from-nothing workstation `--all` errors out at the aibroker prereq because mycelium isn't up. Workaround used in session 85/86: bring mycelium up manually first (`service mycelium start --download`), then run `service slides start` without `--all`. ### 2. `--clear` on proc wipes mycelium even when it was already up If the operator brings mycelium up first then runs `service slides start --all`, the `--all`'s `service_proc start --reset --update --clear` empties the hero_proc registry — including the mycelium registration. The chain then fails because mycelium is no longer registered. ## Proposal Add mycelium between db and aibroker in `--all`'s sequence so the dependency graph is honoured: ``` service_proc start --reset --update --clear service_router start --reset --update service_db start --reset --update service_mycelium start --download # NEW — adds the missing dependency edge # … aibroker auto-bringup, then slides ``` Matches the canonical `service_complete` chain in `packages.nu`. Alternatively: drop `--clear` from `--all`'s proc start (so existing mycelium registration survives) and document the trade-off. ## Acceptance - [ ] On a clean workstation, `service slides start --all` succeeds end-to-end without manual `service mycelium start` first. - [ ] On a workstation that already has mycelium running, `--all` doesn't kill it. - [ ] DEPLOYMENT.md / `service_slides.nu` docstrings updated to reflect the new behaviour. ## Refs - https://forge.ourworld.tf/lhumina_code/hero_slides/issues/49 - `lhumina_code/hero_skills/nutools/modules/services/service_slides.nu` lines ~239-253 (`--all` block) - `lhumina_code/hero_skills/nutools/modules/services/packages.nu` (`service_complete` chain — canonical order)

mik-tf referenced this issue

2026-05-10 12:05:36 +00:00

[meta] broken: dispatcher.nu calls 47 deleted service_*.nu functions after d1825ad — service <name> <verb> non-functional on origin/development #245

mik-tf commented

2026-05-10 12:05:59 +00:00

Author

Owner

Closing as obsolete on origin/development.

The service_slides.nu --all block this issue describes was deleted by commit d1825ad (despiegk, 2026-05-10 08:21) along with 46 other service_*.nu modules and packages.nu. The dispatcher still calls service_slides start --all but the function no longer exists on disk — so on origin/development the symptom this issue documents (mycelium not started by --all) is moot in the sense that service slides start --all itself can't run.

The underlying broken-dispatcher state is tracked in #245. Once that's resolved (whichever resolution path despiegk picks), if the slides flow is restored with --all semantics still in scope, this can be re-filed against the new code path.

Note: anyone still on a runtime clone that hasn't pulled past d1825ad continues to hit this bug as documented; the workaround there remains: bring mycelium up first, then service slides start without --all.

Signed-off-by: mik-tf

Closing as obsolete on origin/development. The `service_slides.nu` `--all` block this issue describes was deleted by commit [`d1825ad`](https://forge.ourworld.tf/lhumina_code/hero_skills/commit/d1825ad) (despiegk, 2026-05-10 08:21) along with 46 other `service_*.nu` modules and `packages.nu`. The dispatcher still calls `service_slides start --all` but the function no longer exists on disk — so on origin/development the symptom this issue documents (mycelium not started by `--all`) is moot in the sense that `service slides start --all` itself can't run. The underlying broken-dispatcher state is tracked in [#245](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/245). Once that's resolved (whichever resolution path despiegk picks), if the slides flow is restored with `--all` semantics still in scope, this can be re-filed against the new code path. Note: anyone still on a runtime clone that hasn't pulled past `d1825ad` continues to hit this bug as documented; the workaround there remains: bring mycelium up first, then `service slides start` without `--all`. Signed-off-by: mik-tf

mik-tf closed this issue

2026-05-10 12:05:59 +00:00