Migrate videre and shepherd onto the tagged runtime in one pass.
Why
Both pin nexum-runtime at rev = d78a368, which is 92 commits and 15 breaking commits behind main.
Neither compiles today: videre-host/src/registry.rs still imports host::actor::{ActorFault, ActorSlot, Liveness, SupervisedStore} and host::extension::{HostService, Installed, ProviderInstance, ProviderKind, downcast_service}, every one deleted by #233, and kind = "venue-adapter" in the adapter manifests refuses at load since #250 retired [component].kind.
A runtime whose only two consumers cannot build against it is not production ready, and nothing in this tracker tracked it until now.
Why it waits for the freeze
Batching is the whole point. The breaking changes are the guest ABI and manifest grammar (#239), the host public Rust API (#241), the metric names and labels (#243), the refusal wording and error_kind set, and the guest-to-guest cut.
Bumping before those land means migrating twice.
Known breaks to carry
Extension loses service, provider and admit_provider; admit_worker loses its providers parameter.
HostService, ServiceKind, ProviderKind, [[services]] and the supervised actor are gone.
[component].kind is retired.
on-event becomes on-trigger and nexum:host goes to 0.2.0.
[[subscription]] becomes [[trigger]].
Videre's venue adapters are the hard part: they are guest wasm driven by host Rust, which is the path #233 removed. That needs the plugin-engine seam, so this issue unblocks the build and the adapters may stay broken until that lands. Say so explicitly rather than discovering it mid-migration.
Done when
Both repos compile and test green against the v1 tag, or their remaining breakage is a named issue rather than a surprise.
Migrate videre and shepherd onto the tagged runtime in one pass.
Why
Both pin
nexum-runtimeatrev = d78a368, which is 92 commits and 15 breaking commits behind main.Neither compiles today:
videre-host/src/registry.rsstill importshost::actor::{ActorFault, ActorSlot, Liveness, SupervisedStore}andhost::extension::{HostService, Installed, ProviderInstance, ProviderKind, downcast_service}, every one deleted by #233, andkind = "venue-adapter"in the adapter manifests refuses at load since #250 retired[component].kind.A runtime whose only two consumers cannot build against it is not production ready, and nothing in this tracker tracked it until now.
Why it waits for the freeze
Batching is the whole point. The breaking changes are the guest ABI and manifest grammar (#239), the host public Rust API (#241), the metric names and labels (#243), the refusal wording and
error_kindset, and the guest-to-guest cut.Bumping before those land means migrating twice.
Known breaks to carry
Extensionlosesservice,providerandadmit_provider;admit_workerloses its providers parameter.HostService,ServiceKind,ProviderKind,[[services]]and the supervised actor are gone.[component].kindis retired.on-eventbecomeson-triggerandnexum:hostgoes to 0.2.0.[[subscription]]becomes[[trigger]].Videre's venue adapters are the hard part: they are guest wasm driven by host Rust, which is the path #233 removed. That needs the plugin-engine seam, so this issue unblocks the build and the adapters may stay broken until that lands. Say so explicitly rather than discovering it mid-migration.
Done when
Both repos compile and test green against the v1 tag, or their remaining breakage is a named issue rather than a surprise.