vinzenz/hyperhive
vinzenz/hyperhive
Watch
0
0
Fork
You've already forked hyperhive
 0
Code Activity

hive-c0re stops owning scheduled prompts; agents lose the schedule tools

Browse source 
Scheduled prompts are swarm-level now (swarmctl schedule, #4936). The hive
keeps the rows it has only so the operator can copy them over.

- hive-c0re no longer fires anything: the scheduled-prompts worker is gone,
  and with it every create/edit/cancel/pause/resume/fire path (dashboard
  endpoints, agent-socket verbs, the SchedulePrompt approval kind and its
  approve handler). The store keeps open + list; GET /api/schedules stays.
- A v6 approvals migration drops schedule_prompt rows.
- Nothing submits approvals any more, so the submit path goes too:
  Approvals::submit_kind (now a test fixture), the ApprovalAdded dashboard
  event and its emitter.
- Coordinator::push_todo, whose only caller was the worker, goes, and with
  it hive_host_sock::agent_todo_socket and hive-c0re's hive-agent-sock /
  hive-sock-client deps.
- The schedules_changed dashboard event goes; nothing mutates the rows.
- Agents lose request_schedule_prompt, cancel_schedule, fire_schedule_now,
  edit_schedule and list_schedules, the `scheduling` tool group, their
  system-prompt line and their transcript formatters. `remind` is unchanged.
This commit is contained in:
atlas 2026-10-04 21:23:18 +02:00
commit c3a54bbdd0
31 changed files with 273 additions and 3124 deletions
Download patch file Download diff file Expand all files Collapse all files

143
hive-c0re/src/coordinator.rs
View file

@ -31,9 +31,9 @@ const LAST_STOPPED_RUNNING_KEY: &str = "last_stopped_running";
pub struct Coordinator {
pub broker: Arc<Broker>,
pub approvals: Arc<Approvals>,
/// Scheduled-prompts queue. One sqlite connection,
/// internal mutex; the worker drains due rows and the manager
/// handlers insert / cancel through the same handle.
/// Scheduled-prompt rows, read-only: the dashboard lists them so the
/// operator can recreate them with `swarmctl schedule`. Nothing on
/// the hive writes or fires them.
pub scheduled_prompts: Arc<crate::scheduled_prompts::ScheduledPrompts>,
/// Full build-log capture. `lifecycle::run` /
/// `lifecycle::prebuild_toplevel` `start()` a row per attempt,
@ -426,17 +426,6 @@ pub struct ApprovalResolved<'a> {
pub description: Option<String>,
}
/// Field-named payload for [`Coordinator::emit_approval_added`].
/// Mirrors the `ApprovalAdded` dashboard-event fields. `agent`
/// borrows from the caller; `approval_kind` is a compile-time
/// constant.
pub struct ApprovalAdded<'a> {
pub id: i64,
pub agent: &'a str,
pub approval_kind: &'static str,
pub description: Option<String>,
}
/// The two ways [`Coordinator::set_paused_by_name`] can fail. See its own
/// doc comment for why this is a distinct type rather than one
/// `anyhow::Error` (`Write` already wraps one, `BadName` never needs to).
@ -587,42 +576,11 @@ impl Coordinator {
});
}
/// Emit a `SchedulesChanged` snapshot event. Called from every
/// schedule mutation site (operator API handlers + the worker
/// after each tick that fires or rearms a row) so the dashboard's
/// scheduled-prompts tab updates live without polling.
pub fn emit_schedules_snapshot(self: &Arc<Self>) {
let mut schedules: Vec<hive_sh4re::schedule::WireSchedule> =
match self.scheduled_prompts.list() {
Ok(rows) => rows
.into_iter()
.map(crate::socket_server::schedule_to_wire_public)
.collect(),
Err(e) => {
tracing::warn!(error = ?e, "emit_schedules_snapshot: list failed");
return;
}
};
// Strip ghost targets (destroyed agents) so the dashboard doesn't
// render dead columns. Best-effort: if the roster cache is
// momentarily contended we emit unfiltered rather than block this
// sync path — the next snapshot / page reload corrects it.
if let Some(live) = self.live_container_names_blocking() {
crate::socket_server::filter_ghost_schedule_targets(&mut schedules, &live);
}
self.emit_dashboard_event(DashboardEvent::SchedulesChanged {
seq: self.next_seq(),
schedules,
});
}
/// Best-effort synchronous snapshot of live container names (the
/// logical agent names from the last `nixos-container list` scan).
/// Returns `None` when the roster cache lock is momentarily
/// contended, so a transient miss is treated as "roster unknown,
/// don't filter" rather than hiding live schedule targets. The async
/// `containers_snapshot` is the reliable path for request handlers;
/// this exists for the sync `emit_schedules_snapshot` SSE emit.
/// contended. The async `containers_snapshot` is the reliable path
/// for request handlers; this exists for the sync SSE snapshot emits.
#[must_use]
pub fn live_container_names_blocking(&self) -> Option<HashSet<String>> {
self.last_containers
@ -772,24 +730,6 @@ impl Coordinator {
self.meta_updates_active.load(Ordering::SeqCst) > 0
}
/// Emit `ApprovalAdded` immediately after the row is inserted in
/// sqlite.
pub fn emit_approval_added(&self, ev: ApprovalAdded<'_>) {
let ApprovalAdded {
id,
agent,
approval_kind,
description,
} = ev;
self.emit_dashboard_event(DashboardEvent::ApprovalAdded {
seq: self.next_seq(),
id,
agent: agent.to_owned(),
approval_kind,
description,
});
}
/// Emit `ApprovalResolved` after `mark_approved` / `mark_denied` /
/// `mark_failed` lands. `resolved_at` is stamped from the system
/// clock here so call sites don't repeat the conversion; if you
@ -904,7 +844,7 @@ impl Coordinator {
.with_context(|| format!("create agent dir {}", agent_dir.display()))?;
let socket_path = Self::socket_path(name);
// Hand the full Coordinator to the per-agent socket — it needs
// more than just the broker (approvals, scheduled_prompts, ...).
// more than just the broker (approvals, ...).
let socket = socket_server::start(name, &socket_path, self.clone())?;
self.agents.lock().unwrap().insert(name.to_owned(), socket);
Ok(agent_dir)
@ -1176,77 +1116,6 @@ impl Coordinator {
.unwrap_or_else(|| hive_sh4re::manager::MANAGER_AGENT.to_owned())
}
/// Push a todo directly into `agent`'s in-container todo store — a
/// best-effort *live* dial of its `hive-agent-sock` socket
/// (`hive_host_sock::agent_todo_socket`), same `UpsertTodo` request
/// shape the built-in in-container producers (matrix/bash/forge-notify)
/// already send — `subsystem` isn't restricted to that set, any
/// producer can use the same shape. This is the migration target for
/// `HelperEvent` variants that are pure "FYI, check when convenient" notices: the
/// event stops being a broker `Message` (which always drives an
/// immediate turn) and becomes a todo instead, with the same
/// dedup-by-key semantics as any other producer.
///
/// Deliberately push, not queue: if the agent's container is down
/// (socket file absent) or the dial otherwise fails, this is a
/// silent no-op (logged at `debug`/`warn`) — no retry, no fallback
/// delivery. An agent that's down doesn't need a todo about
/// something it'll never see appear this way; whatever mechanism
/// resurfaces its state on the next boot is unrelated to this path.
///
/// Returns `Ok(())` on a successful push and `Err(reason)` — a
/// short human-readable string, already logged at `debug`/`warn`
/// by this function — on any of the silent-no-op cases below.
/// Most callers are pure fire-and-forget notices and ignore it
/// (`let _ = coord.push_todo(...).await;`); callers that track a
/// per-target delivery outcome (e.g. the scheduled-prompts worker's
/// `last_result` column) use it instead of assuming success.
/// `reopen_if_acked` forwards to `UpsertTodo` (see its doc comment) —
/// `false` for every caller here except the scheduled-prompts worker.
pub async fn push_todo(
&self,
agent: &str,
subsystem: &str,
key: Option<String>,
summary: String,
source: Option<String>,
reopen_if_acked: bool,
) -> Result<(), String> {
let Ok(ident) = hive_types::Ident::parse(agent) else {
tracing::warn!(%agent, "push_todo: not a valid agent ident, skipping");
return Err(format!("'{agent}' is not a valid agent ident"));
};
let path = hive_host_sock::agent_todo_socket(&ident);
if !path.exists() {
tracing::debug!(%agent, path = %path.display(), "push_todo: agent socket not present (offline?), skipping");
return Err(format!("agent '{agent}' socket not present (offline?)"));
}
let req = hive_agent_sock::Request::UpsertTodo {
subsystem: subsystem.to_owned(),
key,
summary,
source,
reopen_if_acked,
};
match hive_sock_client::request::<_, hive_agent_sock::Response>(
&path,
&req,
hive_sock_client::Retry::None,
)
.await
{
Ok(hive_agent_sock::Response::Err { message }) => {
tracing::warn!(%agent, %message, "push_todo: agent rejected the todo");
Err(format!("agent '{agent}' rejected the todo: {message}"))
}
Err(e) => {
tracing::warn!(%agent, error = ?e, "push_todo: dial failed");
Err(format!("push_todo dial to '{agent}' failed: {e:#}"))
}
Ok(_) => Ok(()),
}
}
/// Push a `HelperEvent` into an arbitrary agent's inbox. Encoded
/// the same way as `notify_manager` (sender = `SYSTEM_SENDER`,
/// body = JSON-encoded event) — e.g. `ContainerCrash`, `NeedsUpdate`.