phase 6: container events + drop the 5s /api/state poll

new DashboardEvent::ContainerStateChanged + ContainerRemoved
close the last refetch loop on the dashboard. Coordinator's
rescan_containers_and_emit diffs a fresh container_view::build_all
against a cached last_containers map and fires per-row events.
called from actions::approve (post-spawn), actions::destroy,
the lifecycle_action wrapper, auto_update::rebuild_agent, and
the existing 10s crash_watch poll.

ContainerView extracted to its own module so coordinator and
dashboard can both build it. dashboard endpoints flip to 200;
container-lifecycle forms carry data-no-refresh. client drops
the periodic poll entirely — initial cold load + SSE for
everything afterwards. pending overlay reads from the existing
transientsState since the new event payload doesn't carry it.

PURG3 + meta-update keep the post-submit refetch since
tombstones + meta_inputs aren't event-derived yet; tracked in
TODO.md.

CLAUDE.md

@@ -193,6 +193,26 @@ Prune freely.
   domain tooling — the agent flake's `inputs` block pulls
   the external flake, `agent.nix` references it via
   `flakeInputs.<name>.packages.${pkgs.system}.default`.
+- **Just landed:** Phase 6 container events. New
+  `DashboardEvent::ContainerStateChanged { container }` +
+  `ContainerRemoved { name }` close the last refetch loop on the
+  dashboard side. `Coordinator::rescan_containers_and_emit` builds a
+  fresh `container_view::build_all` snapshot, diffs it against a
+  cached `last_containers` map, and fires per-row events for the
+  delta. Called from every mutation site: `actions::approve`
+  (post-spawn), `actions::destroy`, the `lifecycle_action` wrapper
+  in `dashboard.rs` (start/stop/restart/rebuild), `auto_update::
+  rebuild_agent`, and the existing 10s `crash_watch` poll loop.
+  `ContainerView` extracted to its own module so coordinator +
+  dashboard can both build it. Dashboard endpoints (`/restart`,
+  `/destroy`, `/kill`, `/rebuild`, `/start`, `/update-all`,
+  `/meta-update`, `/purge-tombstone`) now return 200; matching
+  forms carry `data-no-refresh` where the event coverage is
+  complete (purge + meta-update keep the refetch since tombstones
+  + meta_inputs aren't event-derived yet). Client drops the 5s
+  periodic `/api/state` poll entirely — initial cold load + SSE
+  for everything afterwards; pending overlay reads from
+  `transientsState` since the new event payload doesn't carry it.
 - **Just landed:** dashboard event refactor. New `hive-fr0nt`
   workspace crate hosts shared frontend assets (palette + terminal
   CSS + `window.HiveTerminal.create` JS) so both the dashboard and

TODO.md

@@ -27,7 +27,8 @@
 - Per-agent reminder status (pending, delivered)
 - Reminder query interface for debugging
 - Display reminder delivery errors (failed sends, mark failures)
-- **Phase 6 leftovers** — event-covered endpoints (`/approve`, `/deny`, `/answer-question`, `/cancel-question`, `/request-spawn`) now return 200 (f559441); the matching forms carry `data-no-refresh` so the post-submit `/api/state` refetch is skipped (the SSE event delivers the update). Container-lifecycle endpoints (`/restart`, `/destroy`, `/kill`, `/rebuild`, `/start`, `/api/{cancel,compact,model,new-session}`, `/meta-update`, `/purge-tombstone`) still need a `ContainerListChanged` event before their redirects can drop — `ContainerView` is currently sourced from external `nixos-container list`, so the 5s poll continues to drive that section.
+- **Phase 6 follow-ups** — dashboard side is fully event-driven (Phase 6 leftovers landed); the per-agent web UI's lifecycle endpoints (`/api/{cancel,compact,model,new-session}`, `/login/*`) still 303-redirect-and-poll. Convert them to 200 + `data-no-refresh` so the per-agent page stops refetching `/api/state` on every operator click — `LiveEvent::Note` already covers cancel/compact/model/new-session, login state needs its own `NeedsLogin` / `LoggedIn` events on the per-agent bus.
+- **Tombstones + meta_inputs events**: not yet event-derived. PURG3 + meta-update still trigger a post-submit `/api/state` refetch on the dashboard. Add `TombstoneAdded`/`TombstoneRemoved` + `MetaInputsChanged` so those forms can drop their refetch too and the cold-load is the only `/api/state` fetch in normal operation.
 
 ## Bugs
 

hive-c0re/assets/app.js

@@ -214,6 +214,27 @@
     }
   });
 
+  // Derived container state — cold-loaded from /api/state.containers,
+  // then mutated live by `container_state_changed` (upsert by name)
+  // and `container_removed` (drop by name). The coordinator's rescan
+  // helper fires these after every mutation site + on a periodic poll
+  // in crash_watch. Keyed by ContainerView.name so the lifecycle
+  // forms' POST → 200 → matching event flips the row without a
+  // snapshot refetch.
+  const containersState = new Map();
+  function syncContainersFromSnapshot(s) {
+    containersState.clear();
+    for (const c of s.containers || []) containersState.set(c.name, c);
+  }
+  function applyContainerStateChanged(ev) {
+    if (!ev.container || !ev.container.name) return;
+    containersState.set(ev.container.name, ev.container);
+    renderContainersFromState();
+  }
+  function applyContainerRemoved(ev) {
+    if (containersState.delete(ev.name)) renderContainersFromState();
+  }
+
   // Derived transient state — cold-loaded from /api/state.transients,
   // then mutated live by `transient_set` / `transient_cleared`. Keyed
   // by agent name so add/remove are O(1). `since_unix` is wall-clock so
@@ -251,27 +272,56 @@
     if (s) renderContainers(s);
   }
 
+  // Re-derive port conflicts from the live containers map. Mirrors the
+  // server-side `build_port_conflicts` so the banner reacts to event
+  // updates instead of waiting for a /api/state refetch.
+  function derivePortConflicts(containers) {
+    const byPort = new Map();
+    for (const c of containers) {
+      if (!byPort.has(c.port)) byPort.set(c.port, []);
+      byPort.get(c.port).push(c.name);
+    }
+    const out = [];
+    for (const [port, agents] of byPort) {
+      if (agents.length > 1) {
+        agents.sort();
+        out.push({ port, agents });
+      }
+    }
+    out.sort((a, b) => a.port - b.port);
+    return out;
+  }
+
   // ─── state rendering ────────────────────────────────────────────────────
   function renderContainers(s) {
     const root = $('containers-section');
     root.innerHTML = '';
 
+    // Containers come from the derived map (event-driven) rather than
+    // `s.containers`; `s` still supplies hostname (for the web-ui
+    // link) and tombstones/meta_inputs (not event-derived yet).
+    const containers = Array.from(containersState.values())
+      .sort((a, b) => a.name.localeCompare(b.name));
+    const portConflicts = derivePortConflicts(containers);
+    const anyStale = containers.some((c) => c.needs_update);
+
     // Port-hash collisions: rename one of the listed agents and
     // rebuild. The banner sits above the agent list so it's the
     // first thing the operator sees when something's wedged.
-    if (s.port_conflicts && s.port_conflicts.length) {
+    if (portConflicts.length) {
       const banner = el('div', { class: 'port-conflict' },
         el('strong', {}, '⚠  port collision'), ' — ');
-      const groups = s.port_conflicts.map((c) =>
+      const groups = portConflicts.map((c) =>
         `:${c.port} (${c.agents.join(' + ')})`).join('; ');
       banner.append(groups + '. rename one of each and ↻ R3BU1LD.');
       root.append(banner);
     }
 
-    if (s.any_stale) {
+    if (anyStale) {
       root.append(form(
         '/update-all', 'btn-rebuild', '↻ UPD4TE 4LL',
         'rebuild every stale container?',
+        {}, { noRefresh: true },
       ));
     }
 
@@ -290,15 +340,20 @@
       root.append(ul);
     }
 
-    if (!s.containers.length && !transientsState.size) {
+    if (!containers.length && !transientsState.size) {
       root.append(el('p', { class: 'empty' }, 'no managed containers'));
       return;
     }
 
+    const hostname = (s && s.hostname) || window.location.hostname;
     const ul = el('ul', { class: 'containers' });
-    for (const c of s.containers) {
+    for (const c of containers) {
-      const url = `http://${s.hostname}:${c.port}/`;
+      const url = `http://${hostname}:${c.port}/`;
-      const li = el('li', { class: 'container-row' + (c.pending ? ' pending' : '') });
+      // Pending state is overlaid from the transient store, not from
+      // the container row — `ContainerStateChanged` doesn't carry it,
+      // `TransientSet` / `TransientCleared` do.
+      const pending = transientsState.get(c.name)?.kind || null;
+      const li = el('li', { class: 'container-row' + (pending ? ' pending' : '') });
 
       // ── line 1: identity ─────────────────────────────────────────
       const head = el('div', { class: 'head' });
@@ -307,9 +362,9 @@
         el('span', { class: c.is_manager ? 'role role-m1nd' : 'role role-ag3nt' },
           c.is_manager ? 'm1nd' : 'ag3nt'),
       );
-      if (c.pending) {
+      if (pending) {
         head.append(el('span', { class: 'pending-state' },
-          el('span', { class: 'spinner' }, '◐'), ' ', c.pending + '…'));
+          el('span', { class: 'spinner' }, '◐'), ' ', pending + '…'));
       } else if (c.needs_login) {
         head.append(el('a',
           { class: 'badge badge-warn', href: url, target: '_blank', rel: 'noopener' },
@@ -319,6 +374,7 @@
         head.append(form(
           '/rebuild/' + c.name, 'badge badge-warn btn-inline', 'needs update ↻',
           'rebuild ' + c.name + '? hot-reloads the container.',
+          {}, { noRefresh: true },
         ));
       }
       head.append(el('span', { class: 'meta' }, `${c.container} :${c.port}`));
@@ -333,29 +389,37 @@
       const actions = el('div', { class: 'actions' });
       if (c.running) {
         actions.append(
-          form('/restart/' + c.name, 'btn-restart', '↺ R3ST4RT', 'restart ' + c.name + '?'),
+          form('/restart/' + c.name, 'btn-restart', '↺ R3ST4RT',
+            'restart ' + c.name + '?', {}, { noRefresh: true }),
         );
         if (!c.is_manager) {
           actions.append(
-            form('/kill/' + c.name, 'btn-stop', '■ ST0P', 'stop ' + c.name + '?'),
+            form('/kill/' + c.name, 'btn-stop', '■ ST0P',
+              'stop ' + c.name + '?', {}, { noRefresh: true }),
           );
         }
       } else {
         actions.append(
-          form('/start/' + c.name, 'btn-start', '▶ ST4RT', 'start ' + c.name + '?'),
+          form('/start/' + c.name, 'btn-start', '▶ ST4RT',
+            'start ' + c.name + '?', {}, { noRefresh: true }),
         );
       }
       actions.append(
         form('/rebuild/' + c.name, 'btn-rebuild', '↻ R3BU1LD',
-          'rebuild ' + c.name + '? hot-reloads the container.'),
+          'rebuild ' + c.name + '? hot-reloads the container.',
+          {}, { noRefresh: true }),
       );
       if (!c.is_manager) {
+        // DESTR0Y is event-covered (ContainerRemoved); PURG3 also
+        // wipes tombstone state which isn't event-derived yet, so it
+        // keeps the post-submit refetch.
         actions.append(
           form('/destroy/' + c.name, 'btn-destroy', 'DESTR0Y',
-            'destroy ' + c.name + '? container is removed; state + creds kept.'),
+            'destroy ' + c.name + '? container is removed; state + creds kept.',
+            {}, { noRefresh: true }),
           form('/destroy/' + c.name, 'btn-destroy', 'PURG3',
             'PURGE ' + c.name + '? container, config history, claude creds, '
-            + 'and /state/ notes are all WIPED. no undo.', { purge: 'on' }),
+            + 'and notes are all WIPED. no undo.', { purge: 'on' }),
         );
       }
       li.append(actions);
@@ -1088,10 +1152,11 @@
       // names from here instead of refetching on every keystroke).
       window.__hyperhive_state = s;
       const openDetails = snapshotOpenDetails();
-      // Sync transients first so renderContainers below sees the
+      // Sync transients + containers first so renderContainers below
-      // current derived map (it reads from `transientsState`, not
+      // sees the current derived maps (it reads from
-      // from `s.transients`).
+      // `transientsState` + `containersState`, not from `s.*`).
       syncTransientsFromSnapshot(s);
+      syncContainersFromSnapshot(s);
       renderContainers(s);
       renderTombstones(s);
       // Sync the derived approvals + questions stores from the
@@ -1106,18 +1171,20 @@
       renderMetaInputs(s);
       restoreOpenDetails(openDetails);
       notifyDeltas(s);
-      // Auto-refresh: fast (2s) while a spawn or a per-container
+      // No periodic refresh timer. Phase 6 covers every container
-      // action is in flight, otherwise heartbeat (5s) so newly-queued
+      // mutation with `ContainerStateChanged` / `ContainerRemoved`
-      // approvals from the manager show up without the operator
+      // (lifecycle ops, destroy, rebuild, crash_watch's 10s poll);
-      // having to reload the page. Broker SSE already triggers a
+      // approvals + questions + transients have their own events;
-      // refresh on operator-bound messages; this catches the rest
+      // broker traffic flows through the SSE channel. The only
-      // (approvals, tombstones, questions).
+      // /api/state fetches are the initial cold load and the
-      const anyPending = s.containers.some((c) => c.pending);
+      // post-submit refetch on forms without `data-no-refresh`
-      const next = (transientsState.size || anyPending) ? 2000 : 5000;
+      // (tombstones, meta-input updates).
       if (pollTimer) { clearTimeout(pollTimer); pollTimer = null; }
-      if (next) pollTimer = setTimeout(refreshState, next);
     } catch (err) {
       console.error('refreshState failed', err);
+      // Schedule a single retry on transient errors so the page
+      // recovers from a brief network blip without making the
+      // operator reload.
       pollTimer = setTimeout(refreshState, 5000);
     }
   }
@@ -1171,6 +1238,8 @@
         question_resolved: (ev) => { applyQuestionResolved(ev); },
         transient_set:     (ev) => { applyTransientSet(ev); },
         transient_cleared: (ev) => { applyTransientCleared(ev); },
+        container_state_changed: (ev) => { applyContainerStateChanged(ev); },
+        container_removed:       (ev) => { applyContainerRemoved(ev); },
       },
       // Both history backfill and live frames flow through here, so the
       // inbox section ends up populated correctly on first paint and
@@ -1208,15 +1277,14 @@
       prompt.textContent = stickyTo ? `@${stickyTo}>` : '@—>';
     }
     function knownAgents() {
-      const s = window.__hyperhive_state;
+      // Read live from the derived containers map so newly-spawned
-      if (!s || !Array.isArray(s.containers)) return [];
+      // agents become addressable without an /api/state refetch.
-      // The broker uses the literal recipient `manager` for the
+      // Broker uses the literal recipient `manager` for the manager's
-      // manager's inbox, not the container name `hm1nd`. Swap on
+      // inbox, not the container name `hm1nd`.
-      // suggestion so `@manager` Just Works.
+      const names = Array.from(containersState.values())
-      const names = s.containers.map((c) => (c.is_manager ? 'manager' : c.name));
+        .map((c) => (c.is_manager ? 'manager' : c.name));
-      // `*` fans out the message to every registered agent (server-side
+      // `*` fans out to every registered agent (server-side
-      // broadcast_send). Surface it as a suggestion so operators can
+      // broadcast_send).
-      // type `@*` from the dashboard the same way the manager does.
       names.unshift('*');
       return names;
     }

hive-c0re/src/actions.rs

@@ -85,6 +85,10 @@ pub async fn approve(coord: Arc<Coordinator>, id: i64) -> Result<()> {
                 if let Err(e) = finish_approval(&coord_bg, &approval_bg, result, None) {
                     tracing::warn!(agent = %agent_bg, error = ?e, "spawn approval failed");
                 }
+                // New container row appeared (or didn't, on failure
+                // before nixos-container create completed) — rescan so
+                // dashboards reflect the post-spawn state.
+                coord_bg.rescan_containers_and_emit().await;
             });
             Ok(())
         }
@@ -355,6 +359,9 @@ pub async fn destroy(coord: &Arc<Coordinator>, name: &str, purge: bool) -> Resul
     coord.notify_manager(&HelperEvent::Destroyed {
         agent: name.to_owned(),
     });
+    // Container row disappeared — rescan so the dashboard fires
+    // `ContainerRemoved` for the gone row.
+    coord.rescan_containers_and_emit().await;
     Ok(())
 }
 

hive-c0re/src/auto_update.rs

@@ -95,6 +95,10 @@ pub async fn rebuild_agent(coord: &Arc<Coordinator>, name: &str, current_rev: &s
             // dashboard's meta-input update path — all of which
             // route through rebuild_agent.
             coord.kick_agent(name, "container rebuilt");
+            // Container state (needs_update, deployed_sha) may have
+            // shifted — rescan so dashboards drop the "needs update"
+            // chip without waiting for the next /api/state poll.
+            coord.rescan_containers_and_emit().await;
         }
         Err(e) => {
             coord.notify_manager(&hive_sh4re::HelperEvent::Rebuilt {
@@ -104,6 +108,7 @@ pub async fn rebuild_agent(coord: &Arc<Coordinator>, name: &str, current_rev: &s
                 sha: None,
                 tag: None,
             });
+            coord.rescan_containers_and_emit().await;
         }
     }
     result

hive-c0re/src/container_view.rs

@@ -0,0 +1,121 @@
+//! `ContainerView` + the snapshot builder that turns
+//! `nixos-container list` (plus per-agent state on disk) into the row
+//! shape the dashboard renders. Extracted from `dashboard.rs` so the
+//! coordinator's rescan-and-emit helper can build the same view and
+//! diff against the last snapshot to fire
+//! `ContainerStateChanged` / `ContainerRemoved` events.
+
+use std::collections::HashMap;
+use std::path::Path;
+
+use serde::Serialize;
+
+use crate::coordinator::Coordinator;
+use crate::lifecycle::{self, AGENT_PREFIX, MANAGER_NAME};
+
+#[derive(Serialize, Clone, PartialEq, Eq, Debug)]
+#[allow(clippy::struct_excessive_bools)]
+pub struct ContainerView {
+    /// Logical agent name (no `h-` prefix). Used in action URLs.
+    pub name: String,
+    /// Container name as nixos-container sees it (`h-foo`, `hm1nd`).
+    pub container: String,
+    pub is_manager: bool,
+    pub port: u16,
+    pub running: bool,
+    pub needs_update: bool,
+    pub needs_login: bool,
+    /// First 12 chars of the sha the meta flake currently has locked
+    /// for this agent's input.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub deployed_sha: Option<String>,
+}
+
+/// Build the full container list. Wraps `lifecycle::list()` and
+/// resolves every per-agent attribute the dashboard surfaces.
+pub async fn build_all(coord: &Coordinator) -> Vec<ContainerView> {
+    let raw = lifecycle::list().await.unwrap_or_default();
+    let current_rev = crate::auto_update::current_flake_rev(&coord.hyperhive_flake);
+    let locked = read_meta_locked_revs();
+    let mut out = Vec::new();
+    for c in &raw {
+        let (logical, is_manager) = if c == MANAGER_NAME {
+            (MANAGER_NAME.to_owned(), true)
+        } else if let Some(n) = c.strip_prefix(AGENT_PREFIX) {
+            (n.to_owned(), false)
+        } else {
+            continue;
+        };
+        let needs_update =
+            current_rev.as_deref().is_some_and(|rev| crate::auto_update::agent_needs_update(&logical, rev));
+        let needs_login =
+            !is_manager && !claude_has_session(&Coordinator::agent_claude_dir(&logical));
+        let deployed_sha = locked
+            .get(&format!("agent-{logical}"))
+            .map(|s| s[..s.len().min(12)].to_owned());
+        out.push(ContainerView {
+            port: lifecycle::agent_web_port(&logical),
+            running: lifecycle::is_running(&logical).await,
+            container: c.clone(),
+            name: logical,
+            is_manager,
+            needs_update,
+            needs_login,
+            deployed_sha,
+        });
+    }
+    out
+}
+
+/// Host-side mirror of `hive_ag3nt::login::has_session`. Returns true
+/// if the agent's bound `~/.claude/` dir on disk contains any regular
+/// file. Reads each `build_all()` so a login driven from the agent's
+/// own web UI reflects on the next snapshot.
+pub fn claude_has_session(dir: &Path) -> bool {
+    let Ok(entries) = std::fs::read_dir(dir) else {
+        return false;
+    };
+    entries
+        .flatten()
+        .any(|e| e.file_type().is_ok_and(|t| t.is_file()))
+}
+
+/// Map of `agent-<n>` → locked sha from meta's flake.lock. Used to
+/// render the `deployed:<sha12>` chip per container row.
+fn read_meta_locked_revs() -> HashMap<String, String> {
+    let mut out = HashMap::new();
+    let Ok(raw) = std::fs::read_to_string("/var/lib/hyperhive/meta/flake.lock") else {
+        return out;
+    };
+    let Ok(json) = serde_json::from_str::<serde_json::Value>(&raw) else {
+        return out;
+    };
+    let Some(nodes) = json.get("nodes").and_then(|v| v.as_object()) else {
+        return out;
+    };
+    let Some(root_name) = json.get("root").and_then(|v| v.as_str()) else {
+        return out;
+    };
+    let Some(root_inputs) = nodes
+        .get(root_name)
+        .and_then(|n| n.get("inputs"))
+        .and_then(|v| v.as_object())
+    else {
+        return out;
+    };
+    for alias in root_inputs.keys() {
+        let target_name = match root_inputs.get(alias) {
+            Some(serde_json::Value::String(s)) => s.clone(),
+            _ => continue,
+        };
+        if let Some(rev) = nodes
+            .get(&target_name)
+            .and_then(|n| n.get("locked"))
+            .and_then(|v| v.get("rev"))
+            .and_then(|v| v.as_str())
+        {
+            out.insert(alias.clone(), rev.to_owned());
+        }
+    }
+    out
+}

hive-c0re/src/coordinator.rs

@@ -13,6 +13,7 @@ use tokio::sync::broadcast;
 use crate::agent_server::{self, AgentSocket};
 use crate::approvals::Approvals;
 use crate::broker::Broker;
+use crate::container_view::{self, ContainerView};
 use crate::dashboard_events::DashboardEvent;
 use crate::operator_questions::OperatorQuestions;
 
@@ -64,6 +65,14 @@ pub struct Coordinator {
     /// snapshot.
     dashboard_events: broadcast::Sender<DashboardEvent>,
     event_seq: AtomicU64,
+    /// Last container snapshot seen by `rescan_containers_and_emit`,
+    /// keyed by `ContainerView.name`. The rescan diffs a fresh
+    /// `container_view::build_all` against this map and emits one
+    /// `ContainerStateChanged` per added/changed row and one
+    /// `ContainerRemoved` per disappeared row. Async — guarded by a
+    /// tokio mutex so the rescan can `await` `lifecycle::list` /
+    /// `is_running` without blocking other coordinator paths.
+    last_containers: tokio::sync::Mutex<HashMap<String, ContainerView>>,
 }
 
 /// Per-agent in-progress state that the dashboard surfaces between approve
@@ -142,6 +151,7 @@ impl Coordinator {
             transient: Mutex::new(HashMap::new()),
             dashboard_events,
             event_seq: AtomicU64::new(0),
+            last_containers: tokio::sync::Mutex::new(HashMap::new()),
         })
     }
 
@@ -291,6 +301,68 @@ impl Coordinator {
         });
     }
 
+    /// Rebuild the per-container snapshot, diff it against the last
+    /// one cached on `self`, and emit one
+    /// `DashboardEvent::ContainerStateChanged` per added/changed row
+    /// and one `DashboardEvent::ContainerRemoved` per disappeared row.
+    /// Call after any mutation that could affect what
+    /// `nixos-container list` returns or what a row's
+    /// `running` / `needs_update` / `needs_login` / `deployed_sha`
+    /// resolves to — lifecycle ops, destroy, approve (post-spawn),
+    /// rebuild, meta-update, and the crash-watcher's periodic poll.
+    /// Cheap when nothing changed (one `nixos-container list` + a
+    /// HashMap diff + zero emits).
+    pub async fn rescan_containers_and_emit(self: &Arc<Self>) {
+        let fresh = container_view::build_all(self).await;
+        let mut last = self.last_containers.lock().await;
+        let mut changed_or_new = Vec::new();
+        let mut removed = Vec::new();
+        // Diff into change vs. add.
+        for view in &fresh {
+            match last.get(&view.name) {
+                Some(prev) if prev == view => {} // unchanged
+                _ => changed_or_new.push(view.clone()),
+            }
+        }
+        // Anything in `last` but not in `fresh` is gone.
+        let fresh_names: std::collections::HashSet<&str> =
+            fresh.iter().map(|c| c.name.as_str()).collect();
+        for name in last.keys() {
+            if !fresh_names.contains(name.as_str()) {
+                removed.push(name.clone());
+            }
+        }
+        // Rebuild the cache from the fresh snapshot.
+        last.clear();
+        for c in fresh {
+            last.insert(c.name.clone(), c);
+        }
+        drop(last);
+        for c in changed_or_new {
+            self.emit_dashboard_event(DashboardEvent::ContainerStateChanged {
+                seq: self.next_seq(),
+                container: c,
+            });
+        }
+        for name in removed {
+            self.emit_dashboard_event(DashboardEvent::ContainerRemoved {
+                seq: self.next_seq(),
+                name,
+            });
+        }
+    }
+
+    /// Read-only snapshot of the last cached container view. Used by
+    /// `/api/state` to cold-load page-open clients without re-running
+    /// `nixos-container list` themselves; the
+    /// `rescan_containers_and_emit` calls keep this fresh.
+    pub async fn containers_snapshot(&self) -> Vec<ContainerView> {
+        let last = self.last_containers.lock().await;
+        let mut out: Vec<ContainerView> = last.values().cloned().collect();
+        out.sort_by(|a, b| a.name.cmp(&b.name));
+        out
+    }
+
     pub fn register_agent(self: &Arc<Self>, name: &str) -> Result<PathBuf> {
         // Idempotent: drop any existing listener so re-registration (e.g. on rebuild,
         // or after a hive-c0re restart cleared /run/hyperhive) gets a fresh socket.

hive-c0re/src/crash_watch.rs

@@ -16,10 +16,10 @@
 //! but polling is simpler and a 10s detection delay is fine.
 
 use std::collections::HashSet;
-use std::path::Path;
 use std::sync::Arc;
 use std::time::Duration;
 
+use crate::container_view::claude_has_session;
 use crate::coordinator::{Coordinator, TransientKind};
 use crate::lifecycle::{self, AGENT_PREFIX, MANAGER_NAME};
 
@@ -69,6 +69,12 @@ pub fn spawn(coord: Arc<Coordinator>) {
                 emit_login_transitions(&coord, &prev_logged_in, &current_logged_in, &sub_agents);
                 emit_update_transitions(&coord, &prev_updated, &current_updated, &sub_agents);
             }
+            // Periodic container rescan — catches state flips that
+            // happen outside our mutation surface (operator runs
+            // `nixos-container stop` over ssh, agent logs in via its
+            // own web UI, etc.) so the dashboard converges within one
+            // POLL_INTERVAL. Idempotent + cheap when nothing changed.
+            coord.rescan_containers_and_emit().await;
             prev_running = current_running;
             prev_logged_in = current_logged_in;
             prev_updated = current_updated;
@@ -163,14 +169,3 @@ fn emit_update_transitions(
         });
     }
 }
-
-/// Mirrors `dashboard::claude_has_session`. Lives here too so the
-/// watcher doesn't depend on dashboard internals.
-fn claude_has_session(dir: &Path) -> bool {
-    let Ok(entries) = std::fs::read_dir(dir) else {
-        return false;
-    };
-    entries
-        .flatten()
-        .any(|e| e.file_type().is_ok_and(|t| t.is_file()))
-}

hive-c0re/src/dashboard.rs

@@ -14,7 +14,7 @@ use axum::{
     extract::{Path as AxumPath, State},
     http::{HeaderMap, StatusCode},
     response::{
-        Html, IntoResponse, Redirect, Response,
+        Html, IntoResponse, Response,
         sse::{Event, KeepAlive, Sse},
     },
     routing::{get, post},
@@ -25,8 +25,9 @@ use tokio_stream::wrappers::BroadcastStream;
 use tokio_stream::{Stream, StreamExt};
 
 use crate::actions;
+use crate::container_view::{ContainerView, claude_has_session};
 use crate::coordinator::Coordinator;
-use crate::lifecycle::{self, AGENT_PREFIX, MANAGER_NAME};
+use crate::lifecycle::{self, MANAGER_NAME};
 
 const MANAGER_PORT: u16 = 8000;
 
@@ -200,31 +201,6 @@ struct TombstoneView {
     has_creds: bool,
 }
 
-#[derive(Serialize)]
-#[allow(clippy::struct_excessive_bools)]
-struct ContainerView {
-    /// Logical agent name (no `h-` prefix). Used in action URLs.
-    name: String,
-    /// Container name as nixos-container sees it (`h-foo`, `hm1nd`).
-    container: String,
-    is_manager: bool,
-    port: u16,
-    running: bool,
-    needs_update: bool,
-    needs_login: bool,
-    /// When a lifecycle action is in flight on this container, the kind
-    /// (`starting`, `stopping`, etc.) so the JS can render a spinner +
-    /// disable other buttons.
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pending: Option<&'static str>,
-    /// First 12 chars of the sha the meta flake currently has locked
-    /// for this agent's input. Reflects what's actually deployed; can
-    /// differ from `applied/<n>/main` only between
-    /// `meta::prepare_deploy` and `finalize_deploy` (≤ build duration).
-    #[serde(skip_serializing_if = "Option::is_none")]
-    deployed_sha: Option<String>,
-}
-
 #[derive(Serialize)]
 struct TransientView {
     name: String,
@@ -303,17 +279,20 @@ async fn api_state(headers: HeaderMap, State(state): State<AppState>) -> axum::J
     // to make idempotent, not ours to avoid here.
     let seq = state.coord.current_seq();
 
-    let raw_containers = log_default("nixos-container list", lifecycle::list().await);
+    // Refresh the coordinator's cached container snapshot before
-    let current_rev = crate::auto_update::current_flake_rev(&state.coord.hyperhive_flake);
+    // reading. Cold-load clients then see whatever the latest rescan
+    // produced; live clients converge via the matching
+    // `ContainerStateChanged` / `ContainerRemoved` events the rescan
+    // emits.
+    state.coord.rescan_containers_and_emit().await;
+    let containers = state.coord.containers_snapshot().await;
+    let any_stale = containers.iter().any(|c| c.needs_update);
     let transient_snapshot = state.coord.transient_snapshot();
     let pending_approvals = gc_orphans(
         &state.coord,
         log_default("approvals.pending", state.coord.approvals.pending()),
     );
-
+    let transients = build_transient_views(&containers, &transient_snapshot);
-    let (containers, any_stale) =
-        build_container_views(&raw_containers, current_rev.as_deref(), &transient_snapshot).await;
-    let transients = build_transient_views(&raw_containers, &transient_snapshot);
     let approvals = build_approval_views(pending_approvals).await;
     let approval_history = log_default(
         "approvals.recent_resolved",
@@ -370,96 +349,6 @@ fn build_port_conflicts(containers: &[ContainerView]) -> Vec<PortConflict> {
         .collect()
 }
 
-/// Build `ContainerView`s for every live nixos-container. Returns the
-/// list and whether any container is stale (drives the "↻ UPD4TE 4LL"
-/// banner).
-async fn build_container_views(
-    raw_containers: &[String],
-    current_rev: Option<&str>,
-    transient_snapshot: &std::collections::HashMap<String, crate::coordinator::TransientState>,
-) -> (Vec<ContainerView>, bool) {
-    let mut out = Vec::new();
-    let mut any_stale = false;
-    let locked = read_meta_locked_revs();
-    for c in raw_containers {
-        let (logical, is_manager) = if c == MANAGER_NAME {
-            (MANAGER_NAME.to_owned(), true)
-        } else if let Some(n) = c.strip_prefix(AGENT_PREFIX) {
-            (n.to_owned(), false)
-        } else {
-            continue;
-        };
-        let needs_update =
-            current_rev.is_some_and(|rev| crate::auto_update::agent_needs_update(&logical, rev));
-        if needs_update {
-            any_stale = true;
-        }
-        let needs_login =
-            !is_manager && !claude_has_session(&Coordinator::agent_claude_dir(&logical));
-        let pending = transient_snapshot
-            .get(&logical)
-            .map(|st| transient_label(st.kind));
-        let deployed_sha = locked
-            .get(&format!("agent-{logical}"))
-            .map(|s| s[..s.len().min(12)].to_owned());
-        out.push(ContainerView {
-            port: lifecycle::agent_web_port(&logical),
-            running: lifecycle::is_running(&logical).await,
-            container: c.clone(),
-            name: logical,
-            is_manager,
-            needs_update,
-            needs_login,
-            pending,
-            deployed_sha,
-        });
-    }
-    (out, any_stale)
-}
-
-/// Map of node name → locked sha for nodes the **root** of meta
-/// directly depends on (`hyperhive`, `agent-<n>`). Used by the
-/// container row to render its `deployed:<sha12>` chip per agent.
-/// Distinct from `read_meta_inputs()` which walks deeper for the
-/// flake-input update form.
-fn read_meta_locked_revs() -> std::collections::HashMap<String, String> {
-    let mut out = std::collections::HashMap::new();
-    let Ok(raw) = std::fs::read_to_string("/var/lib/hyperhive/meta/flake.lock") else {
-        return out;
-    };
-    let Ok(json) = serde_json::from_str::<serde_json::Value>(&raw) else {
-        return out;
-    };
-    let Some(nodes) = json.get("nodes").and_then(|v| v.as_object()) else {
-        return out;
-    };
-    let Some(root_name) = json.get("root").and_then(|v| v.as_str()) else {
-        return out;
-    };
-    let Some(root_inputs) = nodes
-        .get(root_name)
-        .and_then(|n| n.get("inputs"))
-        .and_then(|v| v.as_object())
-    else {
-        return out;
-    };
-    for alias in root_inputs.keys() {
-        let target_name = match root_inputs.get(alias) {
-            Some(serde_json::Value::String(s)) => s.clone(),
-            _ => continue,
-        };
-        if let Some(rev) = nodes
-            .get(&target_name)
-            .and_then(|n| n.get("locked"))
-            .and_then(|v| v.get("rev"))
-            .and_then(|v| v.as_str())
-        {
-            out.insert(alias.clone(), rev.to_owned());
-        }
-    }
-    out
-}
-
 #[derive(Serialize, Clone)]
 struct MetaInputView {
     /// Input key in meta's `flake.nix` — `hyperhive`, `agent-<n>`, etc.
@@ -577,16 +466,12 @@ fn walk_meta_inputs(
 /// (`Spawning`). Lifecycle ops on existing containers surface as
 /// `ContainerView.pending` inline; this list only catches pre-creation.
 fn build_transient_views(
-    raw_containers: &[String],
+    containers: &[ContainerView],
     transient_snapshot: &std::collections::HashMap<String, crate::coordinator::TransientState>,
 ) -> Vec<TransientView> {
     transient_snapshot
         .iter()
-        .filter(|(name, _)| {
+        .filter(|(name, _)| !containers.iter().any(|c| &c.name == *name))
-            !raw_containers
-                .iter()
-                .any(|c| c == &format!("{AGENT_PREFIX}{name}") || c == *name)
-        })
         .map(|(name, st)| TransientView {
             name: name.clone(),
             kind: transient_label(st.kind),
@@ -1034,7 +919,10 @@ async fn post_purge_tombstone(
         .fail_pending_for_agent(&name, "agent state purged");
     if errors.is_empty() {
         tracing::info!(%name, "tombstone purged");
-        Redirect::to("/").into_response()
+        // Tombstones aren't event-derived yet, so the client still
+        // refetches /api/state to see this one disappear (matching
+        // form omits `data-no-refresh`).
+        (StatusCode::OK, "ok").into_response()
     } else {
         error_response(&format!("purge {name} partial: {}", errors.join(", ")))
     }
@@ -1086,7 +974,10 @@ async fn post_meta_update(
     tokio::spawn(async move {
         run_meta_update(&coord, &inputs_clone).await;
     });
-    Redirect::to("/").into_response()
+    // Background task — each per-agent rebuild emits its own
+    // `ContainerStateChanged`; the meta inputs panel still relies on
+    // /api/state freshness (matching form omits `data-no-refresh`).
+    (StatusCode::OK, "ok").into_response()
 }
 
 /// Background task: run `nix flake update <inputs>` in meta + commit,
@@ -1260,7 +1151,13 @@ where
     match result {
         Ok(()) => {
             extra(state, &logical);
-            Redirect::to("/").into_response()
+            // Rescan so the running/needs_login/needs_update flip on
+            // the affected row lands on every dashboard's SSE channel
+            // without waiting for a snapshot poll. 200 + matching
+            // `data-no-refresh` on the form skip the post-submit
+            // /api/state refetch.
+            state.coord.rescan_containers_and_emit().await;
+            (StatusCode::OK, "ok").into_response()
         }
         Err(e) => error_response(&format!("{verb} {logical} failed: {e:#}")),
     }
@@ -1336,7 +1233,8 @@ async fn post_update_all(State(state): State<AppState>) -> Response {
         }
     }
     if errors.is_empty() {
-        Redirect::to("/").into_response()
+        // Each rebuild_agent rescanned; no extra refetch needed.
+        (StatusCode::OK, "ok").into_response()
     } else {
         error_response(&format!(
             "update-all partial failure:\n{}",
@@ -1380,8 +1278,11 @@ async fn post_destroy(
 ) -> Response {
     // Checkbox semantics: any non-empty value (axum sends "on") = purge.
     let purge = form.purge.as_deref().is_some_and(|v| !v.is_empty());
+    // `actions::destroy` rescans the container list on success, so the
+    // `ContainerRemoved` event lands before we return 200. The matching
+    // form carries `data-no-refresh`.
     match actions::destroy(&state.coord, &name, purge).await {
-        Ok(()) => Redirect::to("/").into_response(),
+        Ok(()) => (StatusCode::OK, "ok").into_response(),
         Err(e) => error_response(&format!("destroy {name} failed: {e:#}")),
     }
 }
@@ -1429,18 +1330,6 @@ fn gc_orphans(coord: &Coordinator, approvals: Vec<Approval>) -> Vec<Approval> {
         .collect()
 }
 
-/// Host-side mirror of `hive_ag3nt::login::has_session`. Returns true if the
-/// agent's bound `~/.claude/` dir on disk contains any regular file. The
-/// dashboard reads this each render so logins driven from the agent web UI
-/// (Phase 8 step 4) reflect within one auto-refresh cycle.
-fn claude_has_session(dir: &Path) -> bool {
-    let Ok(entries) = std::fs::read_dir(dir) else {
-        return false;
-    };
-    entries
-        .flatten()
-        .any(|e| e.file_type().is_ok_and(|t| t.is_file()))
-}
 
 /// Multi-file unified diff between the currently-deployed tree and
 /// the proposal for this approval. Runs against the applied repo

hive-c0re/src/dashboard_events.rs

@@ -25,6 +25,8 @@
 
 use serde::Serialize;
 
+use crate::container_view::ContainerView;
+
 #[derive(Debug, Clone, Serialize)]
 #[serde(rename_all = "snake_case", tag = "kind")]
 pub enum DashboardEvent {
@@ -121,4 +123,24 @@ pub enum DashboardEvent {
     /// The matching lifecycle action resolved (success or failure).
     /// Clients drop the spinner row.
     TransientCleared { seq: u64, name: String },
+    /// One container row changed — new container appeared (post-spawn
+    /// finalise), an existing one flipped running/needs_update/sha,
+    /// etc. Clients upsert by `container.name`. Payload carries the
+    /// full row so cold-loaded clients and event-driven clients
+    /// converge on the same render.
+    ///
+    /// Fired by `Coordinator::rescan_containers_and_emit`, which diffs
+    /// a fresh `nixos-container list`–derived snapshot against the
+    /// last one cached on the coordinator. Mutation sites (lifecycle
+    /// endpoints, actions::destroy / approve, crash_watch's poll loop)
+    /// call the rescan after their work lands.
+    ContainerStateChanged {
+        seq: u64,
+        container: ContainerView,
+    },
+    /// A container that was in the previous snapshot is gone. Clients
+    /// drop the row by name. Fired alongside any
+    /// `nixos-container destroy` (operator-driven or otherwise) on the
+    /// next rescan.
+    ContainerRemoved { seq: u64, name: String },
 }

hive-c0re/src/main.rs

@@ -11,6 +11,7 @@ mod approvals;
 mod auto_update;
 mod broker;
 mod client;
+mod container_view;
 mod coordinator;
 mod crash_watch;
 mod dashboard;