refactor(#838): consolidate harness state files into hyperhive-harness.json

hive-c0re/src/agent_sockets.rs

@@ -4,8 +4,8 @@
 //! excluded from the map (manager UI routes via the dashboard
 //! upstream, not per-agent `/agent/<name>/`).
 //!
-//! Full mechanism — per-agent subdir bind-mount, `.bound` marker
-//! gate, gateway UDS upstream, transition vs `agent-ports.json`,
+//! Full mechanism — per-agent subdir bind-mount, `hyperhive-socket-bound`
+//! marker gate, gateway UDS upstream, transition vs `agent-ports.json`,
 //! 10s poll loop: `docs/gateway.md::Per-agent unix-socket upstream`.
 
 use std::collections::BTreeMap;
@@ -36,7 +36,13 @@ pub const SOCKET_FILENAME: &str = "web.sock";
 /// publish the unix upstream for this agent yet". Without this gate
 /// the gateway would `proxy_pass` to a non-existent socket for every
 /// sub-agent that hasn't flipped the option yet.
-pub const READY_MARKER: &str = ".bound";
+///
+/// Renamed from `.bound` (legacy) to match the `hyperhive-` prefix
+/// convention for all harness-written state files (#838). `build_map`
+/// checks both names during the transition window so existing containers
+/// don't lose gateway routing before their next rebuild.
+pub const READY_MARKER: &str = "hyperhive-socket-bound";
+const READY_MARKER_LEGACY: &str = ".bound";
 
 #[must_use]
 pub fn host_sockets_path() -> PathBuf {
@@ -72,18 +78,27 @@ pub fn socket_path_for(name: &str) -> PathBuf {
 /// `proxy_pass` to a non-existent socket for every sub-agent that
 /// hasn't yet flipped `hyperhive.web.useUnixSocket = true`.
 ///
+/// Accepts either the new `hyperhive-socket-bound` marker or the legacy
+/// `.bound` marker so existing containers keep their gateway routing
+/// through the transition window (before their next rebuild writes the
+/// new marker name).
+///
 /// `BTreeMap` keeps the JSON output sorted by key so a re-emit
 /// without churn produces byte-identical output — same idempotency
 /// shape `agent_ports::write` relies on.
 #[must_use]
 pub fn build_map(names: &[String]) -> BTreeMap<String, PathBuf> {
-    build_map_with(names, |name| ready_marker_for(name).exists())
+    build_map_with(names, |name| {
+        ready_marker_for(name).exists()
+            || agent_dir_for(name).join(READY_MARKER_LEGACY).exists()
+    })
 }
 
 /// Body of `build_map` with the ready-check parameterised. Tests
 /// pass a predicate they control (no real filesystem access).
 /// Production callers go through `build_map` which wires the
-/// predicate to the on-disk `.bound` marker check.
+/// predicate to the on-disk `hyperhive-socket-bound` (or legacy
+/// `.bound`) marker check.
 fn build_map_with<F>(names: &[String], is_ready: F) -> BTreeMap<String, PathBuf>
 where
     F: Fn(&str) -> bool,
@@ -96,7 +111,7 @@ where
         .collect()
 }
 
-/// Path to the per-agent `.bound` marker file the harness writes
+/// Path to the `hyperhive-socket-bound` marker file the harness writes
 /// after a successful `bind_unix`. Lives next to `web.sock` in the
 /// per-agent subdir so it's covered by the same bind-mount and same
 /// per-agent isolation as the socket itself.
@@ -271,10 +286,9 @@ mod tests {
 
     #[test]
     fn build_map_filters_by_ready_predicate() {
-        // The new gate: only ready agents (with `.bound` marker) get
-        // published. Pin the behaviour so a future refactor that
-        // drops the filter surfaces here, not as a 502-spew in the
-        // gateway.
+        // Only ready agents (with `hyperhive-socket-bound` marker) get
+        // published. Pin the behaviour so a future refactor that drops
+        // the filter surfaces here, not as a 502-spew in the gateway.
         let names: Vec<String> = ["iris", "argus", "atlas"]
             .iter()
             .map(|s| (*s).to_owned())
@@ -296,7 +310,7 @@ mod tests {
         let marker = ready_marker_for("iris");
         let socket = socket_path_for("iris");
         assert_eq!(marker.parent(), socket.parent());
-        assert_eq!(marker, Path::new("/run/hive-agent/iris/.bound"));
+        assert_eq!(marker, Path::new("/run/hive-agent/iris/hyperhive-socket-bound"));
     }
 
     #[test]

hive-c0re/src/container_view.rs

@@ -211,24 +211,31 @@ fn read_dashboard_links(name: &str) -> Vec<DashboardLink> {
     serde_json::from_str::<Vec<DashboardLink>>(&text).unwrap_or_default()
 }
 
-/// Returns true if the agent's harness is currently parked after an API
-/// rate-limit response. Detected via the sentinel file written by
-/// `hive_ag3nt::events::Bus::emit_status("rate_limited")`.
+/// Read `rate_limited` + `needs_login` from the consolidated
+/// `hyperhive-harness.json`. Falls back to the legacy individual
+/// sentinel files written by older harness builds so in-place upgrades
+/// don't lose state during the transition window.
+fn read_harness_flags(name: &str) -> (bool, bool) {
+    let dir = Coordinator::agent_notes_dir(name);
+    if let Ok(raw) = std::fs::read_to_string(dir.join("hyperhive-harness.json")) {
+        if let Ok(v) = serde_json::from_str::<serde_json::Value>(&raw) {
+            let rl = v.get("rate_limited").and_then(|x| x.as_bool()).unwrap_or(false);
+            let nl = v.get("needs_login").and_then(|x| x.as_bool()).unwrap_or(false);
+            return (rl, nl);
+        }
+    }
+    // Legacy fallback: presence of individual sentinel files.
+    let rate_limited = dir.join("hyperhive-rate-limited").exists();
+    let needs_login = dir.join("hyperhive-needs-login").exists();
+    (rate_limited, needs_login)
+}
+
 fn is_rate_limited(name: &str) -> bool {
-    Coordinator::agent_notes_dir(name)
-        .join("hyperhive-rate-limited")
-        .exists()
+    read_harness_flags(name).0
 }
 
-/// True when the harness wrote `{state_dir}/hyperhive-needs-login`
-/// after a 401 mid-turn. Lets the dashboard surface `needs_login` for
-/// agents whose `/root/.claude/` dir still exists (so
-/// `claude_has_session` returns true) but whose OAuth credentials
-/// inside it have actually expired.
 fn auth_failed_sentinel(name: &str) -> bool {
-    Coordinator::agent_notes_dir(name)
-        .join("hyperhive-needs-login")
-        .exists()
+    read_harness_flags(name).1
 }
 
 /// Read the agent's free-text status and the Unix timestamp when it was last set