huskies: merge 1149 story huskies health chat command — surface gateway, sled, matrix, creds, and build-hash status

server/src/chat/transport/matrix/bot/context.rs

@@ -6,6 +6,7 @@ use crate::services::Services;
 use matrix_sdk::ruma::{OwnedEventId, OwnedRoomId, OwnedUserId};
 use std::collections::{BTreeMap, HashSet, VecDeque};
 use std::sync::Arc;
+use std::sync::atomic::AtomicI64;
 use tokio::sync::Mutex as TokioMutex;
 use tokio::sync::RwLock;
 
@@ -104,6 +105,10 @@ pub struct BotContext {
     /// Used by the "rebuild gateway" command to construct the health-check URL
     /// passed to the trampoline.  `None` in standalone single-project mode.
     pub gateway_port: Option<u16>,
+    /// Timestamp (ms since Unix epoch) of the last Matrix event received in any
+    /// configured room.  Updated atomically on every `on_room_message` call so
+    /// the `health` command can detect a stale or dead sync loop.
+    pub last_matrix_event_ms: Arc<AtomicI64>,
 }
 
 impl BotContext {
@@ -299,6 +304,7 @@ mod tests {
                 SEEN_EVENT_IDS_CAP,
             ))),
             gateway_port: None,
+            last_matrix_event_ms: Arc::new(AtomicI64::new(chrono::Utc::now().timestamp_millis())),
         }
     }
 

server/src/chat/transport/matrix/bot/messages/on_room_message.rs

@@ -19,6 +19,20 @@ use super::super::verification::check_sender_verified;
 
 use super::handle_message;
 
+/// Return `true` when the message is a `health` command addressed to the bot.
+///
+/// Recognised case-insensitively as the single word `health` after stripping the bot
+/// mention prefix.  Any trailing whitespace is ignored; extra arguments are not
+/// expected and are silently discarded.
+fn extract_health_command(message: &str, bot_name: &str, bot_user_id: &str) -> bool {
+    let stripped = crate::chat::util::strip_bot_mention(message, bot_name, bot_user_id);
+    let trimmed = stripped
+        .trim()
+        .trim_start_matches(|c: char| !c.is_alphanumeric());
+    let cmd = trimmed.split_whitespace().next().unwrap_or("");
+    cmd.eq_ignore_ascii_case("health")
+}
+
 /// Return `true` when the message is a "rebuild gateway" command addressed to the bot.
 ///
 /// The command is recognised case-insensitively as `rebuild gateway` after stripping
@@ -100,6 +114,12 @@ pub(in crate::chat::transport::matrix::bot) async fn on_room_message(
         return;
     }
 
+    // Update last-event timestamp so the `health` command can detect a stale sync loop.
+    ctx.last_matrix_event_ms.store(
+        chrono::Utc::now().timestamp_millis(),
+        std::sync::atomic::Ordering::Relaxed,
+    );
+
     // Ignore the bot's own messages to prevent echo loops.
     if ev.sender == ctx.matrix_user_id {
         return;
@@ -249,6 +269,7 @@ pub(in crate::chat::transport::matrix::bot) async fn on_room_message(
             "config",
             "project-rebuild",
             "upgrade",
+            "health",
         ];
 
         let stripped = crate::chat::util::strip_bot_mention(
@@ -546,6 +567,26 @@ pub(in crate::chat::transport::matrix::bot) async fn on_room_message(
         return;
     }
 
+    // `health` — async subsystem health report (gateway + standalone).
+    if extract_health_command(
+        &user_message,
+        &ctx.services.bot_name,
+        ctx.matrix_user_id.as_str(),
+    ) {
+        slog!("[matrix-bot] Handling 'health' from {sender}");
+        let response = super::super::super::health::run_health_check(&ctx).await;
+        let html = markdown_to_html(&response);
+        if let Ok(msg_id) = ctx
+            .transport
+            .send_message(&room_id_str, &response, &html)
+            .await
+            && let Ok(event_id) = msg_id.parse()
+        {
+            ctx.bot_sent_event_ids.lock().await.insert(event_id);
+        }
+        return;
+    }
+
     // Check for bot-level commands (help, status, ambient, …) before invoking
     // the LLM.  All commands are registered in commands.rs — no special-casing
     // needed here.

server/src/chat/transport/matrix/bot/run.rs

@@ -6,7 +6,7 @@ use matrix_sdk::ruma::OwnedRoomId;
 use matrix_sdk::{Client, LoopCtrl, config::SyncSettings};
 use std::collections::{HashMap, HashSet};
 use std::sync::Arc;
-use std::sync::atomic::{AtomicBool, AtomicU64, Ordering};
+use std::sync::atomic::{AtomicBool, AtomicI64, AtomicU64, Ordering};
 use tokio::sync::Mutex as TokioMutex;
 use tokio::sync::{RwLock, watch};
 
@@ -336,6 +336,7 @@ pub async fn run_bot(
             super::context::SEEN_EVENT_IDS_CAP,
         ))),
         gateway_port,
+        last_matrix_event_ms: Arc::new(AtomicI64::new(chrono::Utc::now().timestamp_millis())),
     };
 
     slog!(