fix: bind api_server to 127.0.0.1 + fix watchdog process detection

Root cause: api_server refused to bind 0.0.0.0 without API_SERVER_KEY,
so the /health endpoint on :8642 was never available. But the Gateway
process WAS running and connected to Feishu WebSocket.

Fix 1: Change api_server host from 0.0.0.0 to 127.0.0.1 (only local
access needed since entry.py proxy handles external requests)

Fix 2: Watchdog now checks process liveness (kill -0 PID) instead of
HTTP health check, preventing false restart of a running gateway

config.yaml

@@ -13,7 +13,7 @@ platforms:
   api_server:
     enabled: true
     extra:
-      host: 0.0.0.0
+      host: 127.0.0.1
       port: 8642
       cors_origins: "*"
 memory:

start.sh

@@ -129,31 +129,38 @@ echo ""
 
 # ── Start Gateway watchdog in background ──
 (
-    echo "[$(date)] Gateway watchdog started (check interval: 30s)"
+    echo "[$(date)] Gateway watchdog started (check interval: 60s)"
     while true; do
-        sleep 30
-        # Check if gateway health endpoint responds
-        if ! curl -s --max-time 5 http://127.0.0.1:8642/health > /dev/null 2>&1; then
-            echo "[$(date)] Gateway health check FAILED — restarting..."
-            # Kill stale process if any
-            if [ -f /tmp/hermes-gateway.pid ]; then
-                OLD_PID=$(cat /tmp/hermes-gateway.pid)
-                kill -9 "$OLD_PID" 2>/dev/null || true
+        sleep 60
+        # Check if gateway process is still alive
+        GW_ALIVE=false
+        if [ -f /tmp/hermes-gateway.pid ]; then
+            if kill -0 "$(cat /tmp/hermes-gateway.pid)" 2>/dev/null; then
+                GW_ALIVE=true
+            fi
+        fi
+        # Also check via gateway.pid in hermes home
+        if [ "$GW_ALIVE" = false ] && [ -f "$HERMES_HOME/gateway.pid" ]; then
+            if kill -0 "$(cat "$HERMES_HOME/gateway.pid")" 2>/dev/null; then
+                GW_ALIVE=true
+                # Sync the PID file
+                cp "$HERMES_HOME/gateway.pid" /tmp/hermes-gateway.pid
             fi
-            # Append diagnostic info to log
-            echo "--- Gateway watchdog restart $(date) ---" >> /data/hermes/logs/gateway.log
+        fi
+
+        if [ "$GW_ALIVE" = false ]; then
+            echo "[$(date)] Gateway process is DEAD — restarting..."
             # Clean stale state
             rm -f "$HERMES_HOME/gateway.pid" 2>/dev/null
             rm -f "$HERMES_HOME/.gateway_runtime_lock" 2>/dev/null
             rm -f "$HERMES_HOME/.gateway_takeover" 2>/dev/null
             # Restart
             start_gateway
-            # Brief wait then verify
             sleep 15
-            if curl -s --max-time 5 http://127.0.0.1:8642/health > /dev/null 2>&1; then
+            if [ -f /tmp/hermes-gateway.pid ] && kill -0 "$(cat /tmp/hermes-gateway.pid)" 2>/dev/null; then
                 echo "[$(date)] Gateway restarted successfully"
             else
-                echo "[$(date)] Gateway restart failed — will retry in 30s"
+                echo "[$(date)] Gateway restart failed — will retry in 60s"
             fi
         fi
     done