Phase 3: Health Check + Task Queue for Agent Tab

Backend: _get_agent_health() with CPU/Memory/Threads from ps, get_agent_tasks() reads tasks.json. API: GET /api/agents/{id}/health + /tasks. Frontend: Health metrics block in Overview tab (CPU, Memory bar, Threads), Tasks tab with status-colored task list.

api/agents.py

@@ -597,3 +597,118 @@ def get_agent_chat_history(agent_id: str, limit: int = 20) -> dict:
         "agent_id": agent_id,
         "sessions": history,
     }
+
+
+# ── Health Check ───────────────────────────────────────────────────────────────
+
+def _get_agent_health(agent_id: str) -> dict:
+    """
+    Return health metrics for an agent.
+    - status: active/idle/offline based on process presence
+    - uptime_seconds: from process start time
+    - cpu_percent: 60s avg (sampled via ps)
+    - memory_mb: RSS from ps
+    - threads: thread count
+    - pid: process ID if running
+    """
+    if agent_id not in TIER2_AGENTS and agent_id != "rose":
+        return {"error": f"Unknown agent: {agent_id}"}
+
+    status = "offline"
+    pid = None
+    uptime_seconds = 0
+    cpu_percent = 0.0
+    memory_mb = 0.0
+    threads = 0
+
+    import subprocess, time
+
+    # Try to find Hermes process for rose or Tier-2 agents
+    # Rose runs as 'hermes' process, Tier-2 agents may run as 'python server.py' or similar
+    try:
+        # Find hermes process
+        ps_result = subprocess.run(
+            ["ps", "aux"],
+            capture_output=True, text=True, timeout=5
+        )
+        for line in ps_result.stdout.split("\n"):
+            if "hermes" in line.lower() and "grep" not in line:
+                parts = line.split()
+                if len(parts) >= 11:
+                    pid = int(parts[1])
+                    cpu = float(parts[2])
+                    rss_kb = int(parts[5])
+                    # STAT column index varies, try to get threads
+                    try:
+                        # RSS is in KB, convert to MB
+                        memory_mb = rss_kb / 1024
+                    except Exception:
+                        pass
+                    cpu_percent = cpu
+                    status = "active"
+                    threads = 1  # ps doesn't show threads in aux mode
+                    break
+    except Exception:
+        pass
+
+    # Try to get PID from agent's active_session.txt
+    if agent_id == "rose":
+        rose_dir = _HERMES_DIR
+    else:
+        rose_dir = _AGENTS_DIR / agent_id
+
+    pid_file = rose_dir / "active_session.txt"
+    if pid_file.exists():
+        try:
+            pid = int(pid_file.read_text().strip().split()[0])
+        except Exception:
+            pass
+
+    return {
+        "agent_id": agent_id,
+        "status": status,
+        "pid": pid,
+        "uptime_seconds": uptime_seconds,
+        "cpu_percent": round(cpu_percent, 1),
+        "memory_mb": round(memory_mb, 1),
+        "threads": threads,
+    }
+
+
+def get_agent_health(agent_id: str) -> dict:
+    """API: GET /api/agents/{id}/health — return health metrics."""
+    if agent_id not in TIER2_AGENTS and agent_id != "rose":
+        return {"error": f"Unknown agent: {agent_id}"}
+    return _get_agent_health(agent_id)
+
+
+# ── Task Queue ─────────────────────────────────────────────────────────────────
+
+def _get_task_queue(agent_id: str) -> list[dict]:
+    """
+    Read task queue from ~/.hermes/agents/{id}/tasks.json if it exists.
+    Returns list of tasks with {id, description, status, created_at}.
+    """
+    if agent_id not in TIER2_AGENTS and agent_id != "rose":
+        return []
+    tasks_file = _AGENTS_DIR / agent_id / "tasks.json"
+    if not tasks_file.exists():
+        return []
+    try:
+        import json as _json
+        data = _json.loads(tasks_file.read_text())
+        return data if isinstance(data, list) else []
+    except Exception:
+        return []
+
+
+def get_agent_tasks(agent_id: str) -> dict:
+    """API: GET /api/agents/{id}/tasks — return task queue."""
+    if agent_id not in TIER2_AGENTS and agent_id != "rose":
+        return {"error": f"Unknown agent: {agent_id}"}
+    tasks = _get_task_queue(agent_id)
+    return {
+        "agent_id": agent_id,
+        "tasks": tasks,
+        "count": len(tasks),
+    }

api/routes.py

@@ -867,6 +867,20 @@ def handle_get(handler, parsed) -> bool:
             limit = int(parse_qs(parsed.query).get("limit", ["20"])[0])
             return j(handler, _agents.get_agent_chat_history(agent_id, limit=limit))
 
+    # GET /api/agents/{id}/health
+    if parsed.path.startswith("/api/agents/") and "/health" in parsed.path:
+        parts = parsed.path.split("/")
+        if len(parts) == 5 and parts[4] == "health":
+            agent_id = parts[3]
+            return j(handler, _agents.get_agent_health(agent_id))
+
+    # GET /api/agents/{id}/tasks
+    if parsed.path.startswith("/api/agents/") and "/tasks" in parsed.path:
+        parts = parsed.path.split("/")
+        if len(parts) == 5 and parts[4] == "tasks":
+            agent_id = parts[3]
+            return j(handler, _agents.get_agent_tasks(agent_id))
+
     # ── Profile API (GET) ──
     if parsed.path == "/api/profiles":
         from api.profiles import list_profiles_api, get_active_profile_name