Add SSE second-level status push for control monitor.

Co-authored-by: Cursor <cursoragent@cursor.com>

control/backend/app/api/nodes.py

@@ -1,9 +1,13 @@
 from __future__ import annotations
 
+import asyncio
+import json
 import secrets
-from typing import Annotated, Any
+import time
+from typing import Annotated, Any, AsyncIterator
 
-from fastapi import APIRouter, Depends, HTTPException
+from fastapi import APIRouter, Depends, HTTPException, Request
+from fastapi.responses import StreamingResponse
 from pydantic import BaseModel, Field
 
 from ..auth import require_control_user
@@ -47,6 +51,50 @@ def _raise_node_error(code: int, data: Any) -> None:
     )
 
 
+def _health_looks_like_fleet(health: Any) -> bool:
+    return isinstance(health, dict) and "strategy" in health and "mode" in health
+
+
+async def _collect_one_status(
+    node: dict[str, Any],
+    *,
+    timeout: float | None = None,
+) -> dict[str, Any]:
+    probe = await probe_health(node, timeout=timeout)
+    item: dict[str, Any] = {**_public_node(node), **probe}
+    if probe.get("online") and node.get("token_sealed"):
+        health = probe.get("health")
+        # probe_health 在 /health 失败时可能已用 fleet/status 回填到 health
+        if _health_looks_like_fleet(health):
+            code, data = 200, health
+        else:
+            code, data = await call_node(
+                node, "GET", "/api/fleet/status", timeout=timeout
+            )
+        if code == 200 and isinstance(data, dict):
+            item["fleet"] = data
+            item["fleet_ok"] = True
+        else:
+            item["fleet_ok"] = False
+            item["fleet_error"] = _http_detail(data) if data else f"HTTP {code}"
+    elif not node.get("token_sealed"):
+        item["fleet_ok"] = False
+        item["fleet_error"] = "未生成 Token"
+    return item
+
+
+async def collect_all_status(*, timeout: float | None = None) -> list[dict[str, Any]]:
+    db = get_control_db()
+    nodes = db.list_nodes()
+    if not nodes:
+        return []
+    return list(
+        await asyncio.gather(
+            *[_collect_one_status(n, timeout=timeout) for n in nodes]
+        )
+    )
+
+
 class NodeCreate(BaseModel):
     name: str = Field(min_length=1, max_length=64)
     base_url: str = Field(min_length=8, max_length=256)
@@ -79,26 +127,65 @@ async def create_node(
 
 @router.get("/status/all")
 async def status_all(_user: Annotated[str, Depends(require_control_user)]) -> dict:
-    db = get_control_db()
-    items = []
-    for node in db.list_nodes():
-        probe = await probe_health(node)
-        item = {**_public_node(node), **probe}
-        if probe.get("online") and node.get("token_sealed"):
-            code, data = await call_node(node, "GET", "/api/fleet/status")
-            if code == 200 and isinstance(data, dict):
-                item["fleet"] = data
-                item["fleet_ok"] = True
-            else:
-                item["fleet_ok"] = False
-                item["fleet_error"] = _http_detail(data) if data else f"HTTP {code}"
-        elif not node.get("token_sealed"):
-            item["fleet_ok"] = False
-            item["fleet_error"] = "未生成 Token"
-        items.append(item)
+    items = await collect_all_status()
     return {"nodes": items}
 
 
+@router.get("/status/stream")
+async def status_stream(
+    request: Request,
+    _user: Annotated[str, Depends(require_control_user)],
+) -> StreamingResponse:
+    """SSE：中控约每秒并行拉取策略机状态并推送到浏览器。"""
+    settings = get_control_settings()
+    interval = max(0.5, float(settings.control_sse_interval_sec))
+    probe_timeout = max(0.5, float(settings.control_sse_probe_timeout_sec))
+
+    async def event_gen() -> AsyncIterator[str]:
+        last_payload = ""
+        last_heartbeat = 0.0
+        yield f": connected interval={interval}\n\n"
+        while True:
+            if await request.is_disconnected():
+                break
+            started = time.monotonic()
+            try:
+                items = await collect_all_status(timeout=probe_timeout)
+                payload = json.dumps(
+                    {"nodes": items, "ts_ms": int(time.time() * 1000)},
+                    ensure_ascii=False,
+                    separators=(",", ":"),
+                    default=str,
+                )
+                now = time.monotonic()
+                if payload != last_payload:
+                    last_payload = payload
+                    yield f"event: nodes\ndata: {payload}\n\n"
+                    last_heartbeat = now
+                elif now - last_heartbeat >= 5.0:
+                    yield f": heartbeat {int(time.time())}\n\n"
+                    last_heartbeat = now
+            except Exception as e:
+                err = json.dumps(
+                    {"detail": str(e)},
+                    ensure_ascii=False,
+                    separators=(",", ":"),
+                )
+                yield f"event: error\ndata: {err}\n\n"
+            elapsed = time.monotonic() - started
+            await asyncio.sleep(max(0.05, interval - elapsed))
+
+    return StreamingResponse(
+        event_gen(),
+        media_type="text/event-stream",
+        headers={
+            "Cache-Control": "no-cache",
+            "Connection": "keep-alive",
+            "X-Accel-Buffering": "no",
+        },
+    )
+
+
 @router.post("/update-batch")
 async def update_batch(
     body: dict,
@@ -179,13 +266,7 @@ async def node_status(
     node = db.get_node(node_id)
     if not node:
         raise HTTPException(status_code=404, detail="节点不存在")
-    probe = await probe_health(node)
-    out = {**_public_node(node), **probe}
-    if probe.get("online") and node.get("token_sealed"):
-        code, data = await call_node(node, "GET", "/api/fleet/status")
-        if code == 200 and isinstance(data, dict):
-            out["fleet"] = data
-    return out
+    return await _collect_one_status(node)
 
 
 @router.post("/{node_id}/start")