Update app.py

app.py

@@ -1,12 +1,13 @@
 """
-Eroha v6.4.4 — Zero-Daemon Edition
-----------------------------------
-Production-grade FastAPI + Gradio fusion
-No threads, no leaks, 100% graceful lifecycle.
+Eroha v6.5 — Hybrid Enterprise Edition
+--------------------------------------
+Features: Async Lifespan, Prometheus Metrics, Proxy-Aware Limiter, 
+Stable Health Monitor, and Gradio Dashboard.
 """
 
 import asyncio
 import psutil
+import httpx
 import gradio as gr
 from fastapi import FastAPI, Request
 from slowapi import Limiter, _rate_limit_exceeded_handler
@@ -14,149 +15,134 @@ from slowapi.util import get_remote_address
 from slowapi.errors import RateLimitExceeded
 from gradio.routes import mount_gradio_app
 from contextlib import asynccontextmanager
-
+from prometheus_client import make_asgi_app, Counter, Gauge
 
 # ───────────────────────────────
-# 1️⃣ Global metrics state (event-loop safe)
+# 1️⃣ Enterprise Metrics & State
 # ───────────────────────────────
+# Метрики для внешних систем (Prometheus/Grafana)
+REQ_COUNT = Counter("api_requests_total", "Total requests", ["method", "endpoint"])
+SYS_USAGE = Gauge("system_usage_percent", "System metrics", ["resource"])
+
+# Локальный стейт для Gradio UI
 state = {"cpu": 0.0, "ram": 0.0, "timestamp": 0.0}
 
+# ───────────────────────────────
+# 2️⃣ Proxy-Aware Real IP Limiter
+# ───────────────────────────────
+def get_real_ip(request: Request):
+    """Извлекает реальный IP пользователя за прокси (Hugging Face/Nginx)."""
+    forwarded = request.headers.get("x-forwarded-for")
+    if forwarded:
+        return forwarded.split(",")[0]
+    return request.client.host
+
+limiter = Limiter(key_func=get_real_ip)
 
 # ───────────────────────────────
-# 2️⃣ Lifespan manager (async background task)
+# 3️⃣ Lifespan Manager (Resource Control)
 # ───────────────────────────────
 @asynccontextmanager
 async def lifespan(app: FastAPI):
     stop_event = asyncio.Event()
 
     async def background_metrics():
+        """Фоновый цикл сбора данных (без потоков)."""
         while not stop_event.is_set():
             try:
-                state["cpu"] = psutil.cpu_percent()
-                state["ram"] = psutil.virtual_memory().percent
+                cpu = psutil.cpu_percent()
+                ram = psutil.virtual_memory().percent
+                
+                # Обновление для UI
+                state["cpu"], state["ram"] = cpu, ram
                 state["timestamp"] = asyncio.get_event_loop().time()
+                
+                # Обновление для Prometheus
+                SYS_USAGE.labels(resource="cpu").set(cpu)
+                SYS_USAGE.labels(resource="ram").set(ram)
+                
             except Exception as e:
-                print(f"[Metrics] error: {e}")
-            await asyncio.sleep(5)
+                print(f"[Metrics Error] {e}")
+            await asyncio.sleep(15)  # Интервал для Production
 
     task = asyncio.create_task(background_metrics())
-    yield  # Server runs here
+    yield  # Здесь работает приложение
     stop_event.set()
     await asyncio.gather(task, return_exceptions=True)
 
-
 # ───────────────────────────────
-# 3️⃣ FastAPI app with rate limiter
+# 4️⃣ FastAPI Core Setup
 # ───────────────────────────────
-app = FastAPI(title="Eroha v6.4.4 API", lifespan=lifespan)
-limiter = Limiter(key_func=get_remote_address)
+app = FastAPI(title="Eroha v6.5 Enterprise", lifespan=lifespan)
 app.state.limiter = limiter
 app.add_exception_handler(RateLimitExceeded, _rate_limit_exceeded_handler)
 
+# Монтируем эндпоинт для Prometheus
+metrics_app = make_asgi_app()
+app.mount("/metrics/prom", metrics_app)
 
 @app.get("/health")
 async def health():
     return {"status": "ok", "uptime": state["timestamp"]}
 
-
 @app.get("/metrics")
-async def metrics():
+async def get_json_metrics():
+    """Для обратной совместимости с простыми чекерами."""
     return state
 
-
 @app.post("/inference")
 @limiter.limit("10/minute")
 async def inference(request: Request):
+    REQ_COUNT.labels(method="POST", endpoint="/inference").inc()
     data = await request.json()
     prompt = data.get("prompt", "")
-    # simulate model call
-    await asyncio.sleep(0.1)
+    await asyncio.sleep(0.1)  # Имитация работы модели
     return {"reply": f"Echo: {prompt[:120]}", "stats": state}
 
-
 # ─────────���─────────────────────
-# 4️⃣ Gradio dashboard
+# 5️⃣ Gradio Interface (UI)
 # ───────────────────────────────
-with gr.Blocks(title="Eroha v6.4.4 Dashboard") as demo:
-    gr.Markdown("## ⚙️ Eroha v6.4.4 – Zero-Daemon Edition")
-
-    with gr.Row():
-        inp = gr.Textbox(label="Prompt")
-        out = gr.Textbox(label="Response")
-        gr.Button("Send").click(lambda x: f"Echo: {x}", inputs=inp, outputs=out)
-
-    gr.Markdown("### 📊 Live Metrics (5s refresh)")
-    cpu_box = gr.Number(label="CPU %")
-    ram_box = gr.Number(label="RAM %")
-
-    demo.load(lambda: (state["cpu"], state["ram"]), outputs=[cpu_box, ram_box], every=5)
-
-import httpx
-
-# 4️⃣b Health Status Checker
-async def check_health():
-    """Проверка состояния API."""
+async def check_health_ui():
+    """Проверка доступности API через петлю (Self-health check)."""
     try:
-        async with httpx.AsyncClient(timeout=2) as client:
+        async with httpx.AsyncClient(timeout=1) as client:
             resp = await client.get("http://localhost:7860/health")
-            if resp.status_code == 200 and resp.json().get("status") == "ok":
-                return "🟢 ONLINE"
-    except Exception:
+            if resp.status_code == 200:
+                return "<div style='color:lime;font-size:18px;'>🟢 API ONLINE</div>"
+    except:
         pass
-    return "🔴 OFFLINE"
-
-# Добавляем в интерфейс Gradio
-with gr.Blocks(title="Eroha v6.4.5 Dashboard") as demo:
-    gr.Markdown("## ⚙️ Eroha v6.4.5 – Stable Health Edition")
+    return "<div style='color:red;font-size:18px;'>🔴 API OFFLINE</div>"
 
+with gr.Blocks(title="Eroha v6.5 Dashboard", theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# ⚙️ Eroha v6.5 — Hybrid Enterprise")
+    
     with gr.Row():
-        with gr.Column():
-            inp = gr.Textbox(label="Prompt")
-            btn = gr.Button("Send")
-        out = gr.Textbox(label="Response")
+        health_status = gr.HTML("<div style='font-size:18px;'>🟡 Checking...</div>")
     
-    btn.click(fn=lambda x: f"Echo: {x[:200]}", inputs=inp, outputs=out)
-
-# 🩺 System Health Monitor
-gr.Markdown("### 🩺 System Health Monitor")
-
-with gr.Row():
-    with gr.Column(scale=1):
-        health_indicator = gr.HTML("<div style='font-size:20px;'>🟡 Checking...</div>")
-
-# Асинхронная функция проверки состояния API
-async def check_health_ui():
-    try:
-        async with httpx.AsyncClient(timeout=2) as client:
-            resp = await client.get("http://localhost:7860/health")
-            if resp.status_code == 200 and resp.json().get("status") == "ok":
-                return "<div style='color:lime;font-size:20px;'>🟢 ONLINE</div>"
-    except Exception:
-        pass
-    return "<div style='color:red;font-size:20px;'>🔴 OFFLINE</div>"
-
-# Автообновление статуса API каждые 5 секунд
-demo.load(check_health_ui, outputs=[health_indicator], every=5)
-
-# 📊 Live Metrics (5s refresh)
-gr.Markdown("### 📊 Live Metrics (5s refresh)")
-cpu_box = gr.Number(label="CPU %")
-ram_box = gr.Number(label="RAM %")
-
-# Автообновление метрик
-demo.load(lambda: (state["cpu"], state["ram"]), outputs=[cpu_box, ram_box], every=5)
+    with gr.Tabs():
+        with gr.TabItem("Inference"):
+            with gr.Row():
+                inp = gr.Textbox(label="Input Prompt", placeholder="Type here...")
+                out = gr.Textbox(label="Model Response")
+            btn = gr.Button("Run Inference", variant="primary")
+            btn.click(lambda x: f"Processed: {x}", inputs=inp, outputs=out)
+            
+        with gr.TabItem("System Monitor"):
+            gr.Markdown("### 📊 Real-time Resource Usage")
+            with gr.Row():
+                cpu_box = gr.Number(label="CPU %")
+                ram_box = gr.Number(label="RAM %")
+            gr.Markdown("> Metrics are also exported to `/metrics/prom` for Prometheus.")
+
+    # Авто-обновление UI каждые 5 секунд
+    demo.load(check_health_ui, outputs=[health_status], every=5)
+    demo.load(lambda: (state["cpu"], state["ram"]), outputs=[cpu_box, ram_box], every=5)
 
 # ───────────────────────────────
-# 5️⃣ Mount Gradio to FastAPI (single port)
+# 6️⃣ Mounting & Launch
 # ───────────────────────────────
 app = mount_gradio_app(app, demo, path="/")
 
-# No explicit uvicorn.run — HF Spaces handles launch automatically.
-
-# ───────────────────────────────
-# 6️⃣ Entrypoint для Hugging Face & Local
-# ───────────────────────────────
 if __name__ == "__main__":
     import uvicorn
-    # Важно: ��спользуем строку "app:app" для поддержки reload и корректной работы lifespan
     uvicorn.run(app, host="0.0.0.0", port=7860)