Update app.py

app.py

@@ -1,254 +1,109 @@
 import os
 import time
-import threading
-import requests
+import json
 import gradio as gr
-import pandas as pd
-import matplotlib.pyplot as plt
-import base64
-from io import BytesIO
+import requests
 from datetime import datetime
 
-from core.dashboard import ErohaDashboard
-from core.intelligence import update_memory, summarize_context
-from core.selfcheck import evaluate_answer, improve_answer
-from core.learning import analyze_user_input, adapt_answer
-from core.model_selector import choose_model
-
-
-# ==============================
-# 🔐 ENVIRONMENT CONFIG
-# ==============================
-HF_TOKEN = os.getenv("HF_TOKEN", "")
-LOG_FILE = "logs/history.csv"
-os.makedirs("logs", exist_ok=True)
-dashboard = ErohaDashboard()
-
-
-# ==============================
-# 🧠 Router Manager
-# ==============================
-class RouterManager:
-    def __init__(self, token: str):
-        self.token = token
-        self.router_url = "https://router.huggingface.co"
-        self.headers = {"Authorization": f"Bearer {self.token}"}
-        self.is_router_healthy = True
-        self.last_check = None
-        self.reconnect_attempts = 0
-        self.last_token_validation = 0
-
-    def validate_token(self):
-        """Проверяет, действителен ли токен"""
-        try:
-            r = requests.get(self.router_url + "/status", headers=self.headers, timeout=10)
-            if r.status_code == 200:
-                return True
-            elif r.status_code == 401:
-                print("🔒 Ошибка: недействительный токен Hugging Face!")
-                self.is_router_healthy = False
-                return False
-            return False
-        except Exception:
+# =========================
+# ⚙️ НАСТРОЙКИ
+# =========================
+HF_TOKEN = os.getenv("HF_TOKEN") or "hf_your_token_here"
+
+PRIMARY_MODEL = "google/gemma-2-2b-it"
+FALLBACK_MODEL = "mistralai/Mixtral-8x7B-Instruct-v0.1"
+ROUTER_URL = "https://router.huggingface.co"
+
+HEADERS = {"Authorization": f"Bearer {HF_TOKEN}"}
+
+
+# =========================
+# 🧩 ПОЛЕЗНЫЕ ФУНКЦИИ
+# =========================
+
+def check_token():
+    """Проверка валидности токена Hugging Face."""
+    try:
+        res = requests.get("https://router.huggingface.co/status", headers=HEADERS, timeout=8)
+        if res.status_code == 200:
+            return True
+        else:
+            print(f"⚠️ Токен Hugging Face невалиден ({res.status_code})")
             return False
+    except Exception as e:
+        print(f"Ошибка при проверке токена: {e}")
+        return False
 
-    def check_router_health(self):
-        """Проверяет доступность Router API"""
-        try:
-            start = time.time()
-            r = requests.post(self.router_url, headers=self.headers, json={"inputs": "ping"})
-            latency = round(time.time() - start, 2)
-            if r.status_code == 200:
-                self.is_router_healthy = True
-                self.last_check = datetime.now()
-                self.reconnect_attempts = 0
-                print(f"✅ Router API активен ({latency}s)")
-            else:
-                self.is_router_healthy = False
-                print(f"⚠️ Router API ответил {r.status_code}")
-        except Exception as e:
-            self.is_router_healthy = False
-            print("❌ Router недоступен:", str(e))
-
-    def auto_reconnect(self):
-        """Автоматическое восстановление при потере связи"""
-        self.reconnect_attempts += 1
-        delay = min(30, 2 ** self.reconnect_attempts)
-        print(f"🔁 Попытка переподключения через {delay}s...")
-        time.sleep(delay)
-        self.check_router_health()
-
-    def request(self, model_id: str, prompt: str):
-        """Отправляет запрос с failover и auto-heal"""
-        payload = {
-            "model": model_id,
-            "inputs": prompt,
-            "parameters": {"max_new_tokens": 600, "temperature": 0.7},
-        }
-
-        try:
-            response = requests.post(
-                self.router_url,
-                headers=self.headers,
-                json=payload,
-                timeout=60,
-            )
-            if response.status_code == 200:
-                data = response.json()
-                if isinstance(data, list) and "generated_text" in data[0]:
-                    return data[0]["generated_text"]
-                elif isinstance(data, dict) and "generated_text" in data:
-                    return data["generated_text"]
-                else:
-                    raise RuntimeError("Неверный формат ответа")
-            elif response.status_code >= 500:
-                print("⚠️ Ошибка Router 5xx. Переподключение...")
-                self.auto_reconnect()
-                return self.request(model_id, prompt)
-            else:
-                raise RuntimeError(f"Router ошибка: {response.status_code}")
-
-        except Exception as e:
-            print(f"❌ Router ошибка: {e}, активируем fallback...")
-            return self.fallback_request(prompt)
-
-    def fallback_request(self, prompt):
-        """Fallback на стабильную модель"""
-        fallback_model = "mistralai/Mixtral-8x7B-Instruct-v0.1"
-        try:
-            response = requests.post(
-                self.router_url,
-                headers=self.headers,
-                json={"model": fallback_model, "inputs": prompt},
-                timeout=60,
-            )
-            if response.status_code == 200:
-                data = response.json()
-                if isinstance(data, list) and "generated_text" in data[0]:
-                    return f"[⚠️ Fallback {fallback_model}] {data[0]['generated_text']}"
-            raise RuntimeError(f"Fallback ошибка: {response.status_code}")
-        except Exception as e:
-            return f"❌ Ошибка fallback: {e}"
-
-
-# ==============================
-# 🐕 Watchdog Thread
-# ==============================
-def watchdog_loop(router_manager: RouterManager):
-    """Поток, который следит за Router API и токеном"""
-    while True:
-        try:
-            router_manager.check_router_health()
-            now = time.time()
-            # Проверка токена каждые 10 минут
-            if now - router_manager.last_token_validation > 600:
-                router_manager.validate_token()
-                router_manager.last_token_validation = now
-            time.sleep(60)
-        except Exception as e:
-            print("⚠️ Watchdog ошибка:", str(e))
-            time.sleep(30)
-
-
-# ==============================
-# 🧠 Основная генерация
-# ==============================
-router_manager = RouterManager(HF_TOKEN)
-
-
-def generate_response(user_input):
+
+def send_request(model: str, prompt: str):
+    """Отправка запроса к модели через Router API."""
+    payload = {"model": model, "inputs": prompt, "options": {"use_cache": True}}
     try:
         start = time.time()
-        prefs = analyze_user_input(user_input)
-        model_id = choose_model(user_input)
-
-        result = router_manager.request(model_id, user_input)
-        check = evaluate_answer(result)
-        improved = improve_answer(result)
-        personalized = adapt_answer(improved)
-        update_memory(user_input, personalized)
-        context = summarize_context()
-
-        response_time = round(time.time() - start, 2)
-        dashboard.log_request(model_id, prefs["category"], response_time)
-
-        df = pd.DataFrame([{
-            "time": time.strftime("%Y-%m-%d %H:%M:%S"),
-            "model": model_id,
-            "category": prefs["category"],
-            "response_time": response_time,
-            "prompt": user_input,
-            "response": personalized[:1500],
-        }])
-        df.to_csv(LOG_FILE, mode="a", index=False, header=not os.path.exists(LOG_FILE))
-
-        summary = (
-            f"🧠 **Модель:** `{model_id}`\n"
-            f"⚡ **Время отклика:** {response_time} сек\n"
-            f"🔍 **Проверка качества:** {check['result']}\n"
-            f"{'; '.join(check['feedback']) if check['feedback'] else '✅ Всё отлично'}\n\n"
-            f"{context}"
-        )
-
-        return f"{personalized}\n\n{summary}"
-
+        response = requests.post(ROUTER_URL, headers=HEADERS, json=payload, timeout=60)
+        latency = time.time() - start
+
+        if response.status_code == 200:
+            data = response.json()
+            # Универсальный парсинг ответа
+            if isinstance(data, list) and len(data) > 0 and "generated_text" in data[0]:
+                text = data[0]["generated_text"]
+            elif isinstance(data, dict) and "generated_text" in data:
+                text = data["generated_text"]
+            else:
+                text = str(data)
+            return text.strip(), latency, model, None
+        else:
+            return None, latency, model, f"Ошибка API {response.status_code}: {response.text}"
     except Exception as e:
-        return f"❌ Ошибка выполнения: {str(e)}"
-
-
-# ==============================
-# 📈 Аналитика
-# ==============================
-def generate_chart():
-    if not os.path.exists(LOG_FILE):
-        return "⚠️ Нет данных для графика."
-    df = pd.read_csv(LOG_FILE)
-    if df.empty:
-        return "⚠️ История пуста."
-    df["time"] = pd.to_datetime(df["time"])
-    df = df.tail(50)
-
-    plt.figure(figsize=(8, 4))
-    plt.plot(df["time"], df["response_time"], marker="o", linewidth=1.5)
-    plt.title("⚡ Скорость отклика моделей (последние 50 запросов)")
-    plt.xlabel("Время")
-    plt.ylabel("Секунды")
-    plt.grid(True)
-    buffer = BytesIO()
-    plt.savefig(buffer, format="png", bbox_inches="tight")
-    buffer.seek(0)
-    img_base64 = base64.b64encode(buffer.read()).decode("utf-8")
-    plt.close()
-    return f"<img src='data:image/png;base64,{img_base64}'/>"
-
-
-def show_dashboard():
-    metrics_text, df = dashboard.dashboard_ui()
-    chart_html = generate_chart()
-    return metrics_text, df, chart_html
-
-
-# ==============================
-# 🎨 Gradio UI
-# ==============================
-with gr.Blocks(title="Eroha AgentAPI v5.8 — Production-Cluster Core", theme="soft") as app:
-    gr.Markdown("# 🤖 Eroha AgentAPI v5.8 — Production-Cluster Core 🚀")
-    gr.Markdown("**Полный контроль: Router-watchdog + auto-heal + token-validation + fallback.**")
-
-    with gr.Tab("💬 Agent Chat"):
-        user_input = gr.Textbox(label="Введите запрос", placeholder="Например: придумай легенду о ИИ, который нашёл сознание.")
-        output_box = gr.Textbox(label="Ответ", lines=15)
-        submit_btn = gr.Button("🚀 Отправить")
-        submit_btn.click(fn=generate_response, inputs=user_input, outputs=output_box)
-
-    with gr.Tab("📊 Dashboard"):
-        metrics = gr.Markdown(label="📈 Общая статистика")
-        log_table = gr.Dataframe(headers=["time", "model", "category", "response_time"], label="История запросов")
-        chart_box = gr.HTML()
-        refresh = gr.Button("🔄 Обновить дашборд")
-        refresh.click(show_dashboard, outputs=[metrics, log_table, chart_box])
-
-# 🔥 Запуск watchdog-потока
-threading.Thread(target=watchdog_loop, args=(router_manager,), daemon=True).start()
-
-app.launch(server_name="0.0.0.0", server_port=7860)
+        return None, 0, model, str(e)
+
+
+def generate_text(prompt: str):
+    """Основная функция: попытка через основную модель → fallback при ошибке."""
+    if not check_token():
+        return "❌ Токен Hugging Face недействителен. Проверьте переменную HF_TOKEN."
+
+    # 1️⃣ Основная модель
+    output, latency, used_model, error = send_request(PRIMARY_MODEL, prompt)
+    if output:
+        return render_output(output, used_model, latency, success=True)
+    
+    # 2️⃣ Fallback при ошибке
+    output_fb, latency_fb, model_fb, error_fb = send_request(FALLBACK_MODEL, prompt)
+    if output_fb:
+        return render_output(output_fb, model_fb, latency_fb, success=True, fallback=True)
+    else:
+        return f"❌ Ошибка при выполнении запроса:\n- {error}\n- Fallback: {error_fb}"
+
+
+def render_output(text, model, latency, success=False, fallback=False):
+    """Форматированный вывод результата."""
+    emoji = "✅" if success else "⚠️"
+    fb_text = " (через fallback)" if fallback else ""
+    return (
+        f"{emoji} **Модель:** `{model}`{fb_text}\n"
+        f"⏱ **Время отклика:** {latency:.2f} сек\n\n"
+        f"🧠 **Ответ:**\n{text.strip()}"
+    )
+
+
+# =========================
+# 🧭 GRADIO UI
+# =========================
+
+with gr.Blocks(title="🤖 Eroha AgentAPI v5.7 — Stable Router Edition") as demo:
+    gr.Markdown("## 🧠 Eroha AgentAPI v5.7 — Stable Router Edition\n"
+                "Поддержка Router API + AutoFallback + Token Validation 🌐")
+
+    with gr.Row():
+        prompt = gr.Textbox(label="Введите запрос", placeholder="Например: 'Расскажи историю про ИИ, который научился понимать чувства.'", lines=3)
+    output = gr.Markdown(label="Ответ")
+
+    btn = gr.Button("🚀 Отправить", variant="primary")
+    btn.click(generate_text, inputs=prompt, outputs=output)
+
+    gr.Markdown("---")
+    gr.Markdown("🧩 **Eroha Router Core v5.7** | Автоопределение моделей + безопасный fallback")
+
+demo.launch(server_name="0.0.0.0", server_port=7860)