Update app.py

app.py

@@ -1,139 +1,176 @@
 import os
+import sys
 import json
 import time
-import hashlib
-import gradio as gr
-from fastapi import FastAPI
+import asyncio
+import logging
+import traceback
+from contextlib import asynccontextmanager
+from fastapi import FastAPI, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
-from transformers import pipeline
-from langdetect import detect
-from datetime import datetime
-
-
-# ============ Проверка наличия модулей ============
-MODULES = [
-    "Eroha_PromptCore.md",
-    "Eroha_Dashboard.md",
-    "Eroha_AgentAPI_v1.1.md",
-    "Eroha_MetricsCore_v1.0.md"
-]
-AVAILABLE_MODULES = [m for m in MODULES if os.path.exists(m)]
-MISSING_MODULES = [m for m in MODULES if m not in AVAILABLE_MODULES]
-
-
-# ============ Проверка загрузчика ============
-def loader_health():
-    table = []
-    for mod in MODULES:
-        status = "✅ OK" if mod in AVAILABLE_MODULES else "❌ Missing"
-        table.append(f"| {mod} | {status} |")
-
-    trace_id = hashlib.md5(str(time.time()).encode()).hexdigest()[:8]
-    summary = "\n".join(table)
-
-    return f"""
-## 🔍 Eroha Loader — Health Check
-| Модуль | Статус |
-|---|---|
-{summary}
-
-**TL;DR:** {len(AVAILABLE_MODULES)} из {len(MODULES)} активны.  
-**Вывод:** {'Все работает корректно ✅' if not MISSING_MODULES else 'Есть отсутствующие модули ⚠️'}  
-**Риски:** {', '.join(MISSING_MODULES) if MISSING_MODULES else '—'}  
-**Рекомендации:** Проверь структуру проекта в Hugging Face (Files → Upload)  
-trace_id: {trace_id}
-"""
-
+from pydantic import BaseModel, Field, validator
+from typing import Dict
+import uvicorn
+
+# ==================== Безопасные импорты ====================
+try:
+    from transformers import pipeline
+    from langdetect import detect
+except Exception as e:
+    logging.error(f"[ImportError] transformers/langdetect not available: {e}", file=sys.stderr)
+    pipeline = None
+    detect = lambda x: "en"
+
+# ==================== Логирование ====================
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s [%(levelname)s] %(message)s",
+    handlers=[logging.StreamHandler(sys.stderr)],
+)
 
-# ============ Модель суммаризации ============
-summarizers = {}
+# ==================== Конфигурация ====================
+HF_HOME = "/tmp/huggingface"
+os.environ["HF_HOME"] = HF_HOME
+os.makedirs(HF_HOME, exist_ok=True)
+
+_model_cache: Dict[str, any] = {}
+
+# ==================== Lifespan Context ====================
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+    start = time.time()
+    preload_models = ["facebook/bart-large-cnn", "IlyaGusev/mbart_ru_sum_gazeta"]
+    if pipeline:
+        for model_name in preload_models:
+            try:
+                _model_cache[model_name] = pipeline("summarization", model=model_name, device=-1)
+                logging.info(f"[Warmup] Preloaded model: {model_name}")
+            except Exception as e:
+                logging.error(f"[Warmup] Failed preload {model_name}: {e}")
+    logging.info(f"[Startup] Models initialized in {time.time() - start:.2f}s")
+    yield
+
+
+app = FastAPI(title="Eroha AI Summarizer PRO", version="v3.4", lifespan=lifespan)
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
 
-def get_summarizer(lang: str):
-    if lang == "ru":
-        model_name = "IlyaGusev/mbart_ru_sum_gazeta"
-    elif lang == "kk":
-        model_name = "facebook/mbart-large-50-many-to-many-mmt"
-    else:
-        model_name = "facebook/bart-large-cnn"
+# ==================== Pydantic модели ====================
+class SummarizeRequest(BaseModel):
+    text: str = Field(..., min_length=3, max_length=1_000_000)
 
-    if model_name not in summarizers:
-        summarizers[model_name] = pipeline("summarization", model=model_name)
+    @validator("text")
+    def not_empty(cls, v):
+        if not v.strip():
+            raise ValueError("Text cannot be empty or whitespace only")
+        return v
 
-    return summarizers[model_name]
 
+class CheckRequest(BaseModel):
+    data: str = Field(..., min_length=1, max_length=500_000)
 
-def summarize_text(text: str):
-    if not text.strip():
-        return "❌ Введите текст для анализа."
 
+# ==================== Утилиты ====================
+def safe_detect_lang(text: str) -> str:
     try:
-        lang = detect(text)
+        return detect(text)
     except Exception:
-        lang = "en"
+        return "en"
 
-    model = get_summarizer(lang)
-    result = model(text, max_length=160, min_length=40, do_sample=False)
-    summary = result[0]["summary_text"].replace("▁", " ").strip()
 
-    now = datetime.utcnow().strftime("%Y-%m-%d %H:%M:%S")
+def get_model(lang: str):
+    if not pipeline:
+        raise RuntimeError("Transformers pipeline unavailable")
 
-    json_ld = {
-        "@context": "https://schema.org",
-        "@type": "NewsArticle",
-        "headline": summary[:80],
-        "datePublished": now,
-        "inLanguage": lang,
-        "publisher": {"@type": "Organization", "name": "Eroha AI Publisher"},
+    model_map = {
+        "ru": "IlyaGusev/mbart_ru_sum_gazeta",
+        "kk": "facebook/mbart-large-50-many-to-many-mmt",
+        "de": "facebook/bart-large-cnn",
+        "es": "facebook/mbart-large-50-many-to-many-mmt",
+        "fr": "facebook/mbart-large-50-many-to-many-mmt",
+        "en": "facebook/bart-large-cnn",
     }
+    model_name = model_map.get(lang, "facebook/bart-large-cnn")
 
-        formatted = f"""# 🧠 Eroha Summarizer PRO v2.5 (AutoLang: {lang.upper()})
-**Дата:** {now}
-
-## 📘 Резюме
-{summary}
+    if model_name in _model_cache:
+        return _model_cache[model_name]
 
----
+    logging.info(f"[ModelLoad] Loading model dynamically: {model_name}")
+    model = pipeline("summarization", model=model_name, device=-1)
+    _model_cache[model_name] = model
+    return model
 
-## 🧩 Schema.org JSON-LD
-```json
-{json.dumps(json_ld, ensure_ascii=False, indent=2)}
-
-TL;DR: {summary[:140]}{'...' if len(summary) > 140 else ''}
-"""
-    return formatted
-
-# ============ FastAPI API ============
-
-app = FastAPI(title="Eroha Summarizer PRO v2.5")
-
-app.add_middleware(
-CORSMiddleware,
-allow_origins=["*"],
-allow_methods=["*"],
-allow_headers=["*"]
-)
 
+# ==================== Эндпоинты ====================
 @app.get("/")
 async def home():
-    return {"status": "ok", "message": "Eroha Core PRO v2.5 active"}
+    return {
+        "status": "ok",
+        "version": app.version,
+        "cached_models": list(_model_cache.keys()),
+        "endpoints": ["/ping", "/check", "/summarize", "/warmup"],
+    }
 
-@app.post("/api/summarize")
-async def summarize_api(data: dict):
-return {"summary": summarize_text(data.get("text", ""))}
 
-@app.get("/api/core-checkup")
-async def core_checkup():
-return {"report": loader_health()}
+@app.get("/ping")
+async def ping():
+    return {"status": "healthy", "cached_models": list(_model_cache.keys())}
 
-# ============ Gradio интерфейс ============
 
-iface = gr.Interface(
-fn=summarize_text,
-inputs=gr.Textbox(lines=10, label="Введите текст для анализа"),
-outputs=gr.Markdown(label="Результат"),
-title="Eroha Summarizer PRO v2.5 — Publisher Edition",
-description="AI-инструмент для суммаризации, автоопределения языка (RU, EN, KZ, DE, ES, FR) и SEO-разметки (JSON-LD)"
-)
+@app.get("/warmup")
+async def warmup():
+    return {"status": "warm", "models_ready": len(_model_cache) > 0}
+
 
+@app.post("/check")
+async def check_text(req: CheckRequest):
+    try:
+        lang = safe_detect_lang(req.data)
+        return {
+            "status": "success",
+            "preview": req.data[:150],
+            "length": len(req.data),
+            "language": lang,
+        }
+    except Exception as e:
+        logging.error(f"/check error: {traceback.format_exc()}")
+        raise HTTPException(status_code=500, detail=str(e))
+
+
+@app.post("/summarize")
+async def summarize(req: SummarizeRequest):
+    try:
+        lang = safe_detect_lang(req.text)
+        summarizer = get_model(lang)
+        input_text = req.text[:3000]
+        result = summarizer(input_text, max_length=180, min_length=40, do_sample=False)
+        summary = result[0]["summary_text"].replace("▁", " ").strip()
+
+        json_ld = {
+            "@context": "https://schema.org",
+            "@type": "NewsArticle",
+            "headline": summary[:80],
+            "inLanguage": lang,
+            "publisher": {"@type": "Organization", "name": "Eroha AI Publisher"},
+        }
+
+        return {
+            "status": "success",
+            "language": lang,
+            "summary": summary,
+            "summary_length": len(summary),
+            "original_length": len(req.text),
+            "seo_json_ld": json_ld,
+        }
+    except Exception as e:
+        logging.error(f"/summarize error: {traceback.format_exc()}")
+        raise HTTPException(status_code=500, detail=str(e))
+
+
+# ==================== Запуск ====================
 if __name__ == "__main__":
-iface.launch(server_name="0.0.0.0", server_port=7860)
+    uvicorn.run("app:app", host="0.0.0.0", port=7860, reload=False)