Update app.py

app.py

@@ -1,91 +1,100 @@
 import os
 import sys
-import json
 import time
-import asyncio
 import logging
 import traceback
 from contextlib import asynccontextmanager
+from typing import List, Dict, Any
+
 from fastapi import FastAPI, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
 from pydantic import BaseModel, Field, validator
-from typing import Dict
-import uvicorn
 
-# ==================== Безопасные импорты ====================
+# ==================== Safe optional imports ====================
+
+try:
+    from transformers import pipeline  # type: ignore
+except Exception as e:  # pragma: no cover
+    pipeline = None  # type: ignore
+    logging.error(f"[ImportError] transformers not available: {e}", file=sys.stderr)  # type: ignore
+
 try:
-    from transformers import pipeline
-    from langdetect import detect
-except Exception as e:
-    logging.error(f"[ImportError] transformers/langdetect not available: {e}")
-    pipeline = None
-    detect = lambda x: "en"
-
-# ==================== Логирование ====================
+    from langdetect import detect  # type: ignore
+except Exception as e:  # pragma: no cover
+    detect = None  # type: ignore
+    logging.error(f"[ImportError] langdetect not available: {e}", file=sys.stderr)  # type: ignore
+
+
+# ==================== Logging ====================
+
 logging.basicConfig(
     level=logging.INFO,
     format="%(asctime)s [%(levelname)s] %(message)s",
     handlers=[logging.StreamHandler(sys.stderr)],
 )
 
-# ==================== Конфигурация ====================
+logger = logging.getLogger("eroha")
+
+
+# ==================== Environment / HF cache ====================
+
 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
 
+# ==================== Pydantic models ====================
 
-app = FastAPI(title="Eroha AI Summarizer PRO", version="v3.4", lifespan=lifespan)
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=["*"],
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
 
-# ==================== Pydantic модели ====================
 class SummarizeRequest(BaseModel):
     text: str = Field(..., min_length=3, max_length=1_000_000)
 
     @validator("text")
-    def not_empty(cls, v):
+    def not_empty(cls, v: str) -> str:
         if not v.strip():
             raise ValueError("Text cannot be empty or whitespace only")
         return v
 
 
 class CheckRequest(BaseModel):
-    data: str = Field(..., min_length=1, max_length=500_000)
+    text: str = Field(..., min_length=1, max_length=500_000)
+
+    @validator("text")
+    def not_empty(cls, v: str) -> str:
+        if not v.strip():
+            raise ValueError("Text cannot be empty or whitespace only")
+        return v
+
+
+class HealthResponse(BaseModel):
+    status: str
+    version: str
+    endpoints: List[str]
+    cached_models: List[str]
+
+
+# ==================== Model management ====================
+
+_model_cache: Dict[str, Any] = {}
 
 
-# ==================== Утилиты ====================
-def safe_detect_lang(text: str) -> str:
+def _safe_detect_lang(text: str) -> str:
+    if detect is None:
+        return "en"
     try:
-        return detect(text)
+        return detect(text)  # type: ignore[call-arg]
     except Exception:
         return "en"
 
 
-def get_model(lang: str):
-    if not pipeline:
+def _get_model(lang: str):
+    if pipeline is None:
         raise RuntimeError("Transformers pipeline unavailable")
 
+    if lang in _model_cache:
+        logger.info("[ModelCache] using cached model for %s", lang)
+        return _model_cache[lang]
+
     model_map = {
         "ru": "IlyaGusev/mbart_ru_sum_gazeta",
         "kk": "facebook/mbart-large-50-many-to-many-mmt",
@@ -96,81 +105,171 @@ def get_model(lang: str):
     }
     model_name = model_map.get(lang, "facebook/bart-large-cnn")
 
-    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
-
+    logger.info("[ModelLoad] loading model for %s: %s", lang, model_name)
+    try:
+        # CPU-режим (device=-1) — безопасно для HF Spaces
+        model = pipeline("summarization", model=model_name, device=-1)  # type: ignore[arg-type]
+        _model_cache[lang] = model
+        logger.info("[ModelLoad] model for %s ready", lang)
+        return model
+    except Exception as e:  # pragma: no cover
+        logger.error("[ModelLoad] failed to load %s: %s", lang, e)
+        raise
+
+
+def _summarize(text: str, lang: str) -> Dict[str, Any]:
+    summarizer = _get_model(lang)
+
+    max_input_length = 3000  # защита от слишком длинных текстов
+    input_text = text[:max_input_length]
+
+    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"},
+    }
 
-# ==================== Эндпоинты ====================
-@app.get("/")
-async def home():
     return {
-        "status": "ok",
-        "version": app.version,
-        "cached_models": list(_model_cache.keys()),
-        "endpoints": ["/ping", "/check", "/summarize", "/warmup"],
+        "summary": summary,
+        "summary_length": len(summary),
+        "original_length": len(text),
+        "truncated": len(text) > max_input_length,
+        "seo_json_ld": json_ld,
     }
 
 
+# ==================== Lifespan: warmup on startup ====================
+
+
+@asynccontextmanager
+async def lifespan(app: FastAPI):  # type: ignore[override]
+    logger.info("[Startup] application starting, warming models...")
+    start = time.time()
+    if pipeline is not None:
+        for lang in ("en", "ru"):
+            try:
+                _get_model(lang)
+            except Exception as e:
+                logger.error("[Startup] warmup for %s failed: %s", lang, e)
+    else:
+        logger.warning("[Startup] transformers pipeline unavailable, skipping warmup")
+
+    logger.info("[Startup] warmup finished in %.2fs", time.time() - start)
+    yield
+    logger.info("[Shutdown] application stopping")
+
+
+# ==================== FastAPI app ====================
+
+
+app = FastAPI(
+    title="Eroha AI Summarizer PRO",
+    version="v3.4",
+    lifespan=lifespan,
+)
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+
+# ==================== Routes ====================
+
+
+@app.get("/", response_model=HealthResponse)
+async def root() -> HealthResponse:
+    return HealthResponse(
+        status="ok",
+        version="v3.4",
+        endpoints=["/ping", "/check", "/summarize", "/warmup"],
+        cached_models=list(_model_cache.keys()),
+    )
+
+
 @app.get("/ping")
 async def ping():
-    return {"status": "healthy", "cached_models": list(_model_cache.keys())}
+    return {
+        "status": "healthy",
+        "cached_models": list(_model_cache.keys()),
+        "time": time.time(),
+    }
 
 
 @app.get("/warmup")
-async def warmup():
-    return {"status": "warm", "models_ready": len(_model_cache) > 0}
+async def manual_warmup():
+    if pipeline is None:
+        return {
+            "status": "skipped",
+            "reason": "transformers pipeline unavailable",
+        }
+
+    loaded = []
+    errors = {}
+    for lang in ("en", "ru"):
+        try:
+            _get_model(lang)
+            loaded.append(lang)
+        except Exception as e:  # pragma: no cover
+            errors[lang] = str(e)
+
+    return {
+        "status": "ok" if not errors else "partial",
+        "loaded": loaded,
+        "errors": errors,
+        "cached_models": list(_model_cache.keys()),
+    }
 
 
 @app.post("/check")
 async def check_text(req: CheckRequest):
     try:
-        lang = safe_detect_lang(req.data)
+        lang = _safe_detect_lang(req.text)
         return {
             "status": "success",
-            "preview": req.data[:150],
-            "length": len(req.data),
-            "language": lang,
+            "preview": req.text[:150],
+            "length": len(req.text),
+            "detected_language": lang,
         }
-    except Exception as e:
-        logging.error(f"/check error: {traceback.format_exc()}")
+    except Exception as e:  # pragma: no cover
+        logger.error("/check error: %s", 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"},
-        }
+    if pipeline is None:
+        raise HTTPException(
+            status_code=503,
+            detail="transformers pipeline is not available in this environment",
+        )
 
-        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()}")
+    try:
+        lang = _safe_detect_lang(req.text)
+        logger.info("[Summarize] language=%s, length=%d", lang, len(req.text))
+        data = _summarize(req.text, lang)
+        return {"status": "success", "language": lang, **data}
+    except Exception as e:  # pragma: no cover
+        logger.error("/summarize error: %s", traceback.format_exc())
         raise HTTPException(status_code=500, detail=str(e))
 
 
-# ==================== Запуск ====================
 if __name__ == "__main__":
-    uvicorn.run("app:app", host="0.0.0.0", port=7860, reload=False)
+    import uvicorn
+
+    port = int(os.getenv("PORT", "7860"))
+    uvicorn.run("app:app", host="0.0.0.0", port=port, reload=False)
+