Deployment optimized

Dockerfile

@@ -33,13 +33,17 @@ RUN python -c "import os; os.environ['HF_HOME']='/app/.cache'; from ner import g
 RUN python -c "import os; os.environ['HF_HOME']='/app/.cache'; from sentiment import get_emotion_classifier; get_emotion_classifier()"
 RUN python -c "import os; os.environ['HF_HOME']='/app/.cache'; from translate import get_translator; get_translator()"
 
-# Create startup script
+# Create startup script with model pre-loading
 RUN echo '#!/bin/bash\n\
 export HF_HOME=/app/.cache\n\
 export TRANSFORMERS_CACHE=/app/.cache\n\
 export HF_HUB_CACHE=/app/.cache\n\
 echo "🚀 Starting FastAPI server on port 8000..."\n\
 python -m uvicorn api:app --host 0.0.0.0 --port 8000 &\n\
+echo "⏳ Waiting for FastAPI to start..."\n\
+sleep 10\n\
+echo "🔥 Warming up models..."\n\
+curl -X POST http://localhost:8000/warmup || echo "Warmup failed, continuing..."\n\
 echo "🌊 Starting Gradio web interface on port 7860..."\n\
 python app.py' > start_services.sh
 

api.py

@@ -73,6 +73,38 @@ def health():
     """Alternative health check endpoint"""
     return health_check()
 
+@app.post("/warmup")
+async def warmup_models():
+    """Pre-load all models to reduce first request time"""
+    try:
+        logger.info("Starting model warmup...")
+        
+        # Pre-load all models
+        from classifier import get_classifier
+        from ner import get_ner_pipeline
+        from sentiment import get_emotion_classifier
+        from translate import get_translator
+        
+        classifier = get_classifier()
+        ner = get_ner_pipeline()
+        emotion_clf = get_emotion_classifier()
+        translator = get_translator()
+        
+        # Test with sample data
+        test_text = "Test tweet for model warmup"
+        classifier(test_text, ["test", "not test"])
+        if ner:
+            ner(test_text)
+        emotion_clf(test_text)
+        translator(test_text)
+        
+        logger.info("Model warmup completed successfully")
+        return {"status": "success", "message": "All models loaded and ready"}
+        
+    except Exception as e:
+        logger.error(f"Model warmup failed: {str(e)}")
+        return {"status": "error", "message": str(e)}
+
 # Main analysis endpoint
 @app.post("/analyze", response_model=TweetAnalysisResponse)
 async def analyze_tweets(request: TweetAnalysisRequest):

app.py

@@ -42,8 +42,9 @@ def run_pipeline(limit=20, hazard_type=None, location=None, days_back=1):
         logger.info(f"Fetched {len(tweets)} tweets")
         
         # Process tweets: translate -> classify -> analyze
+        logger.info("🔄 Processing tweets (this may take 1-2 minutes for first request)...")
         results = classify_tweets(tweets)
-        logger.info(f"Processed {len(results)} tweets (translated, classified, and analyzed)")
+        logger.info(f"✅ Processed {len(results)} tweets (translated, classified, and analyzed)")
         
         # Store hazardous tweets in database (optional)
         try:

classifier.py

@@ -7,7 +7,15 @@ import json
 
 model_name = "cross-encoder/nli-deberta-v3-base"
 
-classifier = pipeline("zero-shot-classification", model=model_name, framework="pt")
+# Lazy loading - only load when needed
+classifier = None
+
+def get_classifier():
+    """Lazy load classifier to avoid startup delay"""
+    global classifier
+    if classifier is None:
+        classifier = pipeline("zero-shot-classification", model=model_name, framework="pt")
+    return classifier
 
 def classify_with_model(tweet_text):
     """
@@ -17,7 +25,8 @@ def classify_with_model(tweet_text):
     if not tweet_text or not tweet_text.strip():
         return 0
     candidate_labels = ["report of an ocean hazard", "not an ocean hazard"]
-    result = classifier(tweet_text, candidate_labels)
+    classifier_instance = get_classifier()
+    result = classifier_instance(tweet_text, candidate_labels)
     top_label = result['labels'][0]
     top_score = result['scores'][0]
     if top_label == "report of an ocean hazard" and top_score > 0.75: