Upload folder using huggingface_hub

.gitignore

@@ -12,10 +12,12 @@ wheels/
 .env 
 .env.template
 
+
+#Data & Models (Un-ignored for Hackathon Demo persistence)
 models/
-models 
-.langgraph_api
 data/
 datasets/
+data 
 datasets
-data
+models
+.langgraph_api

Dockerfile

@@ -41,5 +41,9 @@ EXPOSE 7860
 # Set environment variable for HuggingFace
 ENV PORT=7860
 
-# Run API server on port 7860 for HuggingFace Spaces
-CMD ["python", "-c", "import uvicorn; uvicorn.run('main:app', host='0.0.0.0', port=7860)"]
+
+# Set execution permissions for scripts
+RUN chmod +x scripts/start_backend.sh
+
+# Run API server (and ML training) via startup script
+CMD ["/bin/bash", "scripts/start_backend.sh"]

main.py

@@ -1122,14 +1122,35 @@ def get_weather_predictor():
     if _weather_predictor is None:
         try:
             import sys
+            import importlib
             from pathlib import Path
+            
             weather_path = Path(__file__).parent / "models" / "weather-prediction" / "src"
-            sys.path.insert(0, str(weather_path))
+            weather_path_str = str(weather_path)
+            
+            # Ensure path is in sys.path
+            if weather_path_str not in sys.path:
+                sys.path.insert(0, weather_path_str)
+            
+            # CRITICAL FIX: Handle 'components' package name collision
+            # If 'components' is already loaded from another model (e.g. currency), unload it
+            if 'components' in sys.modules:
+                existing_path = getattr(sys.modules['components'], '__file__', '')
+                if existing_path and weather_path_str not in str(existing_path):
+                    logger.warning(f"[WeatherAPI] components collision detected. Unloading {existing_path}")
+                    # Unload generic modules to force reload from new path
+                    for mod in list(sys.modules.keys()):
+                        if mod.startswith('components') or mod.startswith('utils'):
+                            del sys.modules[mod]
+            
+            # Now import fresh
             from components.predictor import WeatherPredictor
             _weather_predictor = WeatherPredictor()
             logger.info("[WeatherAPI] Weather predictor initialized")
         except Exception as e:
             logger.warning(f"[WeatherAPI] Failed to initialize predictor: {e}")
+            import traceback
+            logger.warning(traceback.format_exc())
             _weather_predictor = None
     return _weather_predictor
 

scripts/start_backend.sh

@@ -0,0 +1,21 @@
+#!/bin/bash
+set -e
+
+echo "🚀 Starting ModelX Backend on HuggingFace Space..."
+
+# 1. Run ML Training Pipeline (if models missing)
+# We trust the script to handle logic. For Hackathon, we force run it to ensure fresh state if possible,
+# or we can check if output dir is empty.
+echo "🧠 Checking ML Models..."
+# Create output dir if not exists
+mkdir -p models/anomaly-detection/output
+
+# Run training (standalone script)
+# This will use data from 'datasets/' if available. 
+# If datasets are empty, it might fail/skip, so we allow failure without stopping container.
+python scripts/train_ml_models.py || echo "⚠️ ML Training warning (continuing anyway)..."
+
+# 2. Start Request Server
+# HuggingFace expects us to listen on port 7860
+echo "⚡ Starting FastAPI Server on port $PORT..."
+uvicorn main:app --host 0.0.0.0 --port $PORT

scripts/train_ml_models.py

@@ -0,0 +1,25 @@
+
+import os
+import sys
+from pathlib import Path
+
+# Add the anomaly detection model source to path
+MODEL_DIR = Path(__file__).parent.parent / "models" / "anomaly-detection"
+sys.path.append(str(MODEL_DIR))
+
+try:
+    from src.pipeline.training_pipeline import run_training_pipeline
+    print("🚀 Starting ML Model Training Pipeline (Standalone)...")
+    print(f"📂 Model Directory: {MODEL_DIR}")
+    
+    # Run the pipeline
+    artifact = run_training_pipeline()
+    
+    print("\n✅ Training Complete!")
+    print(f"📊 Model Artifacts stored in: {MODEL_DIR}/output")
+    
+except ImportError as e:
+    print(f"❌ Error: Could not import training pipeline. {e}")
+    print("Ensure you are running this from the project root.")
+except Exception as e:
+    print(f"❌ Training Failed: {e}")

src/utils/utils.py

@@ -518,15 +518,26 @@ def scrape_rivernet_impl(
                             except (ValueError, IndexError):
                                 continue
                     
-                    # Determine status based on keywords (expanded list)
+                    # Determine status based on keywords (refined to avoid false positives)
                     text_lower = page_text.lower()
-                    if any(w in text_lower for w in ["danger", "critical", "evacuate", "red alert", "flood warning"]):
+                    
+                    # Default to normal
+                    river_data["status"] = "normal"
+                    
+                    # DANGER / CRITICAL
+                    if any(w in text_lower for w in ["major flood", "danger level", "critical level", "red alert", "evacuate", "extreme flood"]):
                         river_data["status"] = "danger"
-                    elif any(w in text_lower for w in ["warning", "alert", "high", "amber", "caution"]):
+                        
+                    # WARNING (Stricter: removed generic "high", "alert")
+                    elif any(w in text_lower for w in ["minor flood", "warning level", "flood alert", "amber alert", "high risk", "flood warning"]):
                         river_data["status"] = "warning"
-                    elif any(w in text_lower for w in ["rising", "increasing", "elevated"]):
+                        
+                    # RISING
+                    elif any(w in text_lower for w in ["water level rising", "rising trend", "level is rising"]):
                         river_data["status"] = "rising"
-                    elif any(w in text_lower for w in ["normal", "safe", "stable", "low", "green"]):
+                    
+                    # explicitly check for normal keywords to confirm (optional, as we default to normal)
+                    elif any(w in text_lower for w in ["normal", "safe", "stable", "low", "green", "decreasing"]):
                         river_data["status"] = "normal"
                     
                     results["rivers"].append(river_data)