Upload backend/hue_portal/hue_portal/preload_models.py with huggingface_hub

backend/hue_portal/hue_portal/preload_models.py

@@ -0,0 +1,113 @@
+"""
+Preload all models when worker process starts.
+This module is imported by wsgi.py to ensure models are loaded before first request.
+"""
+import os
+import sys
+
+def preload_all_models():
+    """Preload all models (embedding, LLM, reranker) in worker process."""
+    print('[PRELOAD] 🔄 Starting model preload in worker process...', flush=True)
+    try:
+        # 1. Preload Embedding Model (BGE-M3)
+        try:
+            print('[PRELOAD] 📦 Preloading embedding model (BGE-M3)...', flush=True)
+            from hue_portal.core.embeddings import get_embedding_model
+            embedding_model = get_embedding_model()
+            if embedding_model:
+                print('[PRELOAD] ✅ Embedding model preloaded successfully', flush=True)
+            else:
+                print('[PRELOAD] ⚠️ Embedding model not loaded', flush=True)
+        except Exception as e:
+            print(f'[PRELOAD] ⚠️ Embedding model preload failed: {e}', flush=True)
+        
+        # 2. Preload LLM Model (llama.cpp)
+        llm_provider = os.environ.get('DEFAULT_LLM_PROVIDER') or os.environ.get('LLM_PROVIDER', '')
+        if llm_provider.lower() == 'llama_cpp':
+            try:
+                print('[PRELOAD] 📦 Preloading LLM model (llama.cpp)...', flush=True)
+                from hue_portal.chatbot.llm_integration import get_llm_generator
+                llm_gen = get_llm_generator()
+                if llm_gen and hasattr(llm_gen, 'llama_cpp') and llm_gen.llama_cpp:
+                    print('[PRELOAD] ✅ LLM model preloaded successfully', flush=True)
+                else:
+                    print('[PRELOAD] ⚠️ LLM model not loaded (may load on first request)', flush=True)
+            except Exception as e:
+                print(f'[PRELOAD] ⚠️ LLM model preload failed: {e} (will load on first request)', flush=True)
+        else:
+            print(f'[PRELOAD] ⏭️ Skipping LLM preload (provider is {llm_provider or "not set"}, not llama_cpp)', flush=True)
+        
+        # 3. Preload Reranker Model
+        try:
+            print('[PRELOAD] 📦 Preloading reranker model...', flush=True)
+            from hue_portal.core.reranker import get_reranker
+            reranker = get_reranker()
+            if reranker:
+                print('[PRELOAD] ✅ Reranker model preloaded successfully', flush=True)
+            else:
+                print('[PRELOAD] ⚠️ Reranker model not loaded (may load on first request)', flush=True)
+        except Exception as e:
+            print(f'[PRELOAD] ⚠️ Reranker preload failed: {e} (will load on first request)', flush=True)
+        
+        print('[PRELOAD] ✅ Model preload completed in worker process', flush=True)
+    except Exception as e:
+        print(f'[PRELOAD] ⚠️ Model preload error: {e} (models will load on first request)', flush=True)
+        import traceback
+        traceback.print_exc()
+
+Preload all models when worker process starts.
+This module is imported by wsgi.py to ensure models are loaded before first request.
+"""
+import os
+import sys
+
+def preload_all_models():
+    """Preload all models (embedding, LLM, reranker) in worker process."""
+    print('[PRELOAD] 🔄 Starting model preload in worker process...', flush=True)
+    try:
+        # 1. Preload Embedding Model (BGE-M3)
+        try:
+            print('[PRELOAD] 📦 Preloading embedding model (BGE-M3)...', flush=True)
+            from hue_portal.core.embeddings import get_embedding_model
+            embedding_model = get_embedding_model()
+            if embedding_model:
+                print('[PRELOAD] ✅ Embedding model preloaded successfully', flush=True)
+            else:
+                print('[PRELOAD] ⚠️ Embedding model not loaded', flush=True)
+        except Exception as e:
+            print(f'[PRELOAD] ⚠️ Embedding model preload failed: {e}', flush=True)
+        
+        # 2. Preload LLM Model (llama.cpp)
+        llm_provider = os.environ.get('DEFAULT_LLM_PROVIDER') or os.environ.get('LLM_PROVIDER', '')
+        if llm_provider.lower() == 'llama_cpp':
+            try:
+                print('[PRELOAD] 📦 Preloading LLM model (llama.cpp)...', flush=True)
+                from hue_portal.chatbot.llm_integration import get_llm_generator
+                llm_gen = get_llm_generator()
+                if llm_gen and hasattr(llm_gen, 'llama_cpp') and llm_gen.llama_cpp:
+                    print('[PRELOAD] ✅ LLM model preloaded successfully', flush=True)
+                else:
+                    print('[PRELOAD] ⚠️ LLM model not loaded (may load on first request)', flush=True)
+            except Exception as e:
+                print(f'[PRELOAD] ⚠️ LLM model preload failed: {e} (will load on first request)', flush=True)
+        else:
+            print(f'[PRELOAD] ⏭️ Skipping LLM preload (provider is {llm_provider or "not set"}, not llama_cpp)', flush=True)
+        
+        # 3. Preload Reranker Model
+        try:
+            print('[PRELOAD] 📦 Preloading reranker model...', flush=True)
+            from hue_portal.core.reranker import get_reranker
+            reranker = get_reranker()
+            if reranker:
+                print('[PRELOAD] ✅ Reranker model preloaded successfully', flush=True)
+            else:
+                print('[PRELOAD] ⚠️ Reranker model not loaded (may load on first request)', flush=True)
+        except Exception as e:
+            print(f'[PRELOAD] ⚠️ Reranker preload failed: {e} (will load on first request)', flush=True)
+        
+        print('[PRELOAD] ✅ Model preload completed in worker process', flush=True)
+    except Exception as e:
+        print(f'[PRELOAD] ⚠️ Model preload error: {e} (models will load on first request)', flush=True)
+        import traceback
+        traceback.print_exc()
+