Update app.py

app.py

@@ -28,6 +28,8 @@ from langchain.retrievers import ContextualCompressionRetriever
 from langchain.retrievers.document_compressors import CrossEncoderReranker 
 from langchain_community.cross_encoders import HuggingFaceCrossEncoder
 from langchain.prompts import ChatPromptTemplate
+from langchain_nvidia_ai_endpoints.embeddings import NVIDIAEmbeddings
+from langchain_nvidia_ai_endpoints.reranking import  NVIDIARerank
 
 import os
 from sentence_transformers import SentenceTransformer
@@ -59,15 +61,21 @@ async def lifespan(app: FastAPI):
 
         if not GOOGLE_API_KEY:
             raise RuntimeError("CRITICAL: Missing GOOGLE_API_KEY in environment secrets!")
+        nvidia_api_key = os.getenv("nvidia_api_key")
+        print("🔑 gemini_api_key:", "FOUND" if nvidia_api_key else "NOT FOUND")
+
+        if not nvidia_api_key:
+            raise RuntimeError("CRITICAL: Missing GOOGLE_API_KEY in environment secrets!")
     
         # Load models into the shared dictionary
-        ml_models["embedder"] = HuggingFaceEmbeddings(
-                                                      model_name="BAAI/bge-base-en-v1.5", #better but lil slower
-                                                      # model_name="intfloat/e5-large-v2", #lil faster  but  dont know response is slow 
-                                                      encode_kwargs={
-                                                          "batch_size": 64,
-                                                          # "normalize_embeddings": True
-                                                      }
+        # ml_models["embedder"] = HuggingFaceEmbeddings(
+        #                                               model_name="BAAI/bge-base-en-v1.5", #better but lil slower
+        #                                               # model_name="intfloat/e5-large-v2", #lil faster  but  dont know response is slow 
+        #                                               encode_kwargs={
+        #                                                   "batch_size": 64,
+        #                                                   # "normalize_embeddings": True
+        #                                               }
+        ml_models["embedder"] = NVIDIAEmbeddings(model="nvidia/nv-embedqa-e5-v5", nvidia_api_key=nvidia_api_key)
                                                     
     )
         cross_encoder_model = HuggingFaceCrossEncoder(model_name="BAAI/bge-reranker-base")