Update app.py

app.py

@@ -74,7 +74,8 @@ async def lifespan(app: FastAPI):
         # cross_encoder_model = HuggingFaceCrossEncoder(model_name="BAAI/bge-reranker-large")
         ml_models["reranker_compressor"] = CrossEncoderReranker(model=cross_encoder_model, top_n=6)
         ml_models["llm"] = ChatGoogleGenerativeAI(
-                               model="gemini-1.5-pro",
+                               # model="gemini-1.5-pro",
+                               model="gemini-2.0-flash,
                                api_key=GOOGLE_API_KEY,
                                temperature=0.1,
                                max_output_tokens=300
@@ -166,12 +167,12 @@ async def run_hackrx(req: RunRequest):
     )
 
     # dense_retriever = vectorstore.as_retriever(search_type="mmr",search_kwargs={"k": 8})
-    dense_retriever = vectorstore.as_retriever(search_type="mmr",search_kwargs={"k": 5 ,"lambda_mult": 0.5})
+    dense_retriever = vectorstore.as_retriever(search_type="mmr",search_kwargs={"k": 7 ,"lambda_mult": 0.5})
     
 
     # Create retrievers using the pre-loaded models from our ml_models dictionary
     keyword_retriever = BM25Retriever.from_documents(chunks)
-    keyword_retriever.k = 4
+    keyword_retriever.k = 5
     # dense_retriever = Chroma.from_documents(documents=chunks, embedding=ml_models["embedder"]).as_retriever()
     ensemble_retriever = EnsembleRetriever(retrievers=[keyword_retriever, dense_retriever], weights=[0.35, 0.65])
     ### to make it faster we are now using our built reranker thats why commenting the code below