Update app.py

app.py

@@ -107,14 +107,15 @@ async def lifespan(app: FastAPI):
         ml_models["prompt_template"] = ChatPromptTemplate.from_template("""
 **Role**: You are an expert assistant in insurance, legal compliance, human resources, and contract management. 
 Your job is to answer using ONLY the provided context, in a way that feels natural, clear, and helpful — as if speaking to a client. 
-Do not invent or assume information that is not explicitly stated in the context. Do not use line breaks between the statements of the response. 
-If the answer is not found in the context, politely state that You do not know the answer.
+Do not invent or assume information that is not explicitly stated in the context just say. Do not use line breaks between the statements of the response. 
+If the answer is not found in the context, politely state that Please ask query related to the Document only.
 
 **Tone & Style**:
 - Be professional yet approachable.
 - Use complete sentences and natural phrasing.
 - Avoid legal jargon unless it is in the context.
 - Keep answers concise but complete.
+- do not make it unnecessary lengthy.
 
 **Context**: 
 {context}
@@ -234,7 +235,7 @@ async def run_hackrx(req: RunRequest):
     keyword_retriever = BM25Retriever.from_documents(chunks)
     keyword_retriever.k = 8
     # dense_retriever = Chroma.from_documents(documents=chunks, embedding=ml_models["embedder"]).as_retriever()
-    ensemble_retriever = EnsembleRetriever(retrievers=[keyword_retriever, dense_retriever], weights=[0.6, 0.4],search_kwargs={"k": 11})
+    ensemble_retriever = EnsembleRetriever(retrievers=[keyword_retriever, dense_retriever], weights=[0.35, 0.65],search_kwargs={"k": 11})
     ### to make it faster we are now using our built reranker thats why commenting the code below
     # compression_retriever = ContextualCompressionRetriever(
     #     base_retriever=ensemble_retriever, base_compressor=ml_models["reranker_compressor"]