Update app.py

app.py

@@ -291,33 +291,6 @@ def upload_and_prepare_old(file,user):
   return mm, pdf_links  # Update both outputs
   #return mm
 #create_faiss_index(repo_id, file_input)
-#======================================================================
-def get_document_summary(repo_id,query,llm=None):
-    """Generate summary of all documents in repo"""
-    try:
-        # Load vectorstore
-        embeddings = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")
-        faiss_path = hf_hub_download(repo_id=repo_id, filename="index.faiss", repo_type="dataset")
-        vectorstore = FAISS.load_local(os.path.dirname(faiss_path), embeddings, allow_dangerous_deserialization=True)
-        
-        # Get top documents
-        docs = vectorstore.similarity_search(query, k=20)
-        context = "\n".join([doc.page_content[:200] for doc in docs])
-        
-        # Summarize with your LLM
-        summary_prompt = f"""
-        Summarize these context:
-        
-        {context[:3000]}
-        
-        Summary:
-        """
-        summary = llm(summary_prompt)  # Uses TinyLlama
-        return summary
-    except:
-        return "Summary unavailable"
-
-
 #======================================================================
 def generate_qa_chain(repo_id, embedding_model="sentence-transformers/all-MiniLM-L6-v2", llm=None):
     """
@@ -516,10 +489,6 @@ with gr.Blocks(title="N R L C H A T B O T - for commercial procurement - Supply"
                 ask_question1, 
                 inputs=query_input1, 
                 outputs=answer_output1
-            ).then(  # Auto-trigger after answer
-                get_document_summary,
-                inputs=[repo_id,query_input1,llm1],
-                outputs=summary_output
             )    
     
     with gr.Row():