Update app.py

app.py

@@ -352,60 +352,38 @@ def summarize_web_results(query: str, search_results: List[Dict[str, str]], conv
     except Exception as e:
         return f"An error occurred during summarization: {str(e)}"
 
-def get_response_from_excel(query, model, context, num_calls=1, temperature=0.2):
+def get_response_from_excel(query, model, context, num_calls=3, temperature=0.2):
     logging.info(f"Getting response from Excel using model: {model}")
-
-    # Use embeddings to find the most relevant Excel data
-    embed = get_embeddings()
-    office_database = FAISS.load_local("office_faiss_database", embed, allow_dangerous_deserialization=True)
-    retriever = office_database.as_retriever(search_kwargs={"k": 5})
-    relevant_docs = retriever.get_relevant_documents(query)
-
-    # Prepare the context from relevant documents
-    excel_context = "\n".join([doc.page_content for doc in relevant_docs])
-
-    # Prepare the messages for the AI model
+    
     messages = [
-        {"role": "system", "content": "You are an AI assistant specialized in analyzing Excel data. Your task is to provide accurate and detailed responses based solely on the given Excel data context. Do not make assumptions or add information beyond what is explicitly provided in the context."},
-        {"role": "user", "content": f"Based on the following Excel data:\n\n{excel_context}\n\nPlease answer this question: {query}\n\nProvide a step-by-step analysis if applicable, and ensure your response is factual and directly related to the provided Excel data."}
+        {"role": "system", "content": "You are a highly specialized data analyst with expertise in analyzing and summarizing Excel spreadsheets. Your goal is to provide accurate, detailed, and precise information on the data provided. Focus on identifying key metrics, trends, and significant details relevant to the query. Avoid making assumptions or adding information that is not explicitly supported by the data."},
+        {"role": "user", "content": f"Using the following data extracted from Excel spreadsheets:\n{context}\n\nPlease analyze this data and provide a comprehensive answer to the following question: '{query}'. If appropriate, suggest visualizations that could help illustrate the findings."}
     ]
 
-    full_response = ""
     if model.startswith("duckduckgo/"):
-        for _ in range(num_calls):
-            try:
-                ddg_model = model.split('/')[-1]
-                results = DDGS().chat(messages[-1]["content"], model=ddg_model)
-                full_response += results + "\n"
-                logging.info(f"DuckDuckGo API response received for Excel query. Length: {len(results)}")
-            except Exception as e:
-                logging.error(f"Error in generating response from DuckDuckGo for Excel: {str(e)}")
-                yield f"An error occurred with the {model} model: {str(e)}. Please try again."
-                return
+        # Use DuckDuckGo chat with context
+        return get_response_from_duckduckgo(query, model, context, num_calls, temperature)
     elif model == "@cf/meta/llama-3.1-8b-instruct":
-        for response in get_response_from_cloudflare(prompt="", context=excel_context, query=query, num_calls=num_calls, temperature=temperature, search_type="excel"):
-            yield response
-        return
+        # Use Cloudflare API
+        return get_response_from_cloudflare(prompt="", context=context, query=query, num_calls=num_calls, temperature=temperature, search_type="excel")
     else:
+        # Use Hugging Face API
         client = InferenceClient(model, token=huggingface_token)
+
+        response = ""
         for i in range(num_calls):
-            logging.info(f"API call {i+1}/{num_calls} for Excel query")
+            logging.info(f"API call {i+1}/{num_calls}")
             for message in client.chat_completion(
                 messages=messages,
-                max_tokens=1000,
+                max_tokens=20000,
                 temperature=temperature,
                 stream=True,
                 top_p=0.8,
             ):
                 if message.choices and message.choices[0].delta and message.choices[0].delta.content:
                     chunk = message.choices[0].delta.content
-                    full_response += chunk
-                    yield full_response
-
-    if not full_response.strip():
-        yield "I couldn't generate a response based on the Excel data. Please try rephrasing your question or check if the relevant data is present in the uploaded Excel files."
-    else:
-        yield full_response.strip()
+                    response += chunk
+                    yield response  # Yield partial response
 
         logging.info("Finished generating response for Excel data")
 
@@ -455,38 +433,45 @@ def respond(message, history, model, temperature, num_calls, use_web_search, sel
                 yield "No documents available. Please upload documents to answer questions."
                 return
 
-            # Determine if the query is specifically for Excel data
-            excel_files = [doc for doc in selected_docs if doc.lower().endswith(('.xlsx', '.xls'))]
+            all_relevant_docs = []
+            if pdf_database:
+                pdf_retriever = pdf_database.as_retriever(search_kwargs={"k": 10})
+                all_relevant_docs.extend(pdf_retriever.get_relevant_documents(message))
             
-            if excel_files:
-                # If Excel files are selected, use the Excel-specific function
-                for response in get_response_from_excel(message, model, "", num_calls, temperature):
+            if office_database:
+                office_retriever = office_database.as_retriever(search_kwargs={"k": 10})
+                all_relevant_docs.extend(office_retriever.get_relevant_documents(message))
+
+            relevant_docs = [doc for doc in all_relevant_docs if doc.metadata["source"] in selected_docs]
+
+            if not relevant_docs:
+                yield "No relevant information found in the selected documents. Please try selecting different documents or rephrasing your query."
+                return
+
+            # Separate Excel documents from others
+            excel_docs = [doc for doc in relevant_docs if doc.metadata["source"].lower().endswith(('.xlsx', '.xls'))]
+            other_docs = [doc for doc in relevant_docs if not doc.metadata["source"].lower().endswith(('.xlsx', '.xls'))]
+
+            excel_context = "\n".join([doc.page_content for doc in excel_docs])
+            other_context = "\n".join([doc.page_content for doc in other_docs])
+
+            logging.info(f"Excel context length: {len(excel_context)}")
+            logging.info(f"Other context length: {len(other_context)}")
+
+            # Process Excel documents
+            if excel_docs:
+                for response in get_response_from_excel(message, model, excel_context, num_calls, temperature):
                     yield response
-            else:
-                # Existing logic for PDF and other document types
-                all_relevant_docs = []
-                if pdf_database:
-                    pdf_retriever = pdf_database.as_retriever(search_kwargs={"k": 10})
-                    all_relevant_docs.extend(pdf_retriever.get_relevant_documents(message))
-                
-                if office_database:
-                    office_retriever = office_database.as_retriever(search_kwargs={"k": 10})
-                    all_relevant_docs.extend(office_retriever.get_relevant_documents(message))
-
-                relevant_docs = [doc for doc in all_relevant_docs if doc.metadata["source"] in selected_docs]
-
-                if not relevant_docs:
-                    yield "No relevant information found in the selected documents. Please try selecting different documents or rephrasing your query."
-                    return
-
-                context_str = "\n".join([doc.page_content for doc in relevant_docs])
-                
+
+            # Process other documents (PDF, Word)
+            if other_docs:
                 if model == "@cf/meta/llama-3.1-8b-instruct":
-                    for response in get_response_from_cloudflare(prompt="", context=context_str, query=message, num_calls=num_calls, temperature=temperature, search_type="document"):
+                    for response in get_response_from_cloudflare(prompt="", context=other_context, query=message, num_calls=num_calls, temperature=temperature, search_type="document"):
                         yield response
                 else:
                     for response in get_response_from_pdf(message, model, selected_docs, num_calls, temperature):
                         yield response
+
         except Exception as e:
             logging.error(f"Error with {model}: {str(e)}")
             if "microsoft/Phi-3-mini-4k-instruct" in model:
@@ -495,7 +480,7 @@ def respond(message, history, model, temperature, num_calls, use_web_search, sel
                 yield from respond(message, history, fallback_model, temperature, num_calls, use_web_search, selected_docs)
             else:
                 yield f"An error occurred with the {model} model: {str(e)}. Please try again or select a different model."
-
+        
 logging.basicConfig(level=logging.DEBUG)
 
 def get_response_from_cloudflare(prompt, context, query, num_calls=3, temperature=0.2, search_type="pdf"):