Update app.py

app.py

@@ -232,6 +232,7 @@
 
 import os
 import time
+import logging
 from fastapi import FastAPI, Request
 from fastapi.responses import HTMLResponse
 from fastapi.staticfiles import StaticFiles
@@ -246,6 +247,10 @@ from fastapi.templating import Jinja2Templates
 from simple_salesforce import Salesforce, SalesforceLogin
 from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
 
+# Set up logging
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
 # Define Pydantic model for incoming request body
 class MessageRequest(BaseModel):
     message: str
@@ -294,7 +299,7 @@ Settings.llm = HuggingFaceLLM(
     tokenizer_name="google/flan-t5-small",
     context_window=512,  # flan-t5-small has a max context window of 512 tokens
     max_new_tokens=256,
-    generate_kwargs={"temperature": 0.1, "do_sample": True},
+    generate_kwargs={"temperature": 0.3, "do_sample": True},  # Increased temperature for better responses
     model=AutoModelForSeq2SeqLM.from_pretrained("google/flan-t5-small"),
     tokenizer=tokenizer,
     device_map="auto"  # Automatically use GPU if available, else CPU
@@ -313,15 +318,23 @@ chat_history = []
 current_chat_history = []
 
 def data_ingestion_from_directory():
-    documents = SimpleDirectoryReader(PDF_DIRECTORY).load_data()
-    storage_context = StorageContext.from_defaults()
-    index = VectorStoreIndex.from_documents(documents)
-    index.storage_context.persist(persist_dir=PERSIST_DIR)
+    try:
+        documents = SimpleDirectoryReader(PDF_DIRECTORY).load_data()
+        logger.info(f"Loaded {len(documents)} documents from {PDF_DIRECTORY}")
+        if not documents:
+            logger.warning(f"No documents found in {PDF_DIRECTORY}. Ensure PDF files are present.")
+        storage_context = StorageContext.from_defaults()
+        index = VectorStoreIndex.from_documents(documents)
+        index.storage_context.persist(persist_dir=PERSIST_DIR)
+        logger.info(f"Index persisted to {PERSIST_DIR}")
+    except Exception as e:
+        logger.error(f"Error during data ingestion: {str(e)}")
+        raise
 
 def initialize():
     start_time = time.time()
     data_ingestion_from_directory()  # Process PDF ingestion at startup
-    print(f"Data ingestion time: {time.time() - start_time} seconds")
+    logger.info(f"Data ingestion time: {time.time() - start_time} seconds")
 
 def split_name(full_name):
     # Split the name by spaces
@@ -343,10 +356,10 @@ def split_name(full_name):
 initialize()  # Run initialization tasks
 
 def handle_query(query):
-    # Custom prompt template for flan-t5-small (no chat template)
+    # Custom prompt template for flan-t5-small
     text_qa_template = PromptTemplate(
         """
-        You are Clara, a Redfernstech chatbot. Provide accurate, concise answers (10-15 words) based on company data.
+        You are Clara, a Redfernstech chatbot. Answer the question in 10-15 words based on the provided context.
         Context: {context_str}
         Question: {query_str}
         Answer:
@@ -360,15 +373,22 @@ def handle_query(query):
         if past_query.strip():
             context_str += f"User asked: '{past_query}'\nBot answered: '{response}'\n"
     
+    logger.info(f"Query: {query}")
+    logger.info(f"Context: {context_str}")
+    
     query_engine = index.as_query_engine(text_qa_template=text_qa_template)
     answer = query_engine.query(query)
-
-    if hasattr(answer, "response"):
-        response = answer.response
-    elif isinstance(answer, dict) and "response" in answer:
-        response = answer["response"]
+    
+    logger.info(f"Raw query engine output: {answer}")
+    
+    if hasattr(answer, "response") and answer.response:
+        response = answer.response.strip()
+    elif isinstance(answer, dict) and "response" in answer and answer["response"]:
+        response = answer["response"].strip()
     else:
         response = "Sorry, I couldn't find an answer."
+    
+    logger.info(f"Processed response: {response}")
     current_chat_history.append((query, response))
     return response
 
@@ -380,7 +400,7 @@ async def load_chat(request: Request, id: str):
 async def save_chat_history(history: dict):
     # Check if 'userId' is present in the incoming dictionary
     user_id = history.get("userId")
-    print(user_id)
+    logger.info(f"Received userId: {user_id}")
 
     # Ensure user_id is defined before proceeding
     if user_id is None:
@@ -389,7 +409,7 @@ async def save_chat_history(history: dict):
     # Construct the chat history string
     hist = "".join([f"'{entry['sender']}: {entry['message']}'\n" for entry in history["history"]])
     hist = "You are a Redfernstech summarize model. Your aim is to use this conversation to identify user interests solely based on that conversation: " + hist
-    print(hist)
+    logger.info(f"Chat history: {hist}")
 
     # Get the summarized result
     result = hist
@@ -397,6 +417,7 @@ async def save_chat_history(history: dict):
     try:
         sf.Lead.update(user_id, {"Description": result})
     except Exception as e:
+        logger.error(f"Failed to update lead: {str(e)}")
         return {"error": f"Failed to update lead: {str(e)}"}, 500
     
     return {"summary": result, "message": "Chat history saved"}
@@ -418,8 +439,7 @@ async def receive_form_data(request: Request):
     # Generate a unique ID (for tracking user)
     unique_id = a["id"]
     
-    # Here you can do something with form_data like saving it to a database
-    print("Received form data:", form_data)
+    logger.info(f"Received form data: {form_data}")
     
     # Send back the unique id to the frontend
     return JSONResponse({"id": unique_id})
@@ -427,6 +447,7 @@ async def receive_form_data(request: Request):
 @app.post("/chat/")
 async def chat(request: MessageRequest):
     message = request.message  # Access the message from the request body
+    logger.info(f"Received chat message: {message}")
     response = handle_query(message)  # Process the message
     message_data = {
         "sender": "User",
@@ -435,6 +456,7 @@ async def chat(request: MessageRequest):
         "timestamp": datetime.datetime.now().isoformat()
     }
     chat_history.append(message_data)
+    logger.info(f"Chat response: {response}")
     return {"response": response}
 
 @app.get("/")