import os import json from django.http import JsonResponse from django.views.decorators.csrf import csrf_exempt from langchain_community.document_loaders import PyPDFLoader, CSVLoader, UnstructuredExcelLoader from langchain.text_splitter import RecursiveCharacterTextSplitter from langchain_community.vectorstores import FAISS from langchain_community.chat_models import ChatOpenAI from langchain.memory import ConversationBufferMemory from langchain.chains import ConversationalRetrievalChain from langchain_huggingface import HuggingFaceEmbeddings from dotenv import load_dotenv from user_panel.models import ChatBot, UserLeads, EndUser from .models import Conversation load_dotenv() embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2") VECTORSTORES = {} CONVERSATIONS = {} GREETINGS = { "hi": "Hello! How can I help you today?", "hello": "Hi there! How are you doing?", "hey": "Hey! What can I do for you?", "how are you": "Iā€™m just a bot, but Iā€™m doing great! šŸ˜Š", "bye": "Goodbye! Take care!", "thank you": "You're welcome!", "thanks": "Glad I could help!" } INTENTS = { "order": "LEAD_CAPTURE", "menu": "Would you like to see the menu?", "book": "LEAD_CAPTURE", "booking": "LEAD_CAPTURE", "reservation": "LEAD_CAPTURE", "recommend": "Pasta and desserts are highly recommended!" } def save_lead_from_request(data, chatbot_id): """Save lead information to database""" try: chatbot = ChatBot.objects.filter(id=chatbot_id).first() if not chatbot: return {"error": "Chatbot not found."} user = chatbot.user lead = UserLeads.objects.create( name=data.get("name"), email=data.get("email"), phone_number=data.get("phone_number"), lead=data.get("lead", ""), note=data.get("note", ""), address=data.get("address", ""), type=data.get("type", "general"), status="new", chatbot=chatbot # Add chatbot foreign key ) return {"success": True, "message": "šŸ“ Lead saved successfully! We'll contact you soon.", "lead_id": lead.id} except Exception as e: return {"error": f"Failed to save lead: {str(e)}"} def get_next_lead_question(chat_history, inquiry_type): """Determine what information to ask for next in the lead capture flow""" # Extract collected information from chat history collected_info = {} for msg in chat_history: if msg.get("user"): user_msg = msg["user"].lower() bot_msg = msg.get("bot", "") # Check if user provided name if "what's your name" in bot_msg or "your name" in bot_msg: if len(user_msg.split()) >= 2: # Assuming name has at least 2 words collected_info["name"] = msg["user"] # Check if user provided email elif "email" in bot_msg and "@" in user_msg: collected_info["email"] = msg["user"] # Check if user provided phone elif "phone" in bot_msg and any(char.isdigit() for char in user_msg): collected_info["phone_number"] = msg["user"] # Check if user provided additional notes elif "special requirements" in bot_msg or "anything else" in bot_msg: collected_info["note"] = msg["user"] # Determine next question if not collected_info.get("name"): return f"Great! I'd be happy to help you with your {inquiry_type}. What's your full name?" elif not collected_info.get("email"): return f"Nice to meet you, {collected_info.get('name', 'there')}! What's your email address?" elif not collected_info.get("phone_number"): return "Perfect! What's your phone number?" elif not collected_info.get("note"): return f"Do you have any special requirements or notes for your {inquiry_type}?" else: return "LEAD_COMPLETE" # All information collected def save_lead_from_chat_history(chat_history, chatbot_id, inquiry_type): """Extract lead information from chat history and save it""" try: collected_info = {"type": inquiry_type} for msg in chat_history: if msg.get("user"): user_msg = msg["user"] bot_msg = msg.get("bot", "") # Extract name if "what's your name" in bot_msg or "your name" in bot_msg: if len(user_msg.split()) >= 2: collected_info["name"] = user_msg # Extract email elif "email" in bot_msg and "@" in user_msg: collected_info["email"] = user_msg # Extract phone elif "phone" in bot_msg and any(char.isdigit() for char in user_msg): collected_info["phone_number"] = user_msg # Extract notes elif "special requirements" in bot_msg or "anything else" in bot_msg: collected_info["note"] = user_msg # Save the lead return save_lead_from_request(collected_info, chatbot_id) except Exception as e: return {"error": f"Failed to save lead: {str(e)}"} def load_documents(user_id, chatbot_name): base_path = os.path.join("media", "data", str(user_id), chatbot_name) if not os.path.exists(base_path): return [] documents = [] for filename in os.listdir(base_path): full_path = os.path.join(base_path, filename) if filename.endswith(".pdf"): loader = PyPDFLoader(full_path) elif filename.endswith(".csv"): loader = CSVLoader(full_path) elif filename.endswith(".xlsx"): loader = UnstructuredExcelLoader(full_path) else: continue documents.extend(loader.load()) return documents def build_or_load_vectorstore(user_id, chatbot_name, load_existing=False): vectorstore_path = os.path.join("vectorstores", f"{user_id}_{chatbot_name}") os.makedirs("vectorstores", exist_ok=True) if load_existing and os.path.exists(vectorstore_path): return FAISS.load_local(vectorstore_path, embeddings, allow_dangerous_deserialization=True) docs = load_documents(user_id, chatbot_name) if not docs: return None splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200) split_docs = splitter.split_documents(docs) vectorstore = FAISS.from_documents(split_docs, embeddings) vectorstore.save_local(vectorstore_path) return vectorstore def get_or_create_qa(user_id, chatbot_name, load_existing=False): key = f"{user_id}_{chatbot_name}" if key not in VECTORSTORES: vs = build_or_load_vectorstore(user_id, chatbot_name, load_existing) if not vs: return None, None VECTORSTORES[key] = vs else: vs = VECTORSTORES[key] retriever = vs.as_retriever(search_kwargs={"k": 5}) llm = ChatOpenAI(model="gpt-4o-mini", temperature=0.3) memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True) qa = ConversationalRetrievalChain.from_llm(llm, retriever=retriever, memory=memory) return qa, memory # def save_lead_from_request(data): # required_fields = ["name", "email", "number", "inquiry_type"] # if not all(field in data for field in required_fields): # return {"error": "Missing one or more required lead fields."} # lead_data = { # "name": data["name"], # "email": data["email"], # "number": data["number"], # "inquiry_type": data["inquiry_type"], # } # os.makedirs("data", exist_ok=True) # lead_file = "data/leads.json" # if os.path.exists(lead_file): # with open(lead_file, "r") as f: # leads = json.load(f) # else: # leads = [] # leads.append(lead_data) # with open(lead_file, "w") as f: # json.dump(leads, f, indent=4) # return {"success": True, "message": "šŸ“ Lead saved successfully!"} @csrf_exempt def chat_with_bot(request, chatbot_id): if request.method != "POST": return JsonResponse({"error": "POST method required"}, status=400) try: try: data = json.loads(request.body) except json.JSONDecodeError: return JsonResponse({"error": "Invalid JSON body."}, status=400) query = data.get("query", "").strip() load_existing = data.get("load_existing", True) chatbot = ChatBot.objects.filter(id=chatbot_id).first() if not chatbot: return JsonResponse({"error": "Chatbot not found for this user."}, status=404) user_id = chatbot.user.id # Fixed: access user.id instead of user_id if not user_id or not query: return JsonResponse({"error": "Query is required in the request body."}, status=400) chatbot_name = chatbot.chatbot_name key = f"{user_id}_{chatbot_name}" global CONVERSATIONS if key not in CONVERSATIONS: qa, memory = get_or_create_qa(user_id, chatbot_name, load_existing) if not qa: return JsonResponse({"error": "No documents found for this chatbot."}, status=500) CONVERSATIONS[key] = {"qa": qa, "memory": memory, "chat_history": []} else: qa = CONVERSATIONS[key]["qa"] normalized_query = query.lower() if normalized_query in GREETINGS: response = GREETINGS[normalized_query] else: intent_match = next((v for k, v in INTENTS.items() if k in normalized_query), None) if intent_match: if intent_match == "LEAD_CAPTURE": # Check if this is a lead submission with data if data.get("name") and data.get("email") and data.get("phone_number"): # This is a lead submission lead_result = save_lead_from_request(data, chatbot_id) if lead_result.get("success"): response = lead_result["message"] else: response = f"āŒ {lead_result.get('error', 'Failed to save lead')}" else: # This is an initial order/booking request - prompt for lead info step by step inquiry_type = "order" if any(word in normalized_query for word in ["order", "food", "dish"]) else "booking" response = ( f"Great! I'd be happy to help you with your {inquiry_type}. " "Let me collect your information step by step.\n\n" "**Step 1:** What's your full name?\n" "(Please provide your name in the next message)" ) else: response = intent_match else: response = qa.run(query) if response.lower() in ["i don't know", "i am not sure"]: response = "āš ļø Sorry, I could not find that information in your documents." CONVERSATIONS[key]["chat_history"].append({"user": query, "bot": response}) chat_history = CONVERSATIONS[key]["chat_history"] conversation_obj, created = Conversation.objects.get_or_create( user_id=user_id, domain=chatbot_name, defaults={"chat_history": chat_history} ) if not created: conversation_obj.chat_history = chat_history conversation_obj.save() return JsonResponse({"response": response}) except Exception as e: return JsonResponse({"error": f"Unexpected error: {str(e)}"}, status=500)