Coming Soon : This example demonstrates Nadoo Builder capabilities, currently available in enterprise preview.
Overview Learn how to build a production-ready chatbot using Nadoo Builder’s visual interface. This example covers conversation management, intent detection, and knowledge base integration.
Chatbot Architecture Visual Workflow Design Configure the user input node in Builder:
node_type : UserInput configuration : name : "User Message" input_type : text validation : required : true max_length : 1000 metadata : track_user_id : true timestamp : true
Step 2: Intent Classification Add intent classification node:
node_type : LLMNode configuration : name : "Intent Classifier" model : gpt-3.5-turbo temperature : 0.1 system_prompt : | Classify user intent into one of: - faq: Frequently asked questions - support: Technical support request - sales: Sales inquiry - general: General conversation Respond with only the intent category. output_format : text
Step 3: Conditional Routing Create branches based on intent:
node_type : Conditional configuration : name : "Route by Intent" conditions : - condition : "intent == 'faq'" next_node : "knowledge_base" - condition : "intent == 'support'" next_node : "create_ticket" - condition : "intent == 'sales'" next_node : "sales_handler" - condition : "intent == 'general'" next_node : "general_chat" default : "general_chat"
Step 4: Knowledge Base Search Configure knowledge base retrieval:
node_type : VectorSearch configuration : name : "FAQ Search" index_name : "faq_index" top_k : 3 similarity_threshold : 0.7 fields : - question - answer - category
Step 5: Response Generation Generate final response:
node_type : LLMNode configuration : name : "Response Generator" model : gpt-4 temperature : 0.7 system_prompt : | You are a helpful customer service assistant. Use the provided context to answer questions. Be friendly and concise. context_sources : - knowledge_base - conversation_history max_tokens : 500
Code Equivalent The visual workflow compiles to this Flow Core code:
from nadoo_flow import ( ChainableNode, LLMNode, ConditionalNode, VectorSearchNode, FunctionNode ) class ChatbotWorkflow : def __init__ ( self ): # Intent classification self .intent_classifier = LLMNode( model = "gpt-3.5-turbo" , temperature = 0.1 , system_prompt = """Classify user intent into: - faq, support, sales, general""" ) # Knowledge base self .kb_search = VectorSearchNode( index_name = "faq_index" , top_k = 3 ) # Response generation self .response_gen = LLMNode( model = "gpt-4" , temperature = 0.7 , system_prompt = "Helpful customer service assistant" ) async def process_message ( self , user_input : str , user_id : str ): # Classify intent intent_result = await self .intent_classifier.execute({ "message" : user_input }) intent = intent_result[ "content" ].strip().lower() # Route based on intent if intent == "faq" : # Search knowledge base kb_results = await self .kb_search.execute({ "query" : user_input }) # Generate response with context response = await self .response_gen.execute({ "user_message" : user_input, "context" : kb_results[ "results" ] }) elif intent == "support" : # Create support ticket response = await self ._create_support_ticket(user_input, user_id) elif intent == "sales" : # Route to sales response = await self ._handle_sales_inquiry(user_input, user_id) else : # General conversation response = await self .response_gen.execute({ "message" : user_input }) return { "response" : response[ "content" ], "intent" : intent, "user_id" : user_id } async def _create_support_ticket ( self , message : str , user_id : str ): # Ticket creation logic return { "content" : "Support ticket created. We'll be in touch soon." } async def _handle_sales_inquiry ( self , message : str , user_id : str ): # Sales routing logic return { "content" : "Connecting you with our sales team..." }
Advanced Features Conversation Memory Add context awareness:
node_type : ConversationMemory configuration : name : "Chat History" max_turns : 10 include_metadata : true summary_threshold : 20 storage : type : redis ttl : 3600
Generated code:
from nadoo_flow import ConversationMemoryNode memory = ConversationMemoryNode( max_turns = 10 , storage_backend = "redis" ) # Store conversation await memory.add_turn( user_id = "user123" , role = "user" , content = user_input ) # Get context history = await memory.get_history( user_id = "user123" )
Multi-Language Support Add translation nodes:
node_type : LanguageDetector configuration : name : "Detect Language" supported_languages : - en - ko - ja - zh default : en --- node_type : Translator configuration : name : "Translate Input" target_language : en provider : google_translate --- node_type : Translator configuration : name : "Translate Output" source_language : en target_language : auto # Detect from input provider : google_translate
Sentiment Analysis Add emotion detection:
node_type : SentimentAnalysis configuration : name : "Emotion Detector" model : sentiment-analysis-v1 output_format : structured fields : - sentiment : [ positive , negative , neutral ] - emotion : [ joy , anger , sadness , fear , surprise ] - confidence : float actions : - condition : "sentiment == 'negative' and confidence > 0.8" action : "escalate_to_human"
Escalation Logic Handle complex cases:
node_type : EscalationHandler configuration : name : "Human Escalation" triggers : - negative_sentiment : 0.8 - unresolved_after : 3 - explicit_request : true notification : channel : slack webhook : "https://hooks.slack.com/..." handoff_message : "Connecting you with a human agent..."
Integration Examples Slack Integration node_type : SlackIntegration configuration : name : "Slack Connector" bot_token : "${SLACK_BOT_TOKEN}" channels : - general - support event_types : - message - app_mention response_mode : thread
WhatsApp Business node_type : WhatsAppBusiness configuration : name : "WhatsApp Bot" phone_number_id : "${WHATSAPP_PHONE_ID}" access_token : "${WHATSAPP_TOKEN}" features : - text_messages : true - media_messages : true - interactive_buttons : true - templates : true
node_type : WebChatWidget configuration : name : "Website Chat" appearance : theme : light primary_color : "#4fc3f7" position : bottom-right features : - file_upload : true - typing_indicator : true - read_receipts : true - emoji_support : true rate_limiting : messages_per_minute : 10
Testing the Chatbot Test Cases in Builder test_scenarios : - name : "FAQ Query" input : message : "What are your business hours?" user_id : "test_user_1" expected : intent : faq response_contains : "hours" - name : "Support Request" input : message : "I can't log into my account" user_id : "test_user_2" expected : intent : support ticket_created : true - name : "General Chat" input : message : "Hello, how are you?" user_id : "test_user_3" expected : intent : general sentiment : positive
monitoring : metrics : - response_time_p95 - intent_classification_accuracy - user_satisfaction_score - escalation_rate alerts : - condition : "response_time_p95 > 2000" severity : warning - condition : "escalation_rate > 0.3" severity : critical dashboards : - chatbot_performance - user_engagement
Deployment Production Configuration deployment : environment : production scaling : min_instances : 2 max_instances : 10 target_cpu : 70 resources : memory : 2Gi cpu : 1000m rate_limiting : requests_per_second : 100 burst : 200
API Endpoint Once deployed, your chatbot is available at:
POST https://nadoo.ai/api/v1/chat/completions { "messages" : [ { "role" : "user", "content": "Hello"} ], "user_id" : "user123", "stream" : false }
Response:
{ "id" : "chat-abc123" , "choices" : [{ "message" : { "role" : "assistant" , "content" : "Hello! How can I help you today?" }, "intent" : "general" , "confidence" : 0.95 }], "metadata" : { "response_time_ms" : 247 , "nodes_executed" : 5 } }
Best Practices
Keep intents focused and distinct
Provide clear training examples
Use fallback intents for unclear cases
Regularly review misclassifications
Write clear system prompts
Use knowledge base for consistency
Implement tone guidelines
A/B test different prompts
Provide quick acknowledgments
Use typing indicators
Handle errors gracefully
Offer clear escalation paths
Monitor conversation metrics
Update knowledge base regularly
Collect user feedback
Version control workflows
Next Steps 
