Template Patterns

Initiated Chat

Describes a conversational pattern where an incoming customer message starts an AI-assisted workflow in Stubber.

This is the most common pattern in Stubber.
For any customer enquiries, this is the pattern that templates would likely follow. The pattern is defined by :

• an incoming message received from an external party
• an AI Agent responding and handling some request over one or many states

# Examples

• Sales agent on a website, that responds to sales inquiries
• Customer support via AI over email
• Billing queries initiated over Whatsapp by customers

# Detailed Structure

• The incoming notification platform will be configured to initiate a feedback action on a _create context stub if no stubsession if present
• The _create context stub would define the feedback action with a create_stub task that passes the notification platform message data into a new stub following the default context
• The newly created default context stub will run the _create lifecycle action and have tasks to detect the various incoming notification data
• If the notification data for a particular notification platform is detected, the _create action will run a send_actions task and replay the feedback action on the default stub
• The _create action would also initialize an AI Agent with a system prompt message
• The feedback action would then run on the default stub and trigger gpt_chat_task that will submit the chat to an LLM
• The LLM would respond with a feedback action that would most likely trigger a notification to respond to the incoming message
• This would create a stubsession to the default stub and all subsequent replies from the user would be handled by the default stub
• The LLM Agent will call AI Enabled actions and correspond with the user on the stub until the stubsession times out
• A default stub in this case would represent a single query from a user