Version: Latest
Chitchat and Small Talk
This guide shows you how you can have your assistant handle chitchat / smalltalk.
Enabling Free-form chitchat
The easiest way to enable chitchat is to allow free-form generated responses from an LLM.
To do this, add a flow with the name pattern_chitchat
to override the default chitchat pattern.
Use the utter_free_chitchat_response
action:
flows.yml
flows:
pattern_chitchat:
description: handle interactions with the user that are not task-oriented
name: pattern chitchat
steps:
- action: utter_free_chitchat_response
And a custom prompt as part of the utter_free_chitchat_response
in your domain:
domain.yml
utter_free_chitchat_response:
- text: ""
metadata:
rephrase: True
rephrase_prompt: |
The following is a conversation with
an AI assistant. The assistant is helpful, creative, clever, and very friendly.
The user is making small talk, and the assistant should respond, keeping things light.
Context / previous conversation with the user:
{{history}}
{{current_input}}
Suggested AI Response:
Make sure you have the response rephraser set up with an appropriate LLM in your endpoints file. For example:
endpoints.yml
nlg:
type: rephrase
llm:
model: gpt-4o
And ensure that you have set an API key via an environment variable. For example, if you're using an OpenAI model,
- Linux/MacOS
- Windows
export OPENAI_API_KEY="your-openai-api-key"