Skip to main contentPrompt
This is the prompt for the TTT model of the voice stack.
This generates the output of the voice AI node that will be converted into speech, audio.
This is where you can add tool providers to the LLM.
This gives the LLM the option to call a tool (or tools) during its execution.
Response function
This is required for every voice AI node.
This is a tool that is available to the LLM.
When the response function is called, the LLM moves to the next node in the pipeline.
Another way of saying this: The LLM will not move to the next node in the pipeline UNTIL the response function is called.
The response function needs a name and description which indicates to the LLM when to call the function and move to the next node.
Sometimes some prompt instructions are helpful to ensure the LLM calls the response function at the correct time.
Example:
Name: end-call
Description: Call this tool when the user wants to end the call.
These are variables the LLM will generate before it moves to the next node.
These become pipeline variables which can be referenced elsewhere in the pipeline.
Advanced options
We have an IVR-type navigation system for voice AI nodes.
The purpose of this menu is to navigate between nodes that are not connected by edges.
e.g. say you go down one path, and the conversation requires the AI to switch paths, you’d use the IVR menu to switch path.
Nagivate to: Check this toggle to enable other AI Voice nodes to navigate to the current node through the IVR menu.
Name: Provide a descriptive name
Description: Provide a description.
This will tell the other AI voice node when to navigate to the current node.
Navigate from: Check this box if you want the current AI node to have the ability to navigate to other AI nodes using the IVR menu.
Voice Behaviour
- Allow interruptions: When true, the caller can cut off and interrupt the AI mid sentence.
- Disable repeat on silence: When false, the AI will repeat the last message it outputted if the user does not speak for 10s.
- Should loop: When true the AI will loop until the response function ‘goal’ is met.
- Wait for speech: When true, the AI will wait for the user to speak before outputting its first message. Expected behaviour is true on outbound calls, and false on inbound calls.
