Action Nodes
End-call behavior, builtin functions, and pre/post agent actions.
Action capabilities in Talkif fall into three categories: the End Call node (the only standalone action node on the canvas), builtin functions configured on agent nodes, and pre/post actions attached to agent nodes.
End Call node
The only standalone action node type. Terminate the call when the conversation reaches this node. Optionally configure a goodbye message that plays before the call disconnects.
End Call also exists as a builtin function (end_call) on agent nodes, allowing the LLM to end the call mid-conversation without requiring a separate node.
Builtin functions
Builtin functions are configured as function definitions on agent nodes. The LLM can invoke them during a conversation turn.
| Function | Description |
|---|---|
end_call | End the call. The LLM triggers this when the conversation is complete. |
tts_say | Speak a specific message to the caller. Use for scripted lines the LLM decides to deliver. |
forward_call | Transfer the call to another phone number. Configure with a phoneNumber parameter. |
forward_call
Add forward_call as a function on any agent node. Configure with phoneNumber to transfer the call. Use cases include escalating to a human agent, routing to a department, or transferring to an on-call number.
Pre/post actions
Pre/post actions execute automatically when an agent node is entered (pre) or exited (post). They run without LLM involvement.
| Action type | Description |
|---|---|
tts_say | Speak a scripted message. Supports Handlebars template variables for dynamic content. |
play_audio | Play audio from a URL. |
wait | Pause for a specified duration in milliseconds. |
webhook | Send an HTTP request to an external service. Configure with url and method. |
Webhook configuration
| Setting | Description |
|---|---|
| URL | The endpoint to call |
| Method | HTTP method (GET, POST, PUT, PATCH, DELETE) |
Webhooks can also be configured as custom function execution targets on agent nodes, allowing the LLM to trigger them during conversation.