Objects
ActionResponse
ActionResponse is an object you receive in response to your action request. You can read more about the ActionRequest object here.
Conversational
name
string
Required
The conversational action response is the simplest action response; it indicates that the agent has successfully generated a new response for the conversation.
Generated by action request:
conversational
{
"name": "conversational"
}
Human Takeover
name
string
Required
The human takeover response indicates that the agent has detected a request for human intervention and has responded according to its human takeover settings.
Generated by action request:
conversational
{
"name": "human-takeover"
}
Offense
name
string
Required
The offense response indicates that our systems have detected offensive language in the user's message.
The agent's response is adjusted to inform the user that it cannot address queries containing offensive content.
The agent's response is adjusted to inform the user that it cannot address queries containing offensive content.
Generated by action request:
conversational
{
"name": "offense"
}
Unrelated Query
name
string
Required
The unrelated query response indicates that our systems have detected the user's message as unrelated to the business.
The agent's response is adjusted to inform the user that it cannot address the query.
The agent's response is adjusted to inform the user that it cannot address the query.
Generated by action request:
conversational
{
"name": "unrelated-query"
}
Token Limiter
name
string
Required
The token limiter response indicates that the agent has exceeded the allowed token threshold.
The agent's response is adjusted to the fallback message as set in the agent's token limiter settings.
The agent's response is adjusted to the fallback message as set in the agent's token limiter settings.
Generated by action request:
conversational
{
"name": "token-limiter"
}
Hallucination
name
string
Required
The hallucination response suggests that our systems have identified a potential hallucination by the agent.
The agent's response is revised to request the user for more details regarding their query.
The agent's response is revised to request the user for more details regarding their query.
Generated by action request:
conversational
{
"name": "hallucination"
}
Continue
name
string
Required
The continue action response indicates that the agent has successfully continued from its latest message in the conversation.
Generated by action request:
continue
{
"name": "continue"
}