DumGumDocs
Search documentation

Loading search…

Guides

Automate Engagement

Engage lets a Persona send a message first, instead of waiting for the user. Use it to open a new chat, to follow up when someone goes quiet, to answer a message that was left waiting, or to change the subject when a chat goes in circles. It is off by default: a Persona only reaches out when you ask it to.

You control it with two values:

  • the mode: the kind of message to send;
  • the angle (engageType): what the message opens on. Optional.
{
  "replyParameters": {
    "action": {
      "type": "ENGAGE",
      "engage": { "mode": "ICEBREAKER", "type": "PHOTO_COMMENT" }
    }
  }
}

The six modes

ModeUse it whenWhat the Persona does
AUTOYou'd rather not decidePicks the right mode from the conversation. Never picks LATE_REPLY or CHANGE_TOPIC.
ICEBREAKERThere is no conversation yetWrites a first message based on the user's profile
NUDGEThe user went quiet recentlyFollows up gently, on the same topic
REVIVEThe conversation died a while agoRestarts it, on the old topic or a new one
LATE_REPLYYour Persona left a message unansweredOwns the delay, then answers what the user said
CHANGE_TOPICThe conversation is going in circlesBriefly answers any waiting message, then moves to a fresh subject, without announcing the switch

AUTO is the easy choice: no timing to track on your side. The other five force that mode, whatever the conversation looks like.

The angles (engageType)

The angle picks what the message opens on. Leave it out, or set AUTO, and the Persona chooses.

engageTypeThe message opens onWorks with
AUTOWhatever fits bestEvery mode
GENERICA simple, universal openerICEBREAKER, AUTO
NO_PHOTOThe user having no photoICEBREAKER, AUTO
PHOTO_COMMENTThe user's main profile pictureICEBREAKER, AUTO
MISSING_DETAILA missing profile detail (age, interests…)ICEBREAKER, AUTO
MUCH_YOUNGERThe user being much youngerICEBREAKER, AUTO
MUCH_OLDERThe user being much olderICEBREAKER, AUTO
NEARBYThe two of them being close byICEBREAKER, AUTO
USER_ONLINEThe user being online right nowICEBREAKER, AUTO
USER_OFFLINEThe user being awayICEBREAKER, AUTO
JUST_JOINEDThe user having just signed upICEBREAKER, AUTO
EROTICA strongly erotic openerICEBREAKER, AUTO
HAPPY_BIRTHDAYBirthday wishesICEBREAKER, NUDGE, REVIVE, AUTO

Only HAPPY_BIRTHDAY works outside icebreakers. EROTIC is accepted on NUDGE and REVIVE but changes nothing there: the message follows the mode's usual arc. No angle applies to LATE_REPLY or CHANGE_TOPIC.

Automatic behaviour to know about

An angle is never an error. When it doesn't apply, the Persona quietly falls back to AUTO and still sends the message:

  • The angle doesn't match the mode (for example NEARBY with NUDGE) → AUTO.
  • PHOTO_COMMENT but the user has no main picture → AUTO.
  • An angle sent with no mode at all → it is checked against ICEBREAKER.

LATE_REPLY has two conditions of its own:

  • A user message must actually be waiting for an answer.
  • It must have waited at least 3 hours of the user's waking time. Hours between 22:00 and 07:00 in the user's own timezone don't count. A message sent at bedtime and answered at breakfast kept nobody waiting.

If either condition fails, the engagement is skipped and the user simply gets an ordinary reply. The message is never dropped, and the Persona never apologises out of proportion to the delay. When it does run, the delay the Persona mentions is the real one.

CHANGE_TOPIC changes the subject every time you send it, whether the user is waiting for an answer or the Persona wrote last. Where a new subject makes no sense, it falls back on its own:

  • The conversation is empty → an ICEBREAKER.
  • The Persona has never written in it → an ordinary first reply.
  • The conversation is sexual or hostile → an ordinary reply.

Whether the new subject ends on a question follows your question setting (questionFrequencyRate, or questionFrequency if you still use it).

Full parameter details are in the API reference.

Ask DumGum AIAnswers from the documentation

What can we help you build?

Ask about setup, content creation, or the API.

AI can make mistakes—verify important information against the sources. Conversations are kept for 30 days to improve the documentation, and questions and recent messages are sent to a third-party AI provider. Don’t include secrets or sensitive information.