Documentation›Chat, memory, and characters
18 Everyday use

When She Messages You First

Understand proactive messages, when they arrive, how she goes quiet by herself, and how to keep them welcome.

Local Waifu 1.8.0 macOS Windows
plain English first

Follow the bold button names and numbered steps. Technical names are included only when you need to recognize a model or file inside the app.

Local Waifu can start a conversation instead of only answering one. These are called proactive messages, and they are one of the clearest differences between a companion and a chat box.

The four triggers

TriggerWhen it fires
Morning greetingOnce a day, between 07:00 and 09:00 local time
End of work check-inOnce a day, between 17:00 and 18:00 local time
Evening wind-downOnce a day, between 22:00 and 23:00 local time
Long idleYou have been quiet for the active character’s idle threshold, four hours by default. Only considered between 10:00 and 22:00

A deliberately cold thread can also surface as a message about that topic instead of a generic “you there?”, so a callback sometimes arrives with a subject attached.

Earlier versions also pinged you before a calendar event. The calendar feature was removed in 1.8.0, so there is no calendar trigger and no calendar tab. Local Waifu does not read your calendar.

What keeps them from becoming noise

  • At most three proactive messages per hour, across every trigger.
  • One per day for each of the three clock-based triggers.
  • Held while you are typing. A nudge that arrives mid-message waits until the conversation goes quiet, and it carries a deadline past which it is dropped rather than delivered late. A morning greeting that could only be delivered at 23:00 is not delivered.
  • Held while she is quiet (see the next section).
  • Quiet hours per character. Each character learns when you tend to be asleep, and long-idle nudges respect that.
  • The idle threshold adapts to the character rather than being a fixed number for everyone.

She goes quiet by herself

Goal

Understand when she holds back everything you did not ask for, and control it.

While Quiet Mode is on, she holds back proactive messages, routine deliveries, speech bubbles, and idle motion. Replies to your own messages and anything you start yourself are never held. The sidebar shows a Quiet badge, and a held nudge looks again a few minutes later.

She goes quiet automatically in these situations:

SituationmacOSWindows
A fullscreen app is in front (a game, a film)YesYes
A meeting app is in front (Zoom, Teams, FaceTime, Webex, a Slack huddle, Google Meet or Teams in a browser tab)YesYes (Zoom, Teams, Webex, Slack huddles, and Meet in a browser)
The screen is lockedYesYes
The screen is being recordedYesNo
The Mac is asleepYesNo
An app you listed is in frontYesYes

Steps

  1. Open Settings → General.
  2. Find Go quiet during fullscreen, meetings and games. It is on by default.
  3. To add your own apps, type one per line under Also stay quiet when one of these apps is in front, for example Steam or Final Cut Pro.
  4. The line under the switch tells you whether she is quiet right now (Right now: quiet. or Right now: she can speak.).

Expected result

While one of those situations applies she stays silent, and Telegram messages from routines are held too on macOS. She speaks again when it ends.

If something goes wrong

  • If she stays quiet when she should not, check which app is in front. A fullscreen window of any app counts.
  • Turn the switch off to disable Quiet Mode entirely.
  • On macOS, reading a browser tab title for Meet or Teams needs Screen Recording permission. Without it the browser case may not be detected.

Turn proactive messages off, or tone them down

Settings → General → Let her message you first is a single switch: turned off, she sends no greetings, check-ins, or idle nudges and only answers when you write. Routines you set up still run. The finer controls are:

  1. Long idle: this fires only when the app is running and you have been quiet past the character’s threshold. Closing the app is a complete off-switch, because the trigger lives inside the app process.
  2. The learned behaviour: repeated dismissing, along with how you actually write, feeds the character’s adaptive dials, including its idle threshold and quiet hours (Chapter 20). A companion that has been left alone for a while becomes less forward on its own.
  3. During onboarding you chose how often she should reach out first. That choice only sets the starting point.

If something goes wrong

  • If a message arrives while you are mid-sentence, that is the deferral system failing its deadline, not a new setting. Note the time and report it.
  • If a proactive message seems to answer something you typed several minutes earlier, it most likely started generating before your message landed. The current generation is not cancelled by a new message, and this is a known limitation rather than a setting.
  • A good-morning message whose delivery failed is retried.

What a proactive message is not

It is not a notification service and it is not a scheduled task. Each one is a real generation from your active model, produced locally like any other reply, and it appears in the conversation as a normal message you can answer, edit, or ignore. For things she does on a schedule, see Chapter 25.