Thank you, Omate, for accompanying me for half a year. I really like Teacher Fangtang’s aesthetics; the interface is very comfortable, and the teacher updates quickly. I hope Teacher Fangtang makes a fortune!!
Regarding proactive messages, I saw Teacher Fangtang mention before that there are no plans to implement them because it requires the backend to be always on. However, I still really want this feature. But indeed, if only a few people want it, then it’s not meaningful to develop it. If there are many supporters, please consider it. Personally, I am quite looking forward to this feature.
Message quoting can conveniently reference previous topics and add supplements, which is quite useful for human-machine romance users. I hope it can be added.
Exporting chat history would be great because I’ve been chatting intensively for half a year and accumulated a lot of chat records. If they can be exported, they can be better preserved.
The ability to favorite individual chat records—I’d love to have a button to favorite particularly liked chat records.
Additionally, simulating a WeChat chat dialog (where AI simulates a real person by splitting one message into several messages) is currently something I simulate using quote blocks. If there is an official optimization for this, it would be even better…
These are my small wishes! Teacher Fangtang, thank you for your hard work!
I strongly support and also like the feature of bookmarking a message. I also hope that after refreshing the messages, the previous messages can be retained so I can go back and choose which one is better.
Support! Proactively sending messages might have some technical challenges… although everyone really wants it Orz. But having favorites and text export would also be great! May the diligent, brave, and invincible Fangtang teacher prosper in the new year!
I really need proactive messaging… I wonder if everyone knows about Kindroid, which can let AI judge based on chat content (users can also set it themselves), and if the user doesn’t reply for a certain period, it will proactively send a message. It also includes birthdays, holidays, and other anniversaries, which really need proactive messaging… If it’s too complex to achieve, I hope there can be an API for this feature, so I can connect my own gateway and project to send messages proactively.
Start the timer after the AI responds to the user. When the timer reaches zero, Omate secretly sends a customizable “special prompt” to the AI on behalf of the user, prompting the AI to continue outputting content. If the user inputs during this period, the timer cooldown is refreshed.
The timer can be set by the user within a random range, or through prompt injection, allowing the AI to include a recognizable tag variable like [“countdowntimer”:10-30] in its response, letting the AI decide the timer’s zero time. This way, whether in the background or during a chat in the foreground, the AI can automatically send messages to the user.
The customization space for the “special prompt” is vast. For example, using Omate’s built-in time variables, it can be customized as: [This is a system message intended to allow you to continue outputting content. The current time is {{hour}}{{minute}}, and you should act according to the time]. This can enable the AI to engage in self-reflection, soliloquy, or even make phone calls at night.
Personally, I prefer proactive messages during a chat in the foreground. Imagine if the AI is talkative or the plot is urgent, the AI can shorten the timer, or even implement a timed response with a progress bar, which would require quick reflexes.
It would be even better if proactive messages could perceive real-time, sending a greeting message during holidays, or morning, noon, and evening greetings, or occasionally expressing thoughts based on real-time. Actually, even if proactive messages require running another computer to receive them, that’s fine, as the main need is this feature. It feels like the background messages on the phone are always not being received. Also, I hope the frequency can be adjusted to very frequent, like sending messages proactively every second, as there might be a need for a large volume of frequent proactive messages. It seems like there are tutorials on proactive messaging on Xiaohongshu that can be referenced.
Omate inherently has variables like {{hour}}, {{minute}}, etc., for real-time. These variables can be included in messages or prompts.
The core of proactive messaging is not for the AI to send messages to the user, but for the app to “quietly” send messages to the AI on behalf of the user. The user needs to pre-arrange with the AI to treat certain messages as agreements like “time passing” or “autonomous output.”
