Why we choose to limit user amount?

Not sure how you know the existence of this API platform, but actually we have already ran for about 5 months! Okay, so let's go straight to the point, why do we choose to limit users instead of trying to scale?

Resource is limited

We are not OpenAI or any big company (note that this is a hobby project tbh), at first, I just want to share some spare resources to those we don't know how to find. In the beginning, the amount of users grow slowly, like about 1 people a day. But however one day, about 100 people registered the platform in a few hours, we realized that if this keeps going we will run out of resources in a short time, and this is not we want.

The form system

We start the form registration system from about 300 users, and it do control the volume.

From the above chart, you can see how many active users are in the platform. There is a big leap at about 01.15, and that's the time we got about 100 users in a few hours mentioned above. Also, at about March 13th, we got a huge leap of acitve users again. And this time, we implemented the form system. There is always a challenge for us to balance to users and the resources, and we will keep trying to improve the experience for our platform.

Note: Users who do multiple success requests in a few hour will count as Active users in the chart.

Peak usage

Throttling may be a thing, and we definitely need to keep them in control.

The peak usage is around 500 requests recently, and for most of the time the system is doing fine (our virtual machine has 2CPU+12GB RAM). This research is just to make sure throttling is in place to avoid overload.

Note: Usage in time interval means the total success request in an hour.

Premium Access

As known as Aqua Current Plan, it gives you access to high tier models such as GPT 5.5, Claude Opus 4.6, Gemini 3.1 Pro, and more. However, it is not a small cost on our side, so controlling these models' usage is the most important of all mentioned above. According to our data, 109 users have access currently to premium text models, that is about 30% of all users.

Conclusion

Most platform try to scale, but force to end up adding paid plans, and shrink free quota. We want to provide a long term AI access platform to users, therefore we choose the oppostie side: limit the amount of users. This is necessary if we still want to give free access without paywalls.