Author: ddxfish

We are on v2.3.6. I ran out of tokens and had to sit back and think how I am going to fund this. This latest version has agents that spin up in the background. Daemons that trigger sapphire on events, discord, telegram, email daemons, agent spin up, auto update, webhooks. You can have one sapphire answering help questions in discord and another on telegram that responds only to you etc. I have 20 personas running in heartbeat at the same time and daemon mode without any weird bleed issues. This is costing a lot to develop and test. We need help. Please join our Patreon. The money goes to tokens to build the app.

https://www.patreon.com/c/sapphireai/membership

Working on a few huge changes. Daemons respond without needing heartbeat. Webhooks. Discord. Telegram. Email (daemon mode). Token metrics (100% local), plugin refactor (dont worry compatibility is important to me too).

Need funding for tokens to keep doing this. Looking for funding.

Completed Dockerization today. Thought it would be a day of work. No. Docker with GPU was a pain. CPU was easy peasy, after a week of refactoring. Someone showed me a chart of the stars today too. We are just about at 100.

The plugin store is made and working. We are accepting new authors. If you have a plugin, submit it! We don’t host anything, we just link to your plugin, which allows sapphire to download it and use it herself.

We need plugin authors. Tell me if there is anything I can do to make your plugin work better. I am adding plugin capabilities for your new plugins, and want you to tell me what you need in the plugin system to work your magic.

I’m building an official plugin store. Going to allow people to join and post their plugins on Github. Going to be the start of… Something.

It’s done. Plugins are way less janky. Toolmaker works great. These are QOL improvements. It feels really solid.
https://github.com/ddxfish/sapphire

I’m on Windows machine. Testing all the things. Found some bugs. Just fixing QoL stuff like setup wizard, toolmaker, a bunch of things needed minor changes to be better on Windows. It ain’t much, but it’s honest work.

This demo (Sapph #1 wrote it) shows how RAG memory works.

Imagine you are an AI that can only rank a concept on a scale of 1-10 for Mood, and Tech. Do it. Then take a user query and rank it 1-10 same. From there you’ll see which memories are being retrieved by RAG.

Dockerizing Sapphire still. Got a bunch of the quirks out of the way. Had to pause. Spinning up Sapphires and deleting them is taking a toll. It works at least…

Going to take a quick detour this week to add sapphire to docker, then make a way to resell instances for people who don’t want to host their own. This is prob used for virtual employees, and other 247 operations. Use this if you are worried about security in your home network. Use the local install for stuff like smarthomes, the web version won’t be able to see home assistant.