CommunityOne Update: Spark Microservice & Collab Revamp
Summary
The CommunityOne Builders Help community announced significant updates regarding their Spark microservice and Collab features. Spark is being refactored into a faster microservice, with infrastructure upgrades planned, and they are likely moving to better OpenAI models. Additionally, Collab is receiving UI overhauls and transitioning into a server listing for better cross-discovery, alongside enhanced user data privacy controls allowing one-click deletion of all stored data.
Hey guys! Update for this week:
- We are week 1 into Spark sprint. Here's a bit more details:
🇦 Spark will be repackaged as a microservice, this means that once we are done, you can have the same spark with differnt knowledge doc on different channels, you can test spark on the dashboard etc and spark will respond much faster vs right now.
It's a great thing after we paid off our technical debt.
🇧 The rest of the features will be a lot faster to build once we upgrade our infra. That would take about a week.
🇨 In terms of model selection, we haven't made the final decisions, but we are likely to upgrade everyone from the current OpenAI solution. There are simply better models right now with a better retrieval algorithm.
-
Collab: we are redoing a lot of UI and adding new features for Collab. See #┠🟣・bot-change-log on some of the UI changes that have been made. We will transition our collab feature into a "server listing" to allow you guys to cross discover and grow better.
-
In light of recent events where some bots are taken off by Discord due to TOS violation, we decided to do better at communicating our privacy policy. Here are some updates:
(a) At any point, you can click a button and delete all your data from our server. It is in your profile page on the dashboard. No need to email us, find us on discord. You are 100% in control of your data. (b) We don't have any proprietary AI model. We are a really good AI wrapper using the existing market solution. So we have never, and will never, use your data to train any AI model. Honestly, it is both expensive and meaningless to train your own custom AI for most companies (other than getting funding)
