New and Improved… Things!
Release Notes, Version 23.4.4 — grow up and fix things.
Our engineering organization is leveling up! We‘re shifting from “semi-stealth” to “building in public”, and so we’re excited to start sharing our weekly releases here, on the #blog. We’ve got a few cool new elements to share with you, and away… we… GO
Admin UI
As an API-first company, we build features first that deliver value to our clients immediately, then display those features through the UI so even non-technical teammates can play with them. From here, you can set policies to do all the things:
- Block or Redact
-PII
-Passwords
-Keywords
-Our Pre-set Policies (like toxicity, adult content, etc.)
-Whatever you want!
Organizations & Group Policy Management
We all know how complicated big companies can be, but with a little ontology we can bulk apply a policy once, and make sure it’s applied to the right people. Through our new UI, you can now also set moderation and model access policies to organizations (like Marketing, for example), and then groups under that organization (like your Content Marketers, or Video team, as examples). By way of example…
Marketing
Content Marketers
Block: Competitor Mentions, Toxicity, Adult Content, Passwords, PII
Allow Access to: LlaMa, AI21, GPT-4
Allow Data Type: Text Only
Video Team
Block: Competitor Mentions, Toxicity, Adult Content, Passwords, PII
Allow Access to: LlaMa, AI21, GPT-4, Dall-E, Stable Diffusion
Allow Data Type: Text, Video, Image
Improved Model Routing!
We’re also working every day to make our latest feature more sophisticated. In this release, we…
Added a bunch of new metrics! We all know how challenging LLM evaluation can be, so we don’t expect clients to accept our optimization practices out of the box; they’re customizeable based on your preferences, and can be adjusted at the Organizations and Groups levels, too. These new metrics help the model router hone in on the most accurate, lowest-cost, lowest-latency model for the task at hand, including specific tuning for tasks like summarization, translation, and many others.
If you’d like to hear more about these improvements, or our model routing demo… get in touch! And don’t forget to join our waitlist while you’re here!
