September 2024 - Feature Recap 🚢
We've launched a lot at lytix over the past few weeks and wanted to take the time to recap what we're most excited about 🚀
⚡ Evaluation Quickstart
Custom evaluation models are great for catching errors that need deep product context to catch. But creating a long and detailed rubric has been a blocker for our users to get started building their own models.
So we shipped Evaluation Quickstart ⚡ You can now use natural language to describe the task, the evaluation you'd like and optionally bring in examples.
Lytix will then generate a net-new model for you - optimized with the latest prompt engineering techniques, using the description and any examples provided.
We hope this will allow teams to setup more evaluations that easier to customize 🚀
📓 Documentation can be found here (opens in a new tab)
💽 Metadata
Your data should be easy to debug, and part of that is isolating data you care about. Part of that can be accomplished via workflows (opens in a new tab) but sometimes you need even more specificity.
We’ve introduced metadata (opens in a new tab) to solve this problem. Add custom metadata such as environment tags and see that reflected in the dashboard.
You can then filter based on metadata tags along with model + workflow to really triage and streamline your observability process.
📓 Documentation can be found here (opens in a new tab)
∫ Integrations
This month we’ve added even more integrations to make lytix easier to get started with!
- Vercel AI
- Anthropic or OpenAI Vercel AI clients can now create a custom client
- Gemini
- With a couple of lines you can quickly start proxying Gemini messages via lytix with minimal code change!
📓 Documentation can be found here (opens in a new tab) for Vercel, and here (opens in a new tab) for Gemini
🤖 New models
OpenAI launched their o1 line of models this month, we’ve added cost + playground support for both!
📓 Documentation can be found here (opens in a new tab)
🌐 Self Hosted Cache
We’ve added the ability to self host your own cache if you want more control over your data and whats being cached. We’ll soon be launching a hosted version, get in touch with founders@lytix.co to get a preview!
📓 Documentation can be found here (opens in a new tab)
✨ Speed + UI Redesign
We’ve redesigned not only our UI but also our internal database structure to scale better with larger amounts of data. We can now quickly filter metadata, models and workflows across 100,000+ events with ease
👨💻
📖 If you liked this, check out our other posts here (opens in a new tab).
📣 If you are building with AI/LLMs, please check out our project lytix (opens in a new tab). It's a observability + evaluation platform for all the things that happen in your LLM.
© lytix