Skip to main content

Product Updates – January 2025

LLMs and improved quality models. 🤩​

  • DeepSeek is HERE! And guess what? Everyone can try it—R1 and V3 included. No one gets left behind!
  • REMi just got a power-up! Now running on a faster version, meaning metrics are processed quicker, evaluations come sooner, and the model is sharper than ever. More accuracy, less waiting—who doesn’t love that? Check out the official announcement here!

Pushing the RAG Agents to a new level 🚀​

Data Augmentation Graph RAG Agent – We unveiled this beast in a webinar last week, and let’s just say… minds were blown. 🤯 You’ll soon be able to build your own custom graph, expanding knowledge connections like never before. Want a sneak peek? Watch it all here!
Stay tuned—the agent goes live in just a couple of weeks!

For Developers and API Lovers ⚒️​

New Indexes, Who Dis?​

NucliaDB just got a major upgrade! Our latest index now runs on a new architecture that separates data storage from indexing and searching. Translation? Faster scaling, better performance, and ultra-smooth searches even under heavy loads. Get all the nerdy details in our latest blog post.

New Status Codes – Because Details Matter​

Our /ask endpoint now returns:
✅ 0: Success! 🎉
❌ 1: Error 😢
🤷 -2: Not enough context for the LLM to answer
🕵️ -3 (NEW!): No data was found in NucliaDB, so the prediction call wasn’t triggered

Previously, -2 and -3 were lumped together, but now you can tell when there’s missing data vs. when the LLM just doesn’t know the answer. Super useful for reviewing activity logs!

SDK Updates:​

Because who doesn’t love more control? Now we support Personal Access Tokens to get auth key for automated scripts that perform admin tasks (like collecting usage logs).

On the UI Front 💅​

✨ Dashboard glow-up – Now with detailed /chat and /find counts
🎯 Label filtering just got smarter – AND/OR support for even more flexibility
📊 Account dashboards leveled up – Now showing previous months' consumption history