Skip to main content

Product Updates – November 2024

INCREDIBLE NEW Features 🤗​

  • Monitor the Quality of Your RAG Stack with REMi
    Keeping your RAG stack in check just leveled up! Say hello to REMi, your new best friend for pro-level monitoring. Now you can:

    • Use a dedicated API to track and fine-tune answer quality.
    • Access detailed activity logs to stay on top of every query.
    • Dive into rich metrics that give you actionable insights into performance.
      Ready to get started? Check out the full article here for all the details.

  • Data Augmentation Agents Are Coming – Everybody is waiting for this and is here!—well, almost. Launching next week, they’re set to make your data preprocessing smarter and your results sharper— also your life will be easier and yes, it is a bold statement.

From the LLM World 🚀​

  • Haiku for All the Anthropic Lovers – Now in a wallet-friendly package. Who doesn’t love poetic AI at a better price? 💸

For Developers and API Lovers ⚒️​

Webinar Alert!
Topic: "Retrieval Strategies for a Better RAG Output"
📅 Thursday, December 5th | 🕛 12 PM ET / 6 PM CET
👉 Join the webinar here

API Updates:

  • Reranker Evolution: Say goodbye to the legacy multimatch_booster. The default is now noop for speedy retrieval, with predict offering premium results (at a performance cost).
    In more detail, the reranker feature was introduced some time ago and received great feedback. The _predict     reranker delivers significantly better results (at the cost of performance), while the noop reranker offers a fast experience without reranking. The legacy multi_match_booster reranker (currently the default) has become less useful and as of now deprecated and will be supported until February 2nd, 2025. Moving forward, the default will be noop for fast out-of-the-box retrieval, with predict available to deliver higher-quality results.

  • Rank Fusion Power-Up: Customize your ranking with new API parameters for /find and /ask endpoints. Predict reranker now supports a tunable result window, so you’re always in control.
    Let's get technical: Rank fusion is now a first-class feature in NucliaDB. A new API parameter has been added to the /find and /ask endpoints, enabling users to customize rank fusion. Additionally, the predict reranker now includes a new parameter to adjust the window of results to rerank.

On the UI Front 💅​

  • Your RAG lab questions are now auto-saved, so you can pick up where you left off anytime.
  • User feedback on specific results? Check!
  • Fancy nested dropdowns for label filters—because why not add a little elegance to your workflow?