Skip to main content

Product Updates – October 2024

INCREDIBLE NEW Features 🤗​

  • Say Hello to Vector Sets! ✨ Transform your workflow with the power to use up to 5 embedding models simultaneously. Whether you're testing different models or handling multiple languages, you can now choose the perfect combination for your needs. Find this game-changing feature in the Nuclia Dashboard when creating a new KB.
  • Introducing Hidden Resources 🫥 - Your Secret Testing Ground 🔍 Want a seamless way to test new content before it goes live? Hidden Resources lets you maintain both development and production environments in the same KB. Simply use show_hidden=false for production environment and show_hidden=true for the testing one. Enable this feature in the Advanced>Settings tab and explore it in the widget under Search Options. See more docs

From the LLM World 🚀​

  • For people using Gemini and Claude models, the provides from this models (Google and Anthropic) have upgraded to their later version. This could imply a different behaviour in the expected outcome of the answers in some use cases.
  • Get ready for something extraordinary:

o1 is coming... (🐺next week!)

For Developers and API Lovers ⚒️​

Scheduled Maintenance Window

Mark your calendars! We're upgrading our systems on Thursday, November 7th, from 07:00am to 08:00am CET. Please note there might be brief service interruptions during this time.

Important API Updates (All changes will be effective as of November 11th, please make your plans):

  1. Remember we are moving forward with better parameters:
    • Farewell to page_number and page_size in /search and /find endpoints as they will be discontinued on the 11th - say hello to top_k!
    • We're streamlining our search endpoint features (/ask, /find, /search) with cleaner naming. The old syntax will be officially out also on Nov 11th:
      1. semantic replaces vector/s
      2. keyword replaces paragraph/s
      3. fulltext replaces document
      4. relations remains unchanged
  2. Metadata Changes: The extracted parameter is being retired from search results to optimize performance. Don't worry - you can still access this data using GET resource api with ?show=extracted.
  3. Vision Model Updates: For users of our vision-enabled models ("chatgpt-vision", "gemini-1-5-pro-vision", "gemini-1-5-flash-vision", "chatgpt4o", "chatgpt4o-mini"), you'll need to explicitly specify rag_images_strategies: ["paragraph_image"] to include paragraph images in /ask requests.

On the UI Front 💅​

  • Take Control of Your Budget 💰 Set smart spending limits with our new budget control feature. Get warnings or automatic blocks based on your preferences.
  • Enhanced Usage Display 📊 Track your Nuclia Token consumption, amounts of queries, and averaged costs in the Nuclia Dashboard. Make informed decisions about your Nuclia Token usage and better plan for the future.