🥳 The Google Cloud Next video is live! It was awesome recording this one with Brian Kaufman!
💡 The video below is the section where I demo Weaviate in combination with Gemini (it's an e-commerce use case), and 𝘢 𝘭𝘰𝘵 ( #gemini , #vectorsearch , #weaviate , #RAG ,…

Prompt engineering is a key part of getting better responses from Large Language Models. The particular wording of the instructions and examples can determine the fate of the performance 😵‍💫

DSPy Compilers simplify the process of writing the perfect prompt! In this blog post,…

⌘ R, ⌘ R+, and Rerank 3 just became available on Sagemaker! Give it a whirl 💪

You can find them here 👇
aws.amazon.com/marketplace/pp…
aws.amazon.com/marketplace/pp…
aws.amazon.com/marketplace/pp…

Did you know about the Binoculars technique to reliably tell apart fake, LLM-generated, text from human-written text? Or have you heard about Modular RAG? What about Matryoshka Embeddings?

These are just a few posts in our new paper reviews page, where we create 1-2 minute…

Trying Weaviate • vector database and DSPy with WizardLM (of course power by ollama)

lets see if wizardlm2 7b can make structured output Connor Shorten

Exciting news for Generative Feedback Loops! ♻️🔥🚀

weaviate.io/blog/generativ…

One key concept to successfully bring your vector database to production is vector quantization.

It reduces deployment costs & increases search speeds with only small performance losses.

There are 3 main vector quantization techniques:
- Binary
- Scalar (int8)
- Product

Organizations around the world are building the next generation of business software with Weaviate. Today, we’re excited to share the story of Neople, a Netherlands-based company using generative AI to create digital co-workers that help businesses achieve their customer service…

Cohere keeps shipping!📦Introducing Compass, our new embedding model!

What's inside:
> works with multi-aspect data like emails, invoices, support tickets, log messages, tabular data
> converts data to JSON to avoid context loss

Compass is in private beta, sign up to try it…

Well, the most important batch task of all is optimizing DSPy programs. Cool, we’ll add infra to leverage this. IMO 50% is not sufficient cut, but we’ll take it.