CShorten

Hey everyone! I am SUPER EXCITED to publish a new episode of the Weaviate Podcast with Amélie Chatelain and Antoine Chaffin on Multi-Vector Search!

Amélie, Antoine, and the LightOn team are on fire! They are making breakthrough after breakthrough in Search with Multi-Vector, Late Interaction retrieval models.

This podcast covers all sorts of topics, starting with the motivation of Multi-Vector Search to its particular successes in code with ColGrep, as well as reasoning-intensive and multimodal retrieval.

We also covered the cost of MaxSim and Multi-Vector Storage and how MUVERA and PLAID can help. If that wasn't enough, the podcast concludes with their new work on ColBERT-Zero and PyLate!

A lot of big takeaways from this one, I hope you find it useful!

YouTube: https://www.youtube.com/watch?v=44GC3E-WbHU

Spotify: https://spotifycreators-web.app.link/e/IQyBapFaK1b

In Hybrid Search, searchers such as BM25 and Vector Search run independently, produce separate result sets, and get merged...

Introducing ... ... Wormhole Vectors!

Instead of merging orthogonal results, traverse between the vector spaces. Run a lexical search, pool the matching documents' dense vectors, and search the dense space from that point.

Hybrid search runs two queries and hopes the right answer was in one of them. Wormhole vectors let you traverse between the spaces instead.

https://www.youtube.com/watch?v=j4JZvKb5pJY

In many software applications, Search is the core feature. The search bar is the universal interface for interacting with data!

So how should you think about designing user experiences around search so that your users are delighted to use your software?

What are the different types of search experiences you should be aware of as someone working in search engineering?

Also, whereas LLMs have typically responded to us with a wall of text, they can now respond with applications created from stacks such as HTML/CSS/JavaScript. How will this shape search experience design?

https://www.youtube.com/watch?v=zenU1lsjRL0

In this clip from the latest Weaviate Podcast, Doug Turnbull explains why simple lexical tools like grep work so well for agents: The transparent input-output relationship lets the agent reason about why results matched, adjust its strategy, and plan its next query.

https://x.com/weaviatepodcast/status/2028570908978262458

Hey everyone! I am super excited to publish a new episode of the Weaviate Podcast with Doug Turnbull and Trey Grainger on AI-Powered Search!

Doug and Trey are both tenured experts in the world of search and relevance engineering. This one is packed with information!

Covering designing search experiences, types of search, user interfaces for search, filters, the nuances of agentic search, using popularity as a feature in learning to rank... and I loved learning about their pioneering ideas on Wormhole Vectors and Reflected Intelligence!

I hope you find the podcast useful! As always more than happy to discuss these things further with you!

YouTube: https://www.youtube.com/watch?v=ZnQv_wBzUa4 Spotify: https://spotifycreators-web.app.link/e/wvisW7tga1b

I am SUPER EXCITED to publish the 132nd episode of the Weaviate Podcast with Thomas van Dongen, head of AI engineering at Springer Nature!

Thomas is the creator of Pyversity, a fast, lightweight open-source python library for diversifying retrieval results!

Diversity is such an underrated topic in AI and Vector Databases. Whether searching through e-Commerce products or Scientific papers, we often want serendipity from our search engine, results that we were not expecting to find!

For example, say you ask "Who has accomplish the most in professional sports?". A relevance optimized search system might return 10 results all about Michale Jordan... whereas a diversity enhanced system would produce information about Michael Jordan, Tom Brady, Tiger Woods, ...

Rather than just return the relevance ranked search results, Pyversity uses methods such as Maximal Marginal Relevnace (MMR) or Determinantal Point Process (DPP) to achieve diverse results.

I learned a lot from this conversation exploring the general topic of diversity in vector spaces, diversification strategies from Maximal Marginal Relevance (MMR) to Determinantal point process (DPP), and more, as well as Thomas' work and thoughts on AI in Scientific Literature!

I hope you find it interesting!

YouTube: https://www.youtube.com/watch?v=L2N1qvfP7tg

Spotify: https://spotifycreators-web.app.link/e/FBaEZ9b5UYb

I am SUPER EXCITED to publish the 132nd episode of the Weaviate Podcast with Thomas van Dongen, head of AI engineering at Springer Nature!

Thomas is the creator of Pyversity, a fast, lightweight open-source python library for diversifying retrieval results!

Diversity is such an underrated topic in AI and Vector Databases. Whether searching through e-Commerce products or Scientific papers, we often want serendipity from our search engine, results that we were not expecting to find!

For example, say you ask "Who has accomplish the most in professional sports?". A relevance optimized search system might return 10 results all about Michale Jordan... whereas a diversity enhanced system would produce information about Michael Jordan, Tom Brady, Tiger Woods, ...

Rather than just return the relevance ranked search results, Pyversity uses methods such as Maximal Marginal Relevnace (MMR) or Determinantal Point Process (DPP) to achieve diverse results.

I learned a lot from this conversation exploring the general topic of diversity in vector spaces, diversification strategies from Maximal Marginal Relevance (MMR) to Determinantal point process (DPP), and more, as well as Thomas' work and thoughts on AI in Scientific Literature!

I hope you find it interesting!

YouTube: https://www.youtube.com/watch?v=L2N1qvfP7tg

Spotify: https://spotifycreators-web.app.link/e/FBaEZ9b5UYb

AI is transforming Database Systems. Perhaps the biggest impact so far has been natural language to query language translations, or Text-to-SQL. However, another massive innovation is brewing.

AI presents new Semantic Operators for our query languages. For example, we are all familiar with the WHERE filter. Now we have AI_WHERE, in which an LLM, or another AI model, computes the filter value without needing it to already be available in the database!

```sql SELECT * FROM podcasts AI_WHERE “Text-to-SQL” in topics ```

Semantic Filters are just the tip of iceberg, the roster of Semantic Operators further includes Semantic Joins, Map, Rank, Classify, Groupby, and Aggregation!

And it doesn’t stop there! One of the core ideas for Relational Algebra and how its influenced Database Systems is query planning and finding the optimal order to apply filters. For example, let’s say you have two filters, the car is red and the car is a BMW. Now let’s say the dataset only contains 100 BMWs, but 50,000 red cars!! Applying the BMW filter first will limit the size of the set for the next filter!

This foundational idea has all sorts of extensions now that LLMs are involved! This opportunity is giving rise to new query engines and declarative optimizers such as Palimpzest, LOTUS, and others!

I am SUPER EXCITED to publish the 131st episode of the Weaviate Podcast with Matthew Russo, a Ph.D. student at MIT!

So many interesting nuggets in this podcast, loved discussing these things with Matthew, and I hope you find it interesting!

YouTube: https://youtu.be/koPBr9W4qU0

Spotify: https://spotifycreators-web.app.link/e/ddUhVMmLoYb

Medium: https://medium.com/@connorshorten300/semantic-query-engines-with-matthew-russo-weaviate-podcast-131-131a42bbc521

I am SUPER EXCITED to publish the 130th episode of the Weaviate Podcast featuring Xiaoqiang Lin, a Ph.D. student at the National University of Singapore! During his time at Meta, Xiaoqiang lead the research behind REFRAG: Rethinking RAG-based Decoding!

Traditional RAG systems use vectors to find relevant contexts with semantic search, but then throw away these vectors when it is time to pass the retrieved information to the LLM! REFRAG instead feeds the LLM these pre-computed vectors, achieving massive gains in long context processing and LLM inference speeds!

REFRAG makes Time-To-First-Token (TTFT) 31x faster and Time-To-Iterative-Token (TTIT) 3x faster, boosting overall LLM throughput by 7x while also being able to handle much longer contexts!

This is such an exciting evolution for the applications of Vector Databases, and Weaviate’s mission to weave AI and Database systems together! I loved diving into the details of REFRAG with Xiaoqiang, I hope you enjoy the podcast!

YouTube: https://www.youtube.com/watch?v=yi7v-UXMg0U

Spotify: https://spotifycreators-web.app.link/e/RWvmvMgRZXb

REFRAG from Meta Superintelligence Labs is a SUPER exciting breakthrough that may spark the second summer of Vector Databases! REFRAG illustrates how Database Systems are becoming even more integral to LLM inference!

By making clever use of how context vectors are integrated with LLM decoding, REFRAG is able to make TTFT (Time-to-First-Token) 31X faster and TTIT (Time-to-Iterative-Token) 3X faster, overall improving LLM throughput by 7x!! REFRAG is also able to process much longer input contexts than standard LLMs!

How does it work?

Most of the RAG systems today that are built with Vector Databases, such as Weaviate, throw away the associated vector with retrieved search results, only making use of the text content. REFRAG instead passes these vectors to the LLM, instead of the text content!

This is further enhanced with a fine-grained chunk encoding strategy, and a 4-stage training algorithm that includes a selective chunk expansion policy trained with GRPO / PPO.

Here is my review of the paper! I hope you find it useful!

YouTube: https://www.youtube.com/watch?v=Ek0tZootK00

I am SUPER excited to publish the 128th episode of the Weaviate Podcast featuring Charles Pierse!

Charles has lead the development behind the GA release of Weaviate’s Query Agent!

The podcast explores the 6 month journey from alpha release to GA! Starting with the meta from unexpected user feedback, collaboration across teams within Weaviate, and the design of the Python and TypeScript clients.

We then dove deep into the tech! Discussing citations in AI systems, schema introspection, multi-collection routing, and the Compound Retrieval System behind search mode.

Back into the meta around the Query Agent, we ended with its integration with Weaviate's GUI Cloud Console, our case study with MetaBuddy, and some predictions for the future of the Weaviate Query Agent!

I had so much fun chatting about these things with Charles! I really hope you enjoy the podcast!

YouTube: https://www.youtube.com/watch?v=TRTHw6vdVso

Spotify: https://spotifycreators-web.app.link/e/2Rr2Mla5RWb

Hey Hacker News community,

I think this is a really cool example of prompt optimization with DSPy and the newest GEPA optimizer! I hope you find it interesting!

https://www.youtube.com/watch?v=H4o7h6ZbA4o

I am SUPER EXCITED to publish the 127th episode of the Weaviate Podcast featuring Lakshya A. Agrawal!

Lakshya is a Ph.D. student at U.C. Berkeley, where he has lead the research behind GEPA: Reflective Prompt Evolution can Outperform Reinforcement Learning!

GEPA is a huge step forward for automated prompt optimization, DSPy, and the broader scope of integrating LLMs with optimization algorithms!

The podcast discusses all sorts of aspect of GEPA from the Reflective Prompt Mutation to Pareto-Optimal Candidate Selection, Test-Time Training, the LangProBe Benchmark, and more!

I learned so much from discussing these things with Lakshya, and I really hope you enjoy the podcast!

YouTube: https://www.youtube.com/watch?v=fREQrxhBSk0

Spotify: https://creators.spotify.com/pod/profile/weaviate/episodes/GEPA-with-Lakshya-A--Agrawal---Weaviate-Podcast-127-e36qaq1/a-ac3oens