Managed Weaviate on DigitalOcean, Search Agents, and the Future of Retrieval

Hello Weaviate Community! 🤗

What happens when agents become the primary users of search?

This week, we explore why retrieval quality is becoming the defining bottleneck in modern AI systems, how retrieval architectures are evolving for agents, and how your coding agents can now build production-ready RAG applications from a single prompt.

We also have exciting ecosystem news - Weaviate and DigitalOcean announced a partnership, bringing fully managed Weaviate to DigitalOcean in private preview!

Latest AI & tech insights

Explore our recent Weaviate content:

Read

• ✍️ Weaviate 1.37 Release ships a wave of useful and highly requested features for developers:

  • MCP Server - Weaviate now speaks MCP natively, so Claude, Cursor, and other AI agents can query and write to your database without glue code.

  • Diversity Search (MMR) - Solves the “five pizza images” problem. A new query-time reranking step penalizes near-duplicates so RAG pipelines stop retrieving five chunks that say the same thing.

  • Extensible Tokenizers - Accent folding, custom stopwords per property (finally useful for multilingual collections), and a new /tokenize endpoint so you can actually see what the analyzer produces before committing to a schema.

  • Query Profiling - Per-shard timing breakdowns on any search request. When a query is slow, you now know exactly where the time went.

  Also in 1.37: incremental backups, Gemini audio embeddings, BlobHash for storing only a SHA-256 instead of full blobs, and a new collection export API. Read the blog

• 📃 Your LLM Is Only as Good as What It Retrieves: A researcher's perspective on why retrieval quality is the bottleneck in modern RAG systems and how to measure what actually matters. Read the blog

Watch

• 🎙 Search Agents with Nandan Thakur: Dr. Thakur joins us fresh from his dissertation defense to discuss Orbit, a budget-friendly synthetic data pipeline for training search agents, and the shift from neural retrieval to iterative agentic search. Watch the full podcast

• 🎙 AgentIR with Zijian Chen and Xueguang Ma: Why retrieval systems need redesigning when agents, not humans are the users. The team discusses reasoning-aware retrieval, the BrowseComp-Plus benchmark, and Agent-Computer Interaction (AICI). Watch here

• 📽️ Building a production-ready legal RAG app... in one prompt: See how to use the Query Agent to build a complete legal assistant in about 12 minutes using Claude Code, spanning everything from PDF ingestion to a streaming Next.js frontend. Watch the video

Product highlights

☁️ Managed Weaviate on DigitalOcean (Private Preview)

Announced at DO Deploy, developers can now provision managed Weaviate clusters with full API compatibility. Learn more

✍️ Query Agent Updates

If you haven’t seen the latest updates to the query agent, we have been quietly working in the background to give you some new features:

Up to 2x faster responses: Using the latest model architectures, the query agent can now respond up to twice as fast, meaning less time waiting and more time working with your results.

Suggest queries mode: Want to generate some example questions in a chat-bot style application? Simply run:

qa = QueryAgent(
   client=client, 
   collections=["<your_collection_name>"]
)
qa.suggest_queries()

to receive a list of questions tailored to your collections and data. Great for allowing users to generate some sample questions to start exploring in an application! See the full documentation to get started.

Optional answer evaluation: The query agent now optionally gives an analysis of the sources used in its response. Simply pass:

qa.ask(..., result_evaluation="llm")

to use an LLM agent to determine which sources were used and whether the answer is incomplete.

Don’t forget to install the latest version of Weaviate Agents before you get started:

pip install -U weaviate-agents

Company updates

Join our team — We're hiring across various teams! ➡️ Check out our career page for exciting opportunities in engineering, sales, growth, and more.✨

Hungry for more?

Have a question or want to connect? Join our Weaviate Forum to engage in community conversations.

Follow our Blog, LinkedIn, Twitter, and YouTube for more content updates.

See you in two weeks,

Prajjwal