Introducing Search Effort Tiers in Query Agent

Hello Weaviate Community! 🤗

Query Agent's new configurable Search Mode effort tiers let you choose medium, high, or ultrahigh effort per request, giving direct control over the accuracy, latency, and cost tradeoff. Supporting updates add recall and precision filtering choices, suggested queries for chat interfaces, and new model upgrades.

Latest AI & tech insights

Explore our recent Weaviate content:

Read

• ✍️ Scaling Test-Time Compute in Search Mode: Learn how the new medium, high, and ultrahigh effort tiers let you control test-time compute when searching with Query Agent. Read the blog

• 🔍 Query Profiling: Learn how query profiling shows where a query spends its time, helping you diagnose bottlenecks and tune performance. Read the blog

• 🔖 Building Foundry: AI-Powered Creative Workflows: See how semantic search helps creative teams find and reuse assets by meaning, combining embeddings with metadata and previews. Read the blog

Watch

• 🎙 Knowledge Engineering with Dr. Bradley Allen: Dr. Allen explores five decades of AI history, including expert systems, knowledge graphs, LLMs, and why knowledge engineering remains essential for trustworthy enterprise AI. Watch the full podcast

• 🎙 Booking.com and Weaviate: Başak Eskili shares how Booking.com scaled vector search, RAG, and agentic AI to production. She covers the shift from keyword to semantic search, the move to Weaviate, a GenAI messaging agent, and the engineering that powers AI at scale. Watch the full podcast

Product highlights

Check out our latest product updates.

• 🔎 More control and guidance in Query Agent:

  • Recall vs. precision filtering: Use recall, the default multi-query strategy, to retrieve broadly, or precision, a single-query strategy, when the request has strict intent.

  • Suggest Queries: Propose starter questions and contextual follow-ups for chat interfaces, helping users begin and continue conversations.

  • GPT-5.6 Luna and Terra upgrade: Query Agent now uses the upgraded models automatically, improving retrieval quality with no added cost or latency.

  • Learn more

• 🧠 Engram: Give your agents persistent memory that they can write to and search across conversations, users, and topics. Learn 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