# Weaviate announced a three-agent suite while keeping availability stages distinct Weaviate announced its Query, Transformation and Personalization Agents on March 4, 2025; only Query Agent entered public preview in that launch record. Canonical: https://rangoon.ai/insights/archive-weaviate-agents-suite-march-2025/ Author: Rangoon Editorial (https://rangoon.ai/insights/editorial/) Published: 2026-10-24 Updated: 2026-10-24 Event date: 2025-03-04 Tags: [Weaviate](https://rangoon.ai/insights/tags/weaviate/), [AI agents](https://rangoon.ai/insights/tags/ai-agents/), [retrieval](https://rangoon.ai/insights/tags/retrieval/), [vector search](https://rangoon.ai/insights/tags/vector-search/) Source note: Historical reporting: Weaviate announced Weaviate Agents on March 4, 2025 as Query, Transformation and Personalization services. Its launch record says Query Agent was in public preview, while later posts separately introduce the other services. The public documentation provides product context. Product descriptions of natural-language querying and transformations do not prove schema understanding, safe mutations, relevance or availability in every customer collection. Image: Natural-language request routed to separate query, transformation and personalization paths. Original Rangoon explanatory diagram: Source boundary: dated primary release reporting; original editorial analysis. No vendor interface, photograph, measurement result, or endorsement implied. ## Key takeaways - The event introduced a suite concept, but each named agent had its own availability state. - Querying, transformation and personalization change data in different ways and need different review paths. - Natural-language abstraction does not remove schema, retrieval or authorization requirements. ## Weaviate announced its agent suite on March 4, 2025 [source 1](https://weaviate.io/blog/weaviate-agents) Weaviate’s March 4, 2025 launch describes three named services: Query Agent, Transformation Agent and Personalization Agent. The announcement presents them as services that use a Weaviate collection and large language models to help query, transform or rank data. This is a product announcement about an agent suite, not proof that all three services reached the same release stage. The source specifically says Query Agent was available in public preview. Keeping that status visible matters because a suite name can hide material differences in access, support, documentation and operational maturity among its parts. ## Query Agent was presented as a natural-language route into search and aggregation [source 1](https://weaviate.io/blog/weaviate-agents) [source 2](https://weaviate.io/blog/query-agent) The launch says Query Agent interprets a natural-language question, determines relevant information, forms search or aggregation operations and returns ranked results. That is a useful description of a layer between a user request and collection operations. It is not evidence that the agent always chooses the intended filter, understands every schema term or retrieves complete evidence. The following day’s dedicated Query Agent post supplies more product context, while the March 4 announcement establishes the date and public-preview state. Any deployment must still define which fields are accessible, how an answer exposes evidence, what happens on ambiguity and whether generated plans can call operations beyond read-only search. ## Transformation work changes a collection and therefore differs from answering a query [source 1](https://weaviate.io/blog/weaviate-agents) [source 2](https://weaviate.io/blog/query-agent) Weaviate described a Transformation Agent that can use a prompt to create or insert a value for data stored in a collection. That makes its failure modes different from a read-only retrieval assistant. A wrong query can return an unhelpful result; a wrong transformation can create new stored content that affects later search, filters or summaries. The March launch presents the capability direction, and later documentation or releases are needed to establish the availability of individual functions. The product description does not prove that an output is correct for every record or that a generated insertion is authorized. Teams need source-of-truth fields, review paths, reversibility and validation appropriate to their collection before treating a transformation as operationally routine. ## Personalization introduced user context as a distinct retrieval input [source 1](https://weaviate.io/blog/weaviate-agents) [source 2](https://weaviate.io/blog/query-agent) The launch also described a Personalization Agent intended to rerank results using user context and past interactions. Personalization is not only a retrieval-quality feature: it changes which data is treated as input and which result ordering a user sees. That makes provenance, permitted context, consent and evaluation design practical questions for an adopter, even though the announcement itself is only a product record. The suite statement does not say that a generic ranking is universally improved, and it does not prove behavior for a customer’s data. A meaningful evaluation would hold the collection, query set and success criteria visible while comparing the selected personalization configuration with a non-personalized baseline. ## Sources 1. [Weaviate: Meet Weaviate Agents](https://weaviate.io/blog/weaviate-agents) 2. [Weaviate: Query Agent introduction](https://weaviate.io/blog/query-agent) ## Related reading - [Snowflake introduced Cortex Agents in public preview for work across structured and unstructured data](https://rangoon.ai/insights/archive-snowflake-cortex-agents-public-preview-february-2025/) - [Elastic introduced Agent Builder after an October technical-preview release note](https://rangoon.ai/insights/archive-elastic-agent-builder-introduction-october-2025/) - [Qdrant Cloud Inference put embedding generation, storage and indexing behind one API call](https://rangoon.ai/insights/archive-qdrant-cloud-inference-embedding-indexing-july-2025/) ## More information - [Documentation index](https://rangoon.ai/llms.txt) - [AI access and policies](https://rangoon.ai/ai/) - [Source repository](https://github.com/hypler-dev/rangoon)