Weaviate adds agents to its tech stack to ease gen AI app development

Vector database provider Weaviate has added three new agents to its development stack to facilitate the development of generative AI-based applications.  

These agents are modular agentic workflows that use large language models (LLMs) and prompts to interact dynamically with data in Weaviate.

“Each Weaviate Agent uses an LLM pretrained on Weaviate’s APIs to perform tasks. Agents can be easily swapped, rearranged or used together, speeding up development with their simple APIs and powerful functionality,” a Weaviate spokesperson said.

While the vector database provider plans to launch more such agents, for now it has showcased three — Query Agent, Personalization Agent, and Transformation Agent.

The agents, which are expected to be made available as a separate service, could help developers automate their processes for building AI applications and enable natural language interactions with data, said Noel Yuhanna, principal analyst at Forrester.

“These Weaviate agents automate tasks such as vector queries, data transformation, and other processes using natural language while facilitating the search, retrieval, and orchestration of related data,” Yuhanna explained.  

The Futurum Group’s global technology analyst Bradley Shimmin sees these agents as clever marketing and believes that Weaviate is the first vector database provider that is marketing agents as a family of developer-focused tools.

However, he pointed out that they are not the first vendor to use generative AI to augment and automate tasks for developers, database administrators, data scientists, and even business users.

What are the Query, Transformation, and Personalization Agents?

The Query Agent is aimed at helping developers simplify complex query workflows and boost RAG pipelines by querying data using natural language, Weaviate said.

In the absence of the agent, developers would have to construct a query understanding pipeline, which Weaviate claims is challenging to build, maintain, and requires specialized expertise.

The Query Agent, according to Victoria Slocum, machine learning engineer at Weaviate, ditches SQL-to-text querying and uses function calling.

Function calling, in turn, uses an LLM to structure queries using predefined function calls in JSON format, with optional arguments for search, filters, aggregation, and grouping, Slocum explained in a LinkedIn post.

The Query Agent also has the ability to chain commands together, taking the results of a previous query and extending it with a new prompt, the database provider said.

The Transformation and Personalization Agents are targeted at organizing datasets and personalizing agent behavior respectively.

The Transformation Agent allows developers to augment and organize datasets with a single prompt and use cases include cleaning and organizing raw data for AI, generating and enriching metadata, automatically categorizing, labelling and preprocessing data, or even translating the entire dataset, the database provider said.

Changing database landscape, availability, and pricing

The release of Weaviate agents, according to Yuhanna, is in tune with the growing trend of integrating agents to automate various functions and processes to accelerate AI workloads.

“The advantages of these agents are that they can evolve independently, making it easier for enterprises to adopt them quickly than committing to a single, all-encompassing AI solution,” he explained.

However, Shimmin pointed out that there is no shortage of vector databases to choose from right now, particularly among stand-alone players like Weaviate, Qdrant, Faiss, Chroma, and Milvus.  

“This move by Weaviate I think may help the company compete somewhere between stand-alone and traditional database offerings, such as PostgreSQL, Oracle Database, or Microsoft Cosmos DB.  I would imagine Weaviate would favor establishing a rich developer ecosystem, building upon these agents as a means of becoming perhaps the next MongoDB,” Shimmin explained.

When asked about its target customers, a Weaviate spokesperson said that the agents are suitable for businesses of any size that want to make their data operations more efficient.

Weaviate’s Query Agent is currently available in Public Preview and the Transformation and Personalization Agents will be available in Preview later this month. During Preview, these Agents will be free to use as part of Weaviate Serverless Cloud and its free developer sandbox, Weaviate said, adding that detailed pricing will be updated in the future.