How does the Weaviate Query Agent route natural-language questions across collections?

Rather than a developer writing explicit logic to decide which collection a given question is about and what kind of search (vector, keyword, hybrid, filtered) best answers it, the Query Agent is given access to a set of collections and interprets an incoming natural-language question to make those decisions itself.

agent = QueryAgent(client=client, collections=["Products", "Reviews", "FAQs"])
response = agent.run("Are there any complaints about battery life on the X200 model?")

Internally, this involves the agent reasoning about which of the available collections are actually relevant to the question (in this example, likely both Products, to identify the X200, and Reviews, to find complaint-related content), constructing the appropriate underlying queries (likely a hybrid or filtered search rather than pure vector similarity, given the specific product-model constraint), executing them, and synthesizing a coherent answer from whatever's retrieved.

This shifts the burden of query construction from the application developer writing explicit search logic for every anticipated question shape, to the agent interpreting each question dynamically at request time, which trades some of the predictability and fine-grained control of hand-written queries for significantly less upfront engineering effort per new question type an application needs to support.