Plan the question
Tool-calling agents in Google ADK plan, route, and orchestrate queries across relational and document stores to answer complex natural-language questions.
PROJECT 02 Retrieval & reasoning
Agentic RAG AssistantBetter questions. Grounded answers.
A natural-language assistant that plans queries and brings relational and document data into the same conversation.
THE CONTEXT
Useful information lives in both structured tables and unstructured documents. This assistant routes natural-language questions across those sources through a multi-agent workflow.
Retrieve across sources
Hybrid retrieval combines vector search and SQL so the workflow can use both document context and structured data.
Evaluate in production
GCP deployment includes tracing, automated evaluation, LLMOps instrumentation, and inference cost monitoring.
One conversation across different kinds of data.
Talk through the detailsNEXT EXPLORATION