BigQuery Tools

The GoogleAdk.Core.Tools namespace includes two tools for working with Google Cloud BigQuery: - BigQueryMetadataTool: Fetches datasets, tables, and schemas so the LLM can "explore" the database structure. - BigQueryQueryTool: Executes standard SQL queries.

Prerequisites

The host application must be authenticated with Google Cloud (via ADC or GOOGLE_APPLICATION_CREDENTIALS) and have permissions to execute queries in the target project.

Usage

Attach the tools to your agent and provide clear instructions. The LLM will automatically determine when to explore metadata and when to execute a query.

using GoogleAdk.Core.Agents;
using GoogleAdk.Core.Tools;

var analystAgent = new LlmAgent(new LlmAgentConfig
{
    Name = "data_analyst",
    Model = "gemini-2.5-flash",
    Instruction = "You are a data analyst. Explore the dataset structure, then query to answer questions.",
    Tools = [
        new BigQueryMetadataTool(), 
        new BigQueryQueryTool()
    ]
});

Execution Flow

1. User asks: "Top 5 countries in my-project.sales.2024?"
2. LLM calls bigquery_metadata to find columns (e.g., country_code, total_amount).
3. LLM calls bigquery_query with generated SQL: SELECT country_code, SUM(total_amount)...
4. The tool executes the query via the BigQuery client and returns JSON.
5. The LLM formats the final answer.