Available on crate feature
eval only.Expand description
LLM-driven test case generation.
Generates evaluation test cases from natural language descriptions (via LLM)
or from production event logs (direct extraction). Produced cases follow
the standard [TestFile] JSON format and include generation metadata.
§Example
ⓘ
use adk_eval::test_generator::{TestGenerator, GeneratorConfig};
use std::sync::Arc;
let generator = TestGenerator::with_config(model, GeneratorConfig {
cases_per_description: 3,
include_tool_expectations: true,
});
let cases = generator
.generate_from_description("A weather assistant that can look up forecasts")
.await?;Structs§
- Eval
Case Metadata - Metadata for generated eval cases.
- Generator
Config - Configuration for test case generation.
- Test
Generator - Generates evaluation test cases from descriptions or event logs.