Skip to main content

EvalSpec

Agent Runtimes


Agent Runtimes / types/evals / EvalSpec

Interface: EvalSpec

Defined in: src/types/evals.ts:9

Built-in evaluator specification.

Properties

category

category: "Comparison" | "Type Validation" | "Performance" | "LLM-as-a-Judge" | "Span-Based" | "Report"

Defined in: src/types/evals.ts:19

Evaluator family


cost_tier

cost_tier: "free" | "llm"

Defined in: src/types/evals.ts:39

Cost tier for running this evaluator


default_config?

optional default_config?: Record<string, unknown>

Defined in: src/types/evals.ts:47

Suggested baseline configuration


description

description: string

Defined in: src/types/evals.ts:17

Description of the evaluator


evaluator_type

evaluator_type: "case" | "report"

Defined in: src/types/evals.ts:27

Case-level or report-level evaluator


id

id: string

Defined in: src/types/evals.ts:11

Unique eval identifier


latency

latency: "instant" | "fast" | "slow"

Defined in: src/types/evals.ts:41

Expected latency profile


name

name: string

Defined in: src/types/evals.ts:15

Display name


output_kind

output_kind: "boolean" | "boolean_with_reason" | "score" | "score_and_assertion" | "report_table" | "report_curve"

Defined in: src/types/evals.ts:31

Primary output shape


pydantic_class

pydantic_class: string

Defined in: src/types/evals.ts:29

Pydantic evaluator class name


requires?

optional requires?: string[]

Defined in: src/types/evals.ts:43

Runtime requirements (e.g. expected_output, logfire, model)


source

source: string

Defined in: src/types/evals.ts:45

Source documentation URL


version?

optional version?: string

Defined in: src/types/evals.ts:13

Version