@b4run/evals

Supported evaluation definitions, datasets, scorers, reports, and release gates for B4.run application behavior.

Use this when: You are defining repeatable evaluations, scorers, or release gates for a B4.run application.

Install

pnpm add -D @b4run/evals @b4run/testing

Example

import { contains, defineEval, gate, runEval } from "@b4run/evals"
import { createAgentHarness, script } from "@b4run/testing"

const evaluation = defineEval({
  name: "support replies",
  route: "/support#agent",
  dataset: [
    {
      input: "Where is my order?",
      fixtures: script()
        .user("Where is my order?")
        .replies("Your order is in transit."),
    },
  ],
  scorers: [contains("order", { threshold: 1 })],
  gate: gate.perScorer(),
})

await using harness = await createAgentHarness({
  appRoot: process.cwd(),
  route: "/support#agent",
})
const report = await runEval(evaluation, {
  runCase: async (testCase) => {
    if (typeof testCase.input !== "string") throw new TypeError("Expected string input")
    return harness.run({
      input: testCase.input,
      ...(testCase.fixtures !== undefined ? { fixtures: testCase.fixtures } : {}),
    })
  },
})

Runtime and stability

@b4run/evals is a supported node-only testing surface because it resolves JSON and JSONL datasets from disk. Deterministic agent runs come from @b4run/testing; scorer code still executes in replay, record, and live modes. llmJudge() needs a fixture, model credentials, or an injected fetch implementation.

Related

• Evals API reference — exact scorer, runner, and gate semantics.
• Evals — the application evaluation workflow.
• @b4run/testing and Fixtures and Recording — deterministic model calls.

Maturity and support

B4.run is pre-1.0, and its public surface can change. All publishable B4.run packages release together as a fixed group; review the @b4run/evals changelog and upgrading guide before upgrading. For support, use GitHub Discussions; report defects in GitHub Issues.

License

MIT. See the repository license.