@ledgelm/reporter

The TypeScript reporter buffers eval results and sends them to LedgeLM. It does not run evals or calculate baselines.

import { flush, report } from '@ledgelm/reporter';

report({
	name: 'response_grounding',
	type: 'judge',
	score: 0.82,
	passed: true,
	usage: {
		target: { input_tokens: 812, output_tokens: 146 },
		judge: { input_tokens: 1042, output_tokens: 87 }
	},
	input: { messages: [{ role: 'user', content: 'Can I return this purchase?' }] },
	output: 'Yes. Refunds are available within 30 days with a receipt.',
	expected_output: 'Explain the 30-day refund policy and receipt requirement.',
	provenance: {
		target_id: 'support-claude-sonnet',
		target: { provider: 'anthropic', model: 'claude-sonnet' },
		dataset: { name: 'support-golden', version: '2026-07' }
	}
});

const verdict = await flush();

Set LEDGELM_API_URL and LEDGELM_API_TOKEN in GitHub Actions. To target a local or alternate deployment explicitly, pass flush({ api_url: 'https://your-tunnel.example' }); the explicit option takes precedence over the environment variable.

Upload failures warn and return an unknown verdict, so LedgeLM availability does not fail the eval job.

For a GitHub Actions matrix, do not flush in each shard. Export a typed artifact and let the LedgeLM fan-in action flush the complete matrix:

import { reportArtifactSchema, type ReportInput } from '@ledgelm/reporter';

const results: ReportInput[] = await runSuite();
const artifact = reportArtifactSchema.parse({
	schema_version: 1,
	shard: process.env.LEDGELM_SHARD!,
	results
});
await Bun.write('ledgelm-results.json', JSON.stringify(artifact));

The action validates this exact schema before it buffers or uploads any result.