fix(results): tighten result artifact contract

apps/cli/src/commands/eval/run-eval.ts

@@ -571,20 +571,14 @@ export function resolveExperimentNamespace(params: {
   resultGroupName: string;
 }): {
   experiment: string;
-  source: RunRuntimeSourceMetadata['experiment_namespace_source'];
 } {
   if (params.cliExperiment) {
-    return { experiment: params.cliExperiment, source: 'cli' };
+    return { experiment: params.cliExperiment };
   }
   if (params.tagsExperiment) {
-    return { experiment: params.tagsExperiment, source: 'tags' };
+    return { experiment: params.tagsExperiment };
   }
-  const source: RunRuntimeSourceMetadata['experiment_namespace_source'] = params.isMultiEval
-    ? 'multi_eval'
-    : params.suiteName
-      ? 'eval_metadata'
-      : 'eval_filename';
-  return { experiment: params.resultGroupName, source };
+  return { experiment: params.resultGroupName };
 }
 
 /**
@@ -902,11 +896,9 @@ function buildRuntimeSourceMetadata(params: {
   readonly activeTestFiles: readonly string[];
   readonly sourceTests: readonly EvalTest[];
   readonly fileMetadata: ReadonlyMap<string, { readonly options: NormalizedOptions }>;
-  readonly experimentNamespace: string;
-  readonly experimentNamespaceSource: RunRuntimeSourceMetadata['experiment_namespace_source'];
   readonly hasCliRuntimeConfig: boolean;
 }): RunRuntimeSourceMetadata {
-  const evalFiles = uniqueRuntimeSourcePaths(
+  const activeEvalFiles = uniqueRuntimeSourcePaths(
     params.activeTestFiles.map((filePath) => toRuntimeSourcePath(params.cwd, filePath)),
   );
   const activeResolvedFiles = new Set(
@@ -920,30 +912,22 @@ function buildRuntimeSourceMetadata(params: {
     const sourceFile = testSourceEvalPathForComparison(test);
     return sourceFile ? !activeResolvedFiles.has(sourceFile) : false;
   });
-  const kind =
-    params.activeTestFiles.length > 1
-      ? 'multi_eval'
-      : hasImportedSuite || hasNonActiveSourceFile
-        ? 'wrapper_eval'
-        : 'direct_suite';
-  const wrapperEvalFile =
-    kind === 'wrapper_eval'
-      ? toRuntimeSourcePath(params.cwd, params.activeTestFiles[0])
-      : undefined;
+  const isWrapperEval =
+    params.activeTestFiles.length === 1 && (hasImportedSuite || hasNonActiveSourceFile);
+  const wrapperEvalFile = isWrapperEval
+    ? toRuntimeSourcePath(params.cwd, params.activeTestFiles[0])
+    : undefined;
+  const evalFiles = sourceEvalFiles.length > 0 ? sourceEvalFiles : activeEvalFiles;
 
   return {
     schema_version: 'agentv.runtime_source.v1',
-    kind,
     config_source: buildRuntimeConfigSource({
       activeTestFiles: params.activeTestFiles,
       fileMetadata: params.fileMetadata,
       hasCliRuntimeConfig: params.hasCliRuntimeConfig,
     }),
-    experiment_namespace: params.experimentNamespace,
-    experiment_namespace_source: params.experimentNamespaceSource,
     eval_files: evalFiles,
     ...(wrapperEvalFile && { wrapper_eval_file: wrapperEvalFile }),
-    ...(sourceEvalFiles.length > 0 && { source_eval_files: sourceEvalFiles }),
   };
 }
 
@@ -1848,14 +1832,13 @@ export async function runEvalCommand(
     configTags: yamlConfig?.tags,
     cliTags: options.tagMap,
   });
-  const { experiment: resolvedExperimentNamespace, source: experimentNamespaceSource } =
-    resolveExperimentNamespace({
-      cliExperiment: resolvedExperiment.name,
-      tagsExperiment: normalizeString(resolvedTags?.experiment),
-      isMultiEval: resolvedTestFiles.length > 1,
-      suiteName: primarySuite?.metadata?.name,
-      resultGroupName,
-    });
+  const { experiment: resolvedExperimentNamespace } = resolveExperimentNamespace({
+    cliExperiment: resolvedExperiment.name,
+    tagsExperiment: normalizeString(resolvedTags?.experiment),
+    isMultiEval: resolvedTestFiles.length > 1,
+    suiteName: primarySuite?.metadata?.name,
+    resultGroupName,
+  });
   // Normalize once so the row `experiment` field, AGENTV_EXPERIMENT, and the
   // emitted tags map all agree (and any invalid-name error surfaces in one place).
   const normalizedExperiment = normalizeExperimentName(resolvedExperimentNamespace);
@@ -2266,8 +2249,6 @@ export async function runEvalCommand(
     activeTestFiles,
     sourceTests: activeSourceTests,
     fileMetadata,
-    experimentNamespace: normalizeExperimentName(options.experiment),
-    experimentNamespaceSource,
     hasCliRuntimeConfig,
   });
   const hasPerFileRuntimeThresholds =

apps/cli/src/commands/results/report.ts

@@ -86,10 +86,7 @@ function serializeReportResult(
   const runtimeSource = manifestRecord?.runtime_source ?? summaryMetadata?.runtimeSource;
   const resultExperiment = (result as EvaluationResult & { experiment?: string }).experiment;
   const experimentNamespace =
-    runtimeSource?.experiment_namespace ??
-    manifestRecord?.experiment ??
-    summaryMetadata?.experiment ??
-    resultExperiment;
+    manifestRecord?.experiment ?? summaryMetadata?.experiment ?? resultExperiment;
   const fallbackEvalFile =
     normalizeEvalFileLabel(manifestRecord?.eval_file) ??
     summaryMetadata?.evalFile ??
@@ -113,27 +110,13 @@ function serializeReportResult(
     output: result.output,
     assertions: result.assertions,
     experiment: experimentNamespace,
-    experiment_namespace: experimentNamespace,
     runtime_source: runtimeSource,
     runtime_source_label: formatRuntimeSourceLabel(runtimeSource),
     runtime_config_source_label: formatRuntimeConfigSourceLabel(runtimeSource?.config_source),
     eval_file: fallbackEvalFile,
   };
 }
 
-function formatRuntimeKindLabel(kind: RunRuntimeSourceMetadata['kind'] | undefined): string {
-  switch (kind) {
-    case 'direct_suite':
-      return 'Direct suite';
-    case 'wrapper_eval':
-      return 'Wrapper eval';
-    case 'multi_eval':
-      return 'Multi-eval';
-    default:
-      return 'Unknown source';
-  }
-}
-
 function formatRuntimeConfigSourceLabel(
   source: RunRuntimeSourceMetadata['config_source'] | undefined,
 ): string {
@@ -151,36 +134,11 @@ function formatRuntimeConfigSourceLabel(
   }
 }
 
-function formatNamespaceSourceLabel(
-  source: RunRuntimeSourceMetadata['experiment_namespace_source'] | undefined,
-): string {
-  switch (source) {
-    case 'cli':
-      return 'CLI namespace';
-    case 'tags':
-      return 'Tags namespace';
-    case 'eval_metadata':
-      return 'Eval metadata namespace';
-    case 'eval_filename':
-      return 'Eval filename namespace';
-    case 'multi_eval':
-      return 'Multi-eval namespace';
-    default:
-      return '';
-  }
-}
-
 function formatRuntimeSourceLabel(runtimeSource: RunRuntimeSourceMetadata | undefined): string {
   if (!runtimeSource) {
     return '';
   }
-  return [
-    formatRuntimeKindLabel(runtimeSource.kind),
-    formatNamespaceSourceLabel(runtimeSource.experiment_namespace_source),
-    formatRuntimeConfigSourceLabel(runtimeSource.config_source),
-  ]
-    .filter(Boolean)
-    .join(' · ');
+  return formatRuntimeConfigSourceLabel(runtimeSource.config_source);
 }
 
 function uniqueStrings(values: readonly (string | undefined)[]): string[] {
@@ -201,13 +159,7 @@ function escapeHtml(value: string): string {
 
 function formatReportHeaderContext(rows: readonly Record<string, unknown>[]): string {
   const experiments = uniqueStrings(
-    rows.map((row) =>
-      typeof row.experiment_namespace === 'string'
-        ? row.experiment_namespace
-        : typeof row.experiment === 'string'
-          ? row.experiment
-          : undefined,
-    ),
+    rows.map((row) => (typeof row.experiment === 'string' ? row.experiment : undefined)),
   );
   const runtimeSources = uniqueStrings(
     rows.map((row) =>

apps/cli/src/commands/results/serve.ts

@@ -1447,7 +1447,6 @@ async function handleRuns(c: C, { searchDir, agentvDir, projectId }: DataContext
           runtimeSource = deriveDashboardRuntimeSource({
             summaryMetadata,
             records,
-            inferredExperiment: experiment,
           });
         } else {
           // Run is in-progress with 0 results written yet — fall back to the
@@ -1456,7 +1455,6 @@ async function handleRuns(c: C, { searchDir, agentvDir, projectId }: DataContext
           runtimeSource = deriveDashboardRuntimeSource({
             summaryMetadata,
             records: [],
-            inferredExperiment: experiment,
           });
         }
       } catch {
@@ -1524,7 +1522,6 @@ async function handleRunDetail(c: C, { searchDir, projectId }: DataContext) {
     const runtimeSource = deriveDashboardRuntimeSource({
       summaryMetadata,
       records,
-      inferredExperiment: records[0]?.experiment,
     });
     // Surface run_dir + suite_filter for local runs so the UI can launch a
     // Dashboard-side resume against this exact run. Remote runs live in the
@@ -1646,7 +1643,6 @@ function deriveDashboardRuntimeSource(params: {
     runtimeSource?: RunRuntimeSourceMetadata;
     runtime_source?: RunRuntimeSourceMetadata;
   }[];
-  readonly inferredExperiment?: string;
 }): RunRuntimeSourceMetadata | undefined {
   const recordWithRuntimeSource = params.records.find(
     (record) => record.runtimeSource ?? record.runtime_source,
@@ -1659,25 +1655,17 @@ function deriveDashboardRuntimeSource(params: {
     return explicit;
   }
 
-  const experimentNamespace =
-    params.summaryMetadata.experiment ??
-    params.inferredExperiment ??
-    params.records.find((record) => record.experiment)?.experiment ??
-    'default';
   const evalFiles = uniqueRuntimeSourceValues([
     params.summaryMetadata.evalFile,
     ...params.records.map((record) => record.evalPath ?? record.eval_path),
   ]);
-  if (evalFiles.length === 0 && !experimentNamespace) {
+  if (evalFiles.length === 0) {
     return undefined;
   }
 
   return {
     schema_version: 'agentv.runtime_source.v1',
-    kind: evalFiles.length > 1 ? 'multi_eval' : 'direct_suite',
     config_source: 'defaults',
-    experiment_namespace: experimentNamespace,
-    experiment_namespace_source: 'unknown',
     eval_files: evalFiles,
   };
 }
@@ -3173,7 +3161,6 @@ export function createApp(
               runtimeSource = deriveDashboardRuntimeSource({
                 summaryMetadata,
                 records,
-                inferredExperiment: experiment,
               });
             }
           } catch {

apps/cli/test/commands/eval/artifact-writer.test.ts

@@ -26,7 +26,6 @@ import {
   type GradingArtifact,
   type IndexArtifactEntry,
   RESULT_INDEX_FILENAME,
-  RUN_CONFIG_FILENAME,
   type RunSummaryArtifact,
   type TimingArtifact,
   buildAggregateGradingArtifact,
@@ -1177,10 +1176,7 @@ describe('writeArtifactsFromResults', () => {
   it('writes optional runtime source metadata to summary only', async () => {
     const runtimeSource = {
       schema_version: 'agentv.runtime_source.v1' as const,
-      kind: 'direct_suite' as const,
       config_source: 'cli_flags' as const,
-      experiment_namespace: 'cli-smoke',
-      experiment_namespace_source: 'cli' as const,
       eval_files: ['evals/smoke.eval.yaml'],
     };
     const paths = await writeArtifactsFromResults([makeResult({ testId: 'alpha' })], testDir, {
@@ -1199,7 +1195,7 @@ describe('writeArtifactsFromResults', () => {
     expect(indexLine.runtime_source).toBeUndefined();
   });
 
-  it('moves experiment config metadata to an internal run config sidecar', async () => {
+  it('does not write experiment config metadata into public run artifacts', async () => {
     const experimentMetadata = {
       name: 'native-exp',
       target: 'codex-target',
@@ -1214,21 +1210,16 @@ describe('writeArtifactsFromResults', () => {
 
     const summary: RunSummaryArtifact = JSON.parse(await readFile(paths.summaryPath, 'utf8'));
     expect(summary.metadata).not.toHaveProperty('experiment_config');
-    expect(summary.metadata.run_config_path).toBe(`.internal/${RUN_CONFIG_FILENAME}`);
-
-    const runConfig = JSON.parse(
-      await readFile(path.join(paths.testArtifactDir, '.internal', RUN_CONFIG_FILENAME), 'utf8'),
-    );
-    expect(runConfig).toEqual({
-      schema_version: 'agentv.run_config.v1',
-      experiment_config: experimentMetadata,
-    });
+    expect(summary.metadata).not.toHaveProperty('run_config_path');
+    await expect(
+      readFile(path.join(paths.testArtifactDir, '.internal', 'run-config.json'), 'utf8'),
+    ).rejects.toThrow();
 
     await aggregateRunDir(paths.testArtifactDir);
     const rewrittenSummary: RunSummaryArtifact = JSON.parse(
       await readFile(paths.summaryPath, 'utf8'),
     );
-    expect(rewrittenSummary.metadata.run_config_path).toBe(`.internal/${RUN_CONFIG_FILENAME}`);
+    expect(rewrittenSummary.metadata).not.toHaveProperty('run_config_path');
   });
 
   it('omits duplicated root instances from run summary', () => {

apps/cli/test/commands/eval/tag-filtering.test.ts

@@ -144,20 +144,18 @@ describe('resolveExperimentNamespace', () => {
   it('prefers an explicit --experiment over tags.experiment and the default', () => {
     expect(
       resolveExperimentNamespace({ ...base, cliExperiment: 'cli-exp', tagsExperiment: 'tag-exp' }),
-    ).toEqual({ experiment: 'cli-exp', source: 'cli' });
+    ).toEqual({ experiment: 'cli-exp' });
   });
 
   it('uses tags.experiment when no --experiment is given', () => {
     expect(resolveExperimentNamespace({ ...base, tagsExperiment: 'tag-exp' })).toEqual({
       experiment: 'tag-exp',
-      source: 'tags',
     });
   });
 
   it('falls back to the eval-metadata default when neither is set (e.g. --tag experiment=)', () => {
     expect(resolveExperimentNamespace({ ...base })).toEqual({
       experiment: 'my-suite',
-      source: 'eval_metadata',
     });
   });
 
@@ -167,13 +165,13 @@ describe('resolveExperimentNamespace', () => {
         isMultiEval: false,
         resultGroupName: 'dataset',
       }),
-    ).toEqual({ experiment: 'dataset', source: 'eval_filename' });
+    ).toEqual({ experiment: 'dataset' });
   });
 
   it('labels multi-eval runs when no CLI/tags experiment is set', () => {
     expect(
       resolveExperimentNamespace({ isMultiEval: true, resultGroupName: 'multi-eval' }),
-    ).toEqual({ experiment: 'multi-eval', source: 'multi_eval' });
+    ).toEqual({ experiment: 'multi-eval' });
   });
 
   it('lets tags.experiment win over the multi-eval default', () => {
@@ -183,7 +181,7 @@ describe('resolveExperimentNamespace', () => {
         tagsExperiment: 'tag-exp',
         resultGroupName: 'multi-eval',
       }),
-    ).toEqual({ experiment: 'tag-exp', source: 'tags' });
+    ).toEqual({ experiment: 'tag-exp' });
   });
 });
 

apps/cli/test/commands/results/report.test.ts

@@ -160,23 +160,17 @@ describe('results report', () => {
       experiment: 'named-smoke',
       runtimeSource: {
         schema_version: 'agentv.runtime_source.v1',
-        kind: 'wrapper_eval',
         config_source: 'inline_experiment',
-        experiment_namespace: 'named-smoke',
-        experiment_namespace_source: 'eval_metadata',
-        eval_files: ['evals/wrapper.eval.yaml'],
+        eval_files: ['evals/child.eval.yaml'],
         wrapper_eval_file: 'evals/wrapper.eval.yaml',
-        source_eval_files: ['evals/child.eval.yaml'],
       },
     });
 
     const { outputPath } = await writeResultsReport(runDir, undefined, tempDir);
     const html = readFileSync(outputPath, 'utf8');
 
     expect(html).toContain('Experiment namespace: named-smoke');
-    expect(html).toContain(
-      'Runtime source: Wrapper eval · Eval metadata namespace · Inline experiment config',
-    );
+    expect(html).toContain('Runtime source: Inline experiment config');
   });
 
   it('embeds result text containing replacement tokens without corrupting the inline script', async () => {