Lines 29-69javascript
30 targetId: Args.string({ description: "Eval target ID", required: true })
33 parallel: Flags.integer({ description: "Number of parallel eval executions", default: 1 }),
34 "runs-per-eval": Flags.integer({ description: "Number of runs per eval for variance analysis", default: 1 }),
35 iteration: Flags.integer({ description: "Iteration number for tracking", default: 1 }),
36 "skip-baseline": Flags.boolean({ description: "Skip baseline (without_target) runs", default: false }),
37 "skip-grading": Flags.boolean({ description: "Skip LLM grading step", default: false }),
38 model: Flags.string({ description: "Model for eval execution (default: user configured)" }),
39 "grader-model": Flags.string({ description: "Model for grading (default: same as execution model)" }),
40 timeout: Flags.integer({ description: "Timeout per eval in ms", default: 12e4 })
43 const { args, flags } = await this.parse(_EvalRunCommand);
45 const target = await apiGet(`/api/eval-targets/${args.targetId}`);
46 const evals = await apiGet(`/api/eval-targets/${args.targetId}/evals`);
47 if (evals.length === 0) {
48 this.error(`No evals found for target "${target.name}". Create some with \`rulemetric evals create\`.`);
50 this.log(`Running ${evals.length} eval(s) for "${target.name}" v${target.version}`);
51 this.log(`Config: ${flags["runs-per-eval"]} run(s) per eval, parallel=${flags.parallel}, iteration=${flags.iteration}`);
LowEval
Package source references a known benign dynamic code generation pattern.
dist/commands/evals/run.jsView on unpkg · L49 52 if (flags["skip-baseline"]) this.log("Skipping baseline runs");
53 if (flags["skip-grading"]) this.log("Skipping grading");
55 const configurations = [
56 { name: "with_target", content: target.content }
58 if (!flags["skip-baseline"]) {
59 configurations.push({ name: "without_target" });
62 for (const eval_ of evals) {
63 for (const config of configurations) {
64 for (let run = 1; run <= flags["runs-per-eval"]; run++) {
67 configuration: config.name,
68 targetContent: config.content,