Promptfoo eval

Promptfoo Benchmark v0

Actual Promptfoo run for the locked 50-case Benchmark v0 packet. The provider replays existing runtime artifacts; it does not generate new shirts, start live judges, mutate customer data, or spend API tokens.

Promptfoo covers runtime artifact replay invariants for all 50 cases. Live visual quality judges remain a separate approval-gated step; the media packet is linked below but is no longer mislabeled as Promptfoo.

Promptfoo Report

Open the real Promptfoo Cloud report for the Levyo organization. The local HTML/JSON exports stay available as backup evidence.

Open promptfoo.com Open local HTML Open local JSON

Benchmark Surface

Return to the 50-case Benchmark v0 overview or the separate 20-case M8 label queue.

Open Benchmark v0 Open M8 labels

M5 Media Packet

The image-visible no-spend media packet and scorecard remain available as supporting evidence, not as a Promptfoo substitute.

Open media view Open scorecard

Artifact	Status	Ref
Promptfoo config	50 tests	threadly-openai/evals/levyo-benchmark-v0/promptfoo.benchmark-v0.yaml
Promptfoo provider	artifact replay	threadly-openai/evals/levyo-benchmark-v0/providers/benchmark-v0-replay-provider.mjs
Promptfoo cloud	loading	https://www.promptfoo.app/eval/eval-m02-2026-06-23T06:59:33
Local export	backup	/admin/promptfoo/promptfoo-benchmark-v0.html