LIVE · 100 answersLast run: 33h agoJudge: gpt-4o-mini
Run eval
Iterate

Prompt Lab

Take low-scoring sessions, regenerate answers with a candidate prompt, compare metrics.

1Pick weak sessions
16 sessions below 70% pass
2Edit candidate prompt
apps/eval/prompts/candidates/<name>.md (NOT prod)
3Regenerate answers
gpt-4.1-nano · same Q + new prompt
4Compare metrics
OLD vs NEW · ship if better

Low-Scoring Sessions

click Run with new prompt to regenerate all questions in a session using the candidate prompt

Run all weak sessions
SessionTierRoleQ'sPassWorst metric
best buy — Quick Practiceconsumerunknown130%Context Precision: 0.18Run with new prompt
Quick Practiceconsumerunknown30%Context Precision: 0.39Run with new prompt
Stripe — AI enginner (Copy 104)enterpriseunknown40%Opening: 0.27Run with new prompt
Stripe — AI enginner (Copy 96)enterpriseunknown10%Voice Authenticity: 0.25Run with new prompt
Stripe — AI enginner (Copy 88)enterpriseunknown10%Context Precision: 0.22Run with new prompt
Stripe — AI enginner (Copy 87)enterpriseunknown10%Opening: 0.49Run with new prompt
Stripe — AI enginner (Copy 15)enterpriseunknown60%Voice Authenticity: 0.28Run with new prompt
Stripe — AI enginner (Copy 12)enterpriseunknown190%Voice Authenticity: 0.33Run with new prompt
Stripe — AI enginner (Copy 3)enterpriseunknown50%Context Precision: 0.03Run with new prompt
Google — AI ml engineerenterpriseunknown60%Context Precision: 0.07Run with new prompt
Google — AI enginner (Copy 5)enterpriseunknown20%Faithfulness: 0.36Run with new prompt
Google — AI enginner (Copy 6)enterpriseunknown250%Faithfulness: 0.22Run with new prompt
Google — AI enginner (Copy)enterpriseunknown10%Context Precision: 0.26Run with new prompt
Google — AI enginnerenterpriseunknown80%Context Recall: 0.01Run with new prompt
Google — AI enginner (Copy)enterpriseunknown20%Voice Authenticity: 0.44Run with new prompt
Google — AI enginnerenterpriseunknown30%Voice Authenticity: 0.55Run with new prompt

How regeneration works

  1. Copy prod prompt to a candidatecopy apps\server\prompts\_base\VOICE_AND_AUTHENTICITY.md apps\eval\prompts\candidates\my-test.md. Edit the CANDIDATE file, NEVER touch the prod template directly.
  2. Trigger the workflow — click "Run with new prompt" on a session above. Opens GH Actionseval-run.yml. Pick mode "regenerate" and enter your candidate prompt path (relative to repo root) likeapps/eval/prompts/candidates/my-test.md
  3. Pipeline runs — pulls each question from the session, callsgpt-4.1-nanowith the new prompt, scores the new answers with the same 7-metric judge.
  4. Compare + export — once the run finishes, the A/B Compare page shows OLD vs NEW side-by-side. Click Export Markdown to download a self-contained report you can hand to a prompt-review agent. The agent reviews the metrics + Q&A pairs and updates the prod template based on what worked.
See apps/eval/README.md for the full CLI reference (regenerate command, A/B options, threshold tuning).