AI Code Builder — Success Metrics
Success Metrics
Back to: [AI Code Builder — Project README](Research: AI Code Builder)
Threshold minimum & target ideal: [Thresholds](Minimum Score / Threshold)
A. Visual Fidelity
| Metric | Deskripsi | Cara Ukur |
|---|---|---|
| Visual Similarity Score | Kemiripan hasil render code vs screenshot asli | SSIM / perceptual hash |
| Layout Accuracy | Posisi elemen sesuai struktur asli | Bounding box IoU |
| Color & Typography Match | Warna & font sesuai | Delta-E color distance |
B. Functional Correctness
| Metric | Deskripsi | Cara Ukur |
|---|---|---|
| Code Compilability Rate | % kode berhasil di-build tanpa error | Automated build test |
| Component Completeness | % elemen UI berhasil di-generate | Element-level recall |
| Functional Behavior Match | Interaksi sesuai ekspektasi | Automated E2E test |
C. Feature Detection Accuracy
| Metric | Deskripsi | Cara Ukur |
|---|---|---|
| Classification Accuracy | Ketepatan deteksi feature (Auth, dsb) | Precision, Recall, F1-score |
| Edge Case Handling Rate | % edge case (OTP/2FA/social login) tertangani | Manual review test set |
D. Operational & UX
| Metric | Deskripsi | Cara Ukur |
|---|---|---|
| Latency (Time-to-Code) | Waktu drop image → kode muncul | Rata-rata response time |
| Failure Rate | % request gagal diproses | Error / total request |
| Regeneration Rate | Frekuensi user re-generate | Event tracking |
Detail breakdown bagaimana metrik ini dipakai per sub-project: Backend, Mobile, Frontend.