Wayface

AI pilot Solo engineer

A reusable face verification engine — enrollment, verification, and identification policies behind HTTP, CLI, SDK, and test adapters (repo: pareidolia).

Problem

Face verification prototypes glue OpenCV scripts to one deployment shape. Swap the host — HTTP service, batch job, CLI tool, unit test — and you rewrite the integration layer. Quality gates, embedding versioning, and candidate limits scatter across copies.

Wayface centralizes detection, embedding, gallery management, and policy objects in a Python engine with thin adapters for each runtime surface.

Engine surface

FaceVerifier with configurable FaceSettings profiles including CPU-accurate mode
Enrollment, verification, and identification policies with explicit quality bands
Typed errors for no-face, multi-face, low confidence, and embedding version mismatch
Docker Compose API with health checks, doctor command, and localhost-first binding
Mock doctor path for CI without loading production models

Adapter layout

from wayface.engine import (
    FaceVerifier,
    EnrollmentPolicy,
    VerificationPolicy,
    IdentificationPolicy,
    FaceQuality,
    Embedding,
    FaceGallery,
)

__all__ = [
    "FaceVerifier",
    "EnrollmentPolicy",
    "VerificationPolicy",
    "IdentificationPolicy",
    "FaceQuality",
    "FaceGallery",
]

Screenshot placeholder — swap for verification diagnostics or API health output.