v1.0.0

ZepIris v1.0.0 — Open-source face search & content safety

We're excited to share v1.0.0 of ZepIris, an open-source microservice from Zepto Labs for face embedding, vector search, and automated image quality / safety checks.

What it does

ZepIris helps teams ship face-based products without stitching together embedding, storage, search, and safety pipelines from scratch:

• Face embeddings — 512-dimensional, L2-normalized vectors (InsightFace buffalo_l)
• 1-to-N vector search — Milvus-backed COSINE similarity
• Multi-tenant CRUD — insert, upsert, delete, and get under isolated namespaces
• Content safety — nudity, anti-spoof, and blur detection via a separate ML inference service
• Production-shaped architecture — main API (8000) + ML service (8001), MinIO for images, full Docker Compose stack
• Developer-friendly API — FastAPI, OpenAPI/Swagger, requestId on responses

Good fits

• Attendance and workforce check-in
• Onboarding with liveness / quality gates
• Access control with safety validation before enrollment

Get started

git clone https://github.com/zepto-labs/zepiris.git
cd zepiris
# See docs/QUICK_START.md and docs/LOCAL_SETUP_AND_TEST.md

• License: MIT
• Docs: README, API reference, configuration guide in-repo

Issues, PRs, and real-world feedback are welcome — this is the first stable open release and we'd love to hear how you use it.

Repo: https://github.com/zepto-labs/zepiris