Objective

Build a central operator console for Dahua face and access controllers so people are enrolled once, assigned through person groups, and synchronized to the right devices asynchronously — without typing the same person into every terminal.

Tasks I have done

• Full-stack product: Next.js operator console, Express REST API, and BullMQ device worker
• Central person registry with person groups and many-to-many group ↔ device mapping
• Desired-state reconciliation (ENSURE / REMOVE) with per-device sync rows and generation guards
• Dahua CGI adapter — digest auth, concurrency limits, RecNo capture, face enroll, and query encoding quirks
• Queued bulk import (Excel/CSV → zip), dry-run, per-row reports, and face photo normalization
• Migration cutover with conflict review, removal dry-run, and safe removals gating
• JWT/RBAC operator auth, encrypted device credentials, audit trail, and factory-reset lifecycle controls
• In-app Help / user manual for operators

Features

• Central persons, person groups, and device registration for Dahua ASI-series controllers
• Background sync via BullMQ — create, update, remove, and face enroll with progress and cancel
• Bulk onboarding from Excel/CSV with dry-run validation and face enrolment
• Migration tools for multi-device cutover, attribute conflicts, and deferred removals
• Operator RBAC, audit logging, and destructive reset flows with challenge confirmation
• BFF proxy so the browser talks only to Next.js; scripts call the API with an API key

Architecture

Operators manage the desired state in PostgreSQL. The worker pushes that state to Dahua controllers through a throttled Digest CGI client — card first, face enroll separately, with RecNo stored for safe deletes.