Paulo BiaoPaulo BiaoSystems & Cloud Analyst
Projects
In development · Lab validated

PSOP

Physical Security Observability Platform

Operational visibility for cameras, recorders and the infrastructure behind them. PSOP keeps device connectivity, observed function and the freshness of each observation as separate, inspectable evidence.

  • React
  • TypeScript
  • NestJS
  • PostgreSQL
  • Prisma
  • Python
  • Turborepo
PSOP operations dashboard for the lab site: device health, active alerts and recent events
Real screenshot — the operations dashboard running against the lab site.
01 · The problem

“Online” is not the same as working

Collapsing everything into a single online indicator hides the difference between device connectivity and the function an operator actually depends on.

  • A camera can answer on the network without producing usable video.
  • A stream session can be negotiated and still deliver no video packets.
  • A check that passed earlier may no longer describe the site.
Why it exists

Fifteen years of field work with CCTV, alarms and monitoring in Brazil showed the gap firsthand: a recorder reported as online is not proof that a recording exists. PSOP turns that field question into software:

What evidence do we have that this part of the security system is working — and how recent is it?

02 · Architecture

From the device to inspectable evidence

  1. Lab site
    Real hardware

    A recorder, IP cameras and a gateway-managed device group on a lab network

  2. Edge gateway
    Python

    Recorder adapter, ONVIF/RTSP probes and a stream scheduler collect observations

  3. PSOP API
    NestJS · TypeScript

    Authenticated ingestion, organization-scoped devices, sites and alerts

  4. Evidence ledger
    PostgreSQL · Prisma

    Each observation stored with its source, confidence, level and time

  5. Operator dashboard
    React · Vite

    Fleet health, site topology, incidents, activity, sessions and audit

Connectivity

Can the collector reach the device?

Observed function

Is the device doing the job an operator depends on?

Freshness

Is the observation still recent enough to describe the site?

03 · Evidence model

A ladder of proof, not a status light

Every observation records how far up this ladder it reaches. The levels are defined in PSOP's data model; how far each has been proven is stated explicitly.

Defined & ingested

Direct-camera discovery still pending

  1. E0 · Unknown
  2. E1 · Endpoint reachable
  3. E2 · Identity observed
  4. E3 · Profile discovered
  5. E4 · Stream URI obtained
Lab validated

Through the Speco NVR, three Hikvision channels

  1. E5 · Session negotiated
  2. E6 · Media received
Defined, not yet collected

Next evidence to produce

  1. E7 · Recording proven
  2. E8 · Retrieval proven

E5/E6 were first measured through the recorder's own RTSP service, with a stream URI taken from the recorder UI. They are authenticated probe attestations, stored with OBSERVED confidence — never promoted automatically.

04 · What has been built

Implemented, validated and planned — kept apart

Implemented
  • Organization-scoped API — NestJS, PostgreSQL, Prisma
  • Roles: Admin, Operator and Viewer
  • TOTP MFA with recovery codes; mandatory first-login password change
  • Refresh-token rotation, session revocation and replay protection
  • Audit log of administrative activity
  • Sites, devices, telemetry and connectivity events
  • Alerts with a resolution workflow and email notifications
  • Evidence ledger with source, confidence and level
  • Gateway-derived status for gateway-managed devices
  • Site topology, incident rail and activity stream
Lab validated
  • RTSP session negotiation (E5) and received RTP video packets (E6) on three Hikvision channels through a Speco NVR
  • Speco watcher recovery path running against real lab telemetry
Pending & planned
  • Direct-camera hardware validation
  • Sustained periodic-scheduler operation
  • Recording and retrieval proof (E7–E8)
  • Dependency-aware operational topology and incident analysis
  • Additional documented vendor adapters
  • Deployment readiness: delivery, supervision and installation guidance
05 · Engineering evidence

Tested against a real database, on every change

The PSOP CI workflow runs on pull requests and pushes to main. Migrations and the integration suites run against a PostgreSQL 16 service container.

PSOP CI
  1. Prisma schema
    Schema validation
  2. Migrations
    Applied to a PostgreSQL 16 service
  3. Unit tests
    API · Jest
  4. Integration tests
    API e2e suites, including security
  5. Builds
    API build · dashboard type-check & build
17
API unit test files
Run in CI
6
API integration suites
Run in CI against PostgreSQL
17
Python gateway test files
Run locally
Integration suites
  • security
  • fleet
  • telemetry-ingestion
  • device-availability
  • device-intelligence
  • site-fleet-reliability

Run history is public in the repository's Actions tab.

Real screens from the lab site
Operations dashboard
06 · Engineering boundaries

Lab-stage work, stated as such

PSOP is not a deployed commercial platform. Lab validation describes specific hardware observations — not general manufacturer compatibility or production readiness.

  • Runs in a lab environment — not deployed as a production service
  • Validated hardware: one Speco NVR and three Hikvision cameras
  • E5/E6 results are authenticated probe attestations, not independent packet captures
  • The data model covers more device types (access controllers, sensors, intercoms, switches) than the lab has validated