Skip to content
docs.
WORKSPACE

Your documentation

Explore the project catalog
Localzet ServerАсинхронный PHP-серверTriangleФреймворк и компонентыMIX IDИдентификация и интеграцииHack APIПубличные API хакатонаACNПроект и компоненты✓Userlay / UserNetПроект и компонентыAxiomПроект и компонентыМесседжинг через почтуПроект и компонентыLWT / JWT / JOSEПроект и компонентыФинансовая аналитика KimaiПроект и компонентыМультипарадигмальная БДПроект и компонентыZorin PlatformПроект и компонентыZorin TrustПроект и компонентыFloraScope / VegwatchПроект и компонентыaegis-qДокументация проектаaevumdbДокументация проектаauditДокументация проектаauto-invest-managerАвтотрейдерcaine-companionДокументация проектаcertbot-cloudflareДокументация проектаClusterLocalzet Server / ClustercmdveilДокументация проектаConsolePHP CLICronБиблиотеки Localzet Serverdata-connectorДокументация проектаdevice-controlDevice Controlenn-projectДокументация проектаEventsБиблиотеки Localzet ServerflowsintДокументация проектаforgeДокументация проектаHttpClientБиблиотеки Localzet ServerinterpreterСобственный языкinvariaДокументация проектаkeynotesДокументация проектаknotrouteKnotroutelaravel-personal-analyticsДокументация проектаlocalzet-osДокументация проектаmeridianДокументация проектаMQTTБиблиотеки Localzet Servernano-kernelСобственная ОСNoterДокументация проектаpersonal-analyticsДокументация проектаProxyБиблиотеки Localzet ServerPSRБиблиотеки Localzet ServerPSR-7Библиотеки Localzet Serverrschs-monitorДокументация проектаsemsealДокументация проектаshardsealДокументация проектаSocketIOБиблиотеки Localzet ServerSOCKS5Библиотеки Localzet Serverstandalone-analyticsДокументация проектаstatic-php-cliДокументация проектаStorageБиблиотеки Localzet ServersutureДокументация проектаtempoweaveДокументация проектаttsСинтез речиTunnelБиблиотеки Localzet Serverui-kitUI KitvesperДокументация проектаWebAnalyzerWebAnalyzerzorin-access-brokerДокументация проектаzorin-authority-brokerДокументация проектаzorin-authority-sdk-phpДокументация проектаzorin-authority-sdk-pythonДокументация проектаzorin-authority-sdk-tsДокументация проектаzorin-client-intelДокументация проектаzorin-device-core-aospДокументация проектаzorin-host-agentДокументация проектаzorin-native-labДокументация проектаzorin-nodeДокументация проектаzorin-opsДокументация проектаdonstu-python-sdkДокументация проектаfloortraceFloortraceserver-rsLocalzet Server / Cluster
Catalog

Visual Demo Architecture

The visual demo turns ACN from a backend-first research platform into a leadership-friendly demonstration: watch the neural network learn, fail, recover, adapt and improve.

This is a demo layer, not a new distributed training architecture.

Module dependenciesdegradationVisual airplane/ship samplesLive CNN trainingLoss/accuracy/LR telemetryAdaptive controller stateAUTO/MANUAL gateRollback + checkpoint restoreRecovery trainingFinal inference panel
SVG ↗

The viewer sees:

  • current epoch, stage, branch and checkpoint;
  • live loss, accuracy and learning-rate curves;
  • validation images with predicted class, confidence and correctness;
  • checkpoint timeline;
  • adaptive event feed;
  • AUTO/MANUAL mode;
  • manual approve/reject for rollback;
  • upload-based final inference against the current model.
Module dependenciesapps/webVisualTrainingDemoViewFastAPI /api/v1/demo/*SSE /api/v1/demo/eventsVisualDemoSessionTinyVisualClassifierGenerated visual airplane/shipdataIn-memory checkpoint states
SVG ↗

Location:

  • packages/acn/src/acn/experiments/visual_demo.py
  • apps/api/src/acn_api/visual_demo.py

The backend owns a single local VisualDemoSession.

It provides:

  • GET /api/v1/demo/state
  • GET /api/v1/demo/events
  • POST /api/v1/demo/start
  • POST /api/v1/demo/pause
  • POST /api/v1/demo/resume
  • POST /api/v1/demo/rollback
  • POST /api/v1/demo/auto-mode
  • POST /api/v1/demo/approve
  • POST /api/v1/demo/reject
  • POST /api/v1/demo/predict

The demo intentionally uses SSE instead of Redis. A single local browser observing a single local training session does not justify a queue or event bus.

The demo uses lightweight generated visual samples:

  • airplane
  • ship

This keeps the demo reliable without dataset downloads while still being visually understandable. The model is a real small CNN trained with PyTorch; metrics and predictions come from the actual model state.

At the configured degradation stage the demo intentionally applies:

  • a learning-rate spike;
  • label inversion;
  • noisy inputs.

The UI then shows:

  1. validation degradation;
  2. controller detection;
  3. rollback event;
  4. checkpoint restoration;
  5. learning-rate reduction;
  6. recovery training.

Manual mode pauses before rollback and waits for approve/reject. Auto mode executes rollback without operator approval.

Location:

  • apps/web/src/views/VisualTrainingDemoView.tsx
  • apps/web/src/hooks/useVisualDemo.ts
  • apps/web/src/api/visualDemoApi.ts
  • apps/web/src/types/visualDemo.ts

The visual demo is the first navigation item in the dashboard.

Run API:

Окно терминала
make api

Run frontend:

Окно терминала
make web

Open:

http://localhost:5173

Then open Live Demo and press Start.

  • No Kubernetes, Kafka, Celery, Ray or distributed scheduler.
  • No Redis requirement for the demo.
  • No MinIO requirement for the demo.
  • No MLflow requirement for the demo.
  • Checkpoints are in-memory because this is a presentation loop; durable artifact lifecycle remains covered by the core ACN artifact system and real vertical slice.

The demo exists to make ACN understandable. Production artifact persistence, PostgreSQL history and MLflow run comparison remain separate platform concerns.

Source