⚠️ Konzept & Simulation — kein Live-Produkt. Die echte Übersetzung testest du auf der Startseite.
LingradaNext 16 Frontend

Real-time simultaneous translation platform powered by AWS Hybrid & Gemini Live API

Fargate Cluster Pool: IDLE ⚪
Vercel State: READY 🟢

Architectural Specification Registry

Lingrada utilizes an Enterprise WebRTC Hybrid pipeline. Highly optimized for Next.js 16 and Vercel! To minimize cold-starts and client complexity, browser signals are routed straight to automated private AWS ECS Fargate tasks running our Puppeteer translation transcoder.

AWS CDK v2Path: infra/cdk/lib/worker-stack.ts
Der kostenoptimierte AWS-CDK-Stack (TypeScript) für den Fargate-Worker — pro Stage (dev/staging/prod) instanziiert. 0,00 USD Leerlauf-NAT-Kosten durch Public-Subnet-Platzierung in der Default VPC, 1 vCPU / 2 GB RAM auf ARM64/Graviton, desiredCount 0 (per Next.js-API ad-hoc auf 1 skaliert) und sichere, stage-spezifische SSM-Secret-Bindung für den GEMINI_API_KEY.
// infra/cdk/lib/worker-stack.ts — pro Stage instanziiert (dev | staging | prod).
// Fargate Backend Service: idelt bei 0, die Next.js-Route skaliert ad-hoc auf 1.

const vpc = ec2.Vpc.fromLookup(this, 'DefaultVpc', { isDefault: true });

const cluster = new ecs.Cluster(this, 'Cluster', {
  clusterName: `translator-${stage}`, vpc,
});

const taskDef = new ecs.FargateTaskDefinition(this, 'TaskDef', {
  cpu: 1024, memoryLimitMiB: 2048,
  // Graviton: nativer Build auf Apple Silicon, günstigere Runtime
  runtimePlatform: {
    cpuArchitecture: ecs.CpuArchitecture.ARM64,
    operatingSystemFamily: ecs.OperatingSystemFamily.LINUX,
  },
});

taskDef.addContainer('worker', {
  image: ecs.ContainerImage.fromAsset(repoRoot, {
    file: 'infra/Dockerfile', platform: Platform.LINUX_ARM64,
  }),
  environment: { NODE_ENV: 'production', AWS_REGION: 'eu-central-1', STAGE: stage },
  // GEMINI_API_KEY als SecureString aus SSM (nur per ARN referenziert)
  secrets: { GEMINI_API_KEY: ecs.Secret.fromSsmParameter(geminiParam) },
  logging: ecs.LogDriver.awsLogs({ streamPrefix: 'worker', logGroup }),
});

// NAT-frei: Public Subnet + Internet Gateway -> 0,00 USD Leerlaufkosten
new ecs.FargateService(this, 'Service', {
  cluster, taskDefinition: taskDef,
  serviceName: `translator-${stage}-worker`,
  desiredCount: 0, assignPublicIp: true,
  vpcSubnets: { subnetType: ec2.SubnetType.PUBLIC },
});

Translation Cluster Live Sandbox

AWS Status
OFFLINE
Fargate Nodes
0 Instances
Model Latency
---
Transcoder CPU
0.0%
Guide Transmitter (🇩🇪 German)
STATION_LIVE

Geben Sie eine Tour-Erklärung auf Deutsch ein, um das Fargate Core Transcoding & die Gemini API Live-Übersetzung zu simulieren:

Verbindet mit Gemini via Live SDK
AWS VPC Hybrid Dataflow Bridge
1. GERMAN GUIDE
Audio 16kHz PCM
▼ Chime Main Meeting (WebRTC)
2. FARGATE PUPPETEER TASKContinuous Audio Worklet
▼ Bidirectional SSL WebSocket Stream
3. GOOGLE GEMINI LIVE APIModel Ready
▼ Translated Audio Return (24kHz Codec)
4. TARGET TOURISTSLow-Latency WebRTC
AWS Fargate Realtime Telemetry Logs

Active Cloud logs are idle. Click "Spin-Up AWS Cluster" on the toolbar to trigger the AWS ECS tasks.

Tourist Simulator Page Context

Test how tourists receive translation streams using WebAudio / Google Translate simulation.

Target Language:
11:51
LTE

Lingrada

Passenger Voice Channel

PASSIVE_WEBRTC

Join translation channel

Simultaneous WebRTC audio streams directly to your headphones.

Please Start the AWS stack first.

amazon-chime-sdk.min.js loaded

How to test simulated transcode:

  1. Click "Spin-Up AWS Cluster" to boot the 4 simulated translation servers.
  2. Open the Tourist Simulator by clicking "Listen in SPANISH".
  3. Press "Broadcast Audio (PCM)" to send the guide message to Google Gemini API.
  4. Hear the Translated voice speak back and see the translated text arrive!

Enterprise security guard:

Crucially uses standard Vercel AWS ECS launch tunnels to bypass browser microphone exposures and locks API keys inside private AWS VPC Fargate tasks safely.