Frontend Integration Guide

Frontend (subscriber-portal)
  │
  ├─ POST /training/jobs (with config + hyperparameters)
  │
  ▼
eventml-svc
  ├─ 1. Check credits (wallet-svc) ─ 50 base + dynamic
  ├─ 2. Create SageMaker training job (AWS SDK)
  ├─ 3. SSE stream: epoch metrics (loss, accuracy)
  ├─ 4. Job completes → deduct credits
  ├─ 5. Kafka: eventzr.eventml.training.completed.v1
  └─ 6. notify-svc → push notification to user

const useTrainingStream = (jobId: string) => {
  const [metrics, setMetrics] = useState<TrainingMetric[]>([]);
  const [status, setStatus] = useState<'idle' | 'streaming' | 'completed'>('idle');

  useEffect(() => {
    if (!jobId) return;
    setStatus('streaming');
    const es = new EventSource(
      `${EVENTML_URL}/training/jobs/${jobId}/stream`,
      { headers: { Authorization: `Bearer ${token}` } }
    );

    es.addEventListener('progress', (e) => {
      setMetrics((prev) => [...prev, JSON.parse(e.data)]);
    });
    es.addEventListener('completed', (e) => {
      setMetrics((prev) => [...prev, JSON.parse(e.data)]);
      setStatus('completed');
      es.close();
    });
    return () => es.close();
  }, [jobId]);

  return { metrics, status };
};

import { EventmlClient } from '@eventzr/eventml-client';

const client = new EventmlClient({
  baseUrl: process.env.NEXT_PUBLIC_EVENTML_URL,
  tenantId: user.tenantId,
});

// Create model
const model = await client.createModel(tenantId, {
  name: 'event-recommender',
  framework: 'PYTORCH',
  taskType: 'RECOMMENDATION',
  modelType: 'CUSTOM',
});

// Start training
const job = await client.createTrainingJob(tenantId, {
  modelId: model.id,
  jobType: 'FULL_TRAINING',
  resourceConfig: { instanceType: 'ml.m5.large', instanceCount: 1 },
  hyperparameters: { learningRate: 0.001, epochs: 50 },
});

// Real-time prediction
const prediction = await client.predict(tenantId, {
  endpointId: 'endpoint-uuid',
  input: { features: [1.0, 2.5, 3.7] },
});