---
id: "ET-ML-005"
title: "Integración con Backend"
type: "Technical Specification"
status: "Done"
priority: "Alta"
epic: "OQI-006"
project: "trading-platform"
version: "1.0.0"
created_date: "2025-12-05"
updated_date: "2026-01-04"
---

# ET-ML-005: Integración con Backend

## Metadata

| Campo | Valor |
|-------|-------|
| **ID** | ET-ML-005 |
| **Épica** | OQI-006 - Señales ML |
| **Tipo** | Especificación Técnica |
| **Versión** | 1.0.0 |
| **Estado** | Pendiente |
| **Última actualización** | 2025-12-05 |

---

## Propósito

Especificar la integración entre el ML Engine (Python/FastAPI) y el Backend principal (Express.js), incluyendo comunicación HTTP, eventos en tiempo real, y sincronización de datos.

---

## Arquitectura de Integración

```
┌─────────────────────────────────────────────────────────────────────────────┐
│                              FRONTEND (React)                                │
│                                                                              │
│   ┌──────────────┐     ┌──────────────┐     ┌──────────────────────────┐   │
│   │   Charts     │     │   Signals    │     │      Dashboard           │   │
│   │  Component   │     │   Display    │     │      Component           │   │
│   └──────┬───────┘     └──────┬───────┘     └───────────┬──────────────┘   │
│          │                    │                         │                   │
└──────────┼────────────────────┼─────────────────────────┼───────────────────┘
           │                    │                         │
           ▼                    ▼                         ▼
┌──────────────────────────────────────────────────────────────────────────────┐
│                         BACKEND (Express.js)                                  │
│                                                                               │
│   ┌─────────────────────────────────────────────────────────────────────┐   │
│   │                        API Gateway Layer                             │   │
│   │   ┌──────────────┐   ┌──────────────┐   ┌────────────────────────┐ │   │
│   │   │  /ml/signals │   │ /ml/predict  │   │   /ml/indicators       │ │   │
│   │   └──────┬───────┘   └──────┬───────┘   └───────────┬────────────┘ │   │
│   └──────────┼──────────────────┼───────────────────────┼───────────────┘   │
│              │                  │                       │                    │
│   ┌──────────▼──────────────────▼───────────────────────▼───────────────┐   │
│   │                       ML Integration Service                         │   │
│   │   ┌─────────────┐   ┌──────────────┐   ┌──────────────────────────┐│   │
│   │   │  ML Client  │   │ Rate Limiter │   │   Response Transformer   ││   │
│   │   │   (HTTP)    │   │  (per user)  │   │                          ││   │
│   │   └──────┬──────┘   └──────────────┘   └──────────────────────────┘│   │
│   └──────────┼───────────────────────────────────────────────────────────┘   │
│              │                                                               │
│   ┌──────────▼───────────────────────────────────────────────────────────┐   │
│   │                        Event Bus (Redis Pub/Sub)                      │   │
│   │                                                                       │   │
│   │   Channels: signals:BTCUSDT, signals:ETHUSDT, predictions:*          │   │
│   └───────────────────────────────────────────────────────────────────────┘   │
│                                                                               │
└───────────────────────────────────────────────────────────────────────────────┘
           │                           │
           ▼                           ▼
┌──────────────────────────────────────────────────────────────────────────────┐
│                           ML ENGINE (FastAPI)                                 │
│                                                                               │
│   ┌─────────────────────────────────────────────────────────────────────┐   │
│   │                          REST API                                    │   │
│   │   POST /predictions    POST /signals    GET /indicators              │   │
│   └─────────────────────────────────────────────────────────────────────┘   │
│                                                                               │
│   ┌─────────────────────────────────────────────────────────────────────┐   │
│   │                     Signal Publisher (Background)                    │   │
│   │                     Publishes to Redis Pub/Sub                       │   │
│   └─────────────────────────────────────────────────────────────────────┘   │
│                                                                               │
└───────────────────────────────────────────────────────────────────────────────┘
```

---

## ML Client (Backend)

### HTTP Client Service

```typescript
// src/services/ml/ml-client.service.ts
import axios, { AxiosInstance, AxiosError } from 'axios';
import { Injectable } from '@nestjs/common';
import { ConfigService } from '@nestjs/config';
import { RedisService } from '../redis/redis.service';

interface PredictionRequest {
  symbol: string;
  horizon: number;
}

interface SignalRequest {
  symbol: string;
  horizon: number;
  includeRange?: boolean;
  includeTpsl?: boolean;
}

interface MLResponse<T> {
  success: boolean;
  data: T;
  metadata: {
    requestId: string;
    latencyMs: number;
    cached: boolean;
  };
}

@Injectable()
export class MLClientService {
  private client: AxiosInstance;
  private readonly cachePrefix = 'ml:cache:';
  private readonly defaultCacheTTL = 30; // seconds

  constructor(
    private config: ConfigService,
    private redis: RedisService,
  ) {
    this.client = axios.create({
      baseURL: this.config.get('ML_ENGINE_URL'),
      timeout: 30000,
      headers: {
        'X-API-Key': this.config.get('ML_API_KEY'),
        'Content-Type': 'application/json',
      },
    });

    // Add response interceptor for logging
    this.client.interceptors.response.use(
      (response) => {
        console.log(`ML API: ${response.config.url} - ${response.status} - ${response.data?.metadata?.latencyMs}ms`);
        return response;
      },
      (error: AxiosError) => {
        console.error(`ML API Error: ${error.config?.url} - ${error.response?.status}`);
        throw error;
      }
    );
  }

  /**
   * Get price prediction with caching
   */
  async getPrediction(request: PredictionRequest): Promise<MLResponse<PredictionData>> {
    const cacheKey = `${this.cachePrefix}prediction:${request.symbol}:${request.horizon}`;

    // Check cache first
    const cached = await this.redis.get(cacheKey);
    if (cached) {
      return {
        success: true,
        data: JSON.parse(cached),
        metadata: { requestId: 'cached', latencyMs: 0, cached: true }
      };
    }

    // Call ML Engine
    const response = await this.client.post<MLResponse<PredictionData>>(
      '/predictions',
      {
        symbol: request.symbol,
        horizon: request.horizon,
      }
    );

    // Cache response
    await this.redis.setex(
      cacheKey,
      this.defaultCacheTTL,
      JSON.stringify(response.data.data)
    );

    return response.data;
  }

  /**
   * Get trading signal
   */
  async getSignal(request: SignalRequest): Promise<MLResponse<SignalData>> {
    const response = await this.client.post<MLResponse<SignalData>>(
      '/signals',
      {
        symbol: request.symbol,
        horizon: request.horizon,
        include_range: request.includeRange ?? true,
        include_tpsl: request.includeTpsl ?? true,
      }
    );

    return response.data;
  }

  /**
   * Get technical indicators
   */
  async getIndicators(symbol: string): Promise<MLResponse<IndicatorData>> {
    const cacheKey = `${this.cachePrefix}indicators:${symbol}`;

    const cached = await this.redis.get(cacheKey);
    if (cached) {
      return {
        success: true,
        data: JSON.parse(cached),
        metadata: { requestId: 'cached', latencyMs: 0, cached: true }
      };
    }

    const response = await this.client.get<MLResponse<IndicatorData>>(
      `/indicators?symbol=${symbol}`
    );

    await this.redis.setex(cacheKey, 10, JSON.stringify(response.data.data));

    return response.data;
  }

  /**
   * Get signal history
   */
  async getSignalHistory(params: SignalHistoryParams): Promise<SignalHistoryResponse> {
    const queryParams = new URLSearchParams({
      symbol: params.symbol,
      ...(params.horizon && { horizon: params.horizon.toString() }),
      ...(params.type && { type: params.type }),
      limit: (params.limit || 50).toString(),
      offset: (params.offset || 0).toString(),
    });

    const response = await this.client.get<SignalHistoryResponse>(
      `/signals/history?${queryParams}`
    );

    return response.data;
  }

  /**
   * Health check
   */
  async healthCheck(): Promise<boolean> {
    try {
      const response = await this.client.get('/health', { timeout: 5000 });
      return response.data.status === 'healthy';
    } catch {
      return false;
    }
  }
}
```

---

## Rate Limiting por Usuario

```typescript
// src/services/ml/ml-rate-limiter.service.ts
import { Injectable, ForbiddenException } from '@nestjs/common';
import { RedisService } from '../redis/redis.service';

interface RateLimitConfig {
  free: number;
  basic: number;
  pro: number;
  premium: number;
}

@Injectable()
export class MLRateLimiterService {
  private readonly limits: Record<string, RateLimitConfig> = {
    predictions: { free: 3, basic: 10, pro: 100, premium: 1000 },
    signals: { free: 3, basic: 10, pro: 100, premium: 1000 },
    indicators: { free: 10, basic: 50, pro: 500, premium: 5000 },
  };

  private readonly windowSeconds = 86400; // 24 hours

  constructor(private redis: RedisService) {}

  /**
   * Check if user can make request
   */
  async checkLimit(
    userId: string,
    userPlan: string,
    endpoint: 'predictions' | 'signals' | 'indicators'
  ): Promise<{ allowed: boolean; remaining: number; resetAt: Date }> {
    const key = `ratelimit:ml:${endpoint}:${userId}`;
    const limit = this.limits[endpoint][userPlan] || this.limits[endpoint].free;

    const current = await this.redis.incr(key);

    if (current === 1) {
      await this.redis.expire(key, this.windowSeconds);
    }

    const ttl = await this.redis.ttl(key);
    const resetAt = new Date(Date.now() + ttl * 1000);

    return {
      allowed: current <= limit,
      remaining: Math.max(0, limit - current),
      resetAt,
    };
  }

  /**
   * Enforce rate limit (throws if exceeded)
   */
  async enforce(
    userId: string,
    userPlan: string,
    endpoint: 'predictions' | 'signals' | 'indicators'
  ): Promise<void> {
    const { allowed, remaining, resetAt } = await this.checkLimit(userId, userPlan, endpoint);

    if (!allowed) {
      throw new ForbiddenException({
        message: 'Rate limit exceeded',
        limit: this.limits[endpoint][userPlan],
        remaining: 0,
        resetAt: resetAt.toISOString(),
        upgradeUrl: '/subscription/upgrade',
      });
    }
  }

  /**
   * Get current usage stats
   */
  async getUsage(userId: string, userPlan: string): Promise<MLUsageStats> {
    const endpoints = ['predictions', 'signals', 'indicators'] as const;
    const usage: Record<string, { used: number; limit: number; remaining: number }> = {};

    for (const endpoint of endpoints) {
      const key = `ratelimit:ml:${endpoint}:${userId}`;
      const current = parseInt(await this.redis.get(key) || '0');
      const limit = this.limits[endpoint][userPlan] || this.limits[endpoint].free;

      usage[endpoint] = {
        used: current,
        limit,
        remaining: Math.max(0, limit - current),
      };
    }

    return usage;
  }
}
```

---

## API Gateway (Express Routes)

```typescript
// src/routes/ml.routes.ts
import { Router } from 'express';
import { authenticate } from '../middleware/auth';
import { MLController } from '../controllers/ml.controller';

const router = Router();
const controller = new MLController();

// All routes require authentication
router.use(authenticate);

/**
 * POST /api/ml/predictions
 * Get price prediction
 */
router.post('/predictions', controller.getPrediction);

/**
 * POST /api/ml/signals
 * Get trading signal
 */
router.post('/signals', controller.getSignal);

/**
 * GET /api/ml/signals/history
 * Get signal history
 */
router.get('/signals/history', controller.getSignalHistory);

/**
 * GET /api/ml/indicators/:symbol
 * Get technical indicators
 */
router.get('/indicators/:symbol', controller.getIndicators);

/**
 * GET /api/ml/usage
 * Get user's ML API usage stats
 */
router.get('/usage', controller.getUsage);

export default router;
```

```typescript
// src/controllers/ml.controller.ts
import { Request, Response, NextFunction } from 'express';
import { MLClientService } from '../services/ml/ml-client.service';
import { MLRateLimiterService } from '../services/ml/ml-rate-limiter.service';

export class MLController {
  private mlClient: MLClientService;
  private rateLimiter: MLRateLimiterService;

  constructor() {
    this.mlClient = new MLClientService();
    this.rateLimiter = new MLRateLimiterService();
  }

  /**
   * Get price prediction
   */
  getPrediction = async (req: Request, res: Response, next: NextFunction) => {
    try {
      const { userId, plan } = req.user;
      const { symbol, horizon } = req.body;

      // Validate input
      if (!symbol || !horizon) {
        return res.status(400).json({
          success: false,
          error: 'symbol and horizon are required',
        });
      }

      // Check rate limit
      await this.rateLimiter.enforce(userId, plan, 'predictions');

      // Get prediction from ML Engine
      const prediction = await this.mlClient.getPrediction({ symbol, horizon });

      // Transform response for frontend
      res.json({
        success: true,
        data: this.transformPrediction(prediction.data),
        metadata: prediction.metadata,
      });
    } catch (error) {
      next(error);
    }
  };

  /**
   * Get trading signal
   */
  getSignal = async (req: Request, res: Response, next: NextFunction) => {
    try {
      const { userId, plan } = req.user;
      const { symbol, horizon, includeRange, includeTpsl } = req.body;

      await this.rateLimiter.enforce(userId, plan, 'signals');

      const signal = await this.mlClient.getSignal({
        symbol,
        horizon,
        includeRange,
        includeTpsl,
      });

      // Log signal for analytics
      await this.logSignal(userId, signal.data);

      res.json({
        success: true,
        data: this.transformSignal(signal.data),
        metadata: signal.metadata,
      });
    } catch (error) {
      next(error);
    }
  };

  /**
   * Get technical indicators
   */
  getIndicators = async (req: Request, res: Response, next: NextFunction) => {
    try {
      const { userId, plan } = req.user;
      const { symbol } = req.params;

      await this.rateLimiter.enforce(userId, plan, 'indicators');

      const indicators = await this.mlClient.getIndicators(symbol);

      res.json({
        success: true,
        data: indicators.data,
        metadata: indicators.metadata,
      });
    } catch (error) {
      next(error);
    }
  };

  /**
   * Get signal history
   */
  getSignalHistory = async (req: Request, res: Response, next: NextFunction) => {
    try {
      const { symbol, horizon, type, limit, offset } = req.query;

      const history = await this.mlClient.getSignalHistory({
        symbol: symbol as string,
        horizon: horizon ? parseInt(horizon as string) : undefined,
        type: type as string,
        limit: limit ? parseInt(limit as string) : 50,
        offset: offset ? parseInt(offset as string) : 0,
      });

      res.json(history);
    } catch (error) {
      next(error);
    }
  };

  /**
   * Get ML API usage stats
   */
  getUsage = async (req: Request, res: Response, next: NextFunction) => {
    try {
      const { userId, plan } = req.user;
      const usage = await this.rateLimiter.getUsage(userId, plan);

      res.json({
        success: true,
        data: {
          plan,
          usage,
          resetAt: this.getNextResetTime(),
        },
      });
    } catch (error) {
      next(error);
    }
  };

  private transformPrediction(data: any) {
    return {
      ...data,
      formattedPredictedHigh: `$${data.predicted_high.toLocaleString()}`,
      formattedPredictedLow: `$${data.predicted_low.toLocaleString()}`,
    };
  }

  private transformSignal(data: any) {
    return {
      ...data,
      signalEmoji: this.getSignalEmoji(data.signal.type),
      confidenceLabel: this.getConfidenceLabel(data.signal.confidence),
    };
  }

  private getSignalEmoji(type: string): string {
    const emojis = { buy: '📈', sell: '📉', hold: '⏸️' };
    return emojis[type] || '❓';
  }

  private getConfidenceLabel(confidence: number): string {
    if (confidence >= 0.8) return 'Very High';
    if (confidence >= 0.65) return 'High';
    if (confidence >= 0.5) return 'Medium';
    return 'Low';
  }

  private async logSignal(userId: string, signal: any) {
    // Log to database for analytics
  }

  private getNextResetTime(): string {
    const tomorrow = new Date();
    tomorrow.setUTCHours(0, 0, 0, 0);
    tomorrow.setDate(tomorrow.getDate() + 1);
    return tomorrow.toISOString();
  }
}
```

---

## Eventos en Tiempo Real (Redis Pub/Sub)

### Publisher (ML Engine)

```python
# app/tasks/signal_publisher.py
import asyncio
import json
from redis import asyncio as aioredis
from datetime import datetime
from app.services.signal_generator import SignalGeneratorService
from app.config.settings import settings

class SignalPublisher:
    """Publishes signals to Redis for real-time distribution"""

    def __init__(self):
        self.redis = None
        self.signal_gen = SignalGeneratorService()
        self.symbols = settings.SUPPORTED_SYMBOLS
        self.horizons = [6, 18]  # Only short-term for real-time
        self.interval = 60  # seconds between signal generation

    async def start(self):
        """Start the signal publisher background task"""
        self.redis = await aioredis.from_url(settings.REDIS_URL)

        while True:
            try:
                await self.publish_signals()
            except Exception as e:
                print(f"Error publishing signals: {e}")

            await asyncio.sleep(self.interval)

    async def publish_signals(self):
        """Generate and publish signals for all symbols"""
        for symbol in self.symbols:
            for horizon in self.horizons:
                try:
                    signal = await self.signal_gen.generate(
                        symbol=symbol,
                        horizon=horizon,
                        include_range=True,
                        include_tpsl=True
                    )

                    # Publish to Redis channel
                    channel = f"signals:{symbol}"
                    message = json.dumps({
                        'type': 'signal',
                        'data': signal.dict(),
                        'timestamp': datetime.utcnow().isoformat()
                    })

                    await self.redis.publish(channel, message)

                except Exception as e:
                    print(f"Error generating signal for {symbol}: {e}")

    async def stop(self):
        """Stop and cleanup"""
        if self.redis:
            await self.redis.close()
```

### Subscriber (Backend)

```typescript
// src/services/ml/signal-subscriber.service.ts
import Redis from 'ioredis';
import { Server as SocketServer } from 'socket.io';
import { Injectable, OnModuleInit, OnModuleDestroy } from '@nestjs/common';

@Injectable()
export class SignalSubscriberService implements OnModuleInit, OnModuleDestroy {
  private subscriber: Redis;
  private io: SocketServer;
  private channels: string[] = ['signals:BTCUSDT', 'signals:ETHUSDT'];

  constructor() {
    this.subscriber = new Redis(process.env.REDIS_URL);
  }

  async onModuleInit() {
    await this.subscribe();
  }

  async onModuleDestroy() {
    await this.subscriber.quit();
  }

  setSocketServer(io: SocketServer) {
    this.io = io;
  }

  private async subscribe() {
    for (const channel of this.channels) {
      await this.subscriber.subscribe(channel);
    }

    this.subscriber.on('message', (channel: string, message: string) => {
      this.handleMessage(channel, message);
    });

    console.log(`Subscribed to ML signal channels: ${this.channels.join(', ')}`);
  }

  private handleMessage(channel: string, message: string) {
    try {
      const data = JSON.parse(message);
      const symbol = channel.split(':')[1];

      // Broadcast to connected clients subscribed to this symbol
      this.io?.to(`signals:${symbol}`).emit('signal', {
        symbol,
        ...data,
      });

      // Log for monitoring
      console.log(`Signal received: ${symbol} - ${data.data?.signal?.type}`);
    } catch (error) {
      console.error('Error handling ML signal:', error);
    }
  }
}
```

### WebSocket Integration

```typescript
// src/websocket/signals.gateway.ts
import {
  WebSocketGateway,
  WebSocketServer,
  SubscribeMessage,
  OnGatewayConnection,
  OnGatewayDisconnect,
} from '@nestjs/websockets';
import { Server, Socket } from 'socket.io';
import { SignalSubscriberService } from '../services/ml/signal-subscriber.service';

@WebSocketGateway({ namespace: '/signals' })
export class SignalsGateway implements OnGatewayConnection, OnGatewayDisconnect {
  @WebSocketServer()
  server: Server;

  constructor(private signalSubscriber: SignalSubscriberService) {}

  afterInit(server: Server) {
    this.signalSubscriber.setSocketServer(server);
  }

  handleConnection(client: Socket) {
    console.log(`Client connected to signals: ${client.id}`);
  }

  handleDisconnect(client: Socket) {
    console.log(`Client disconnected from signals: ${client.id}`);
  }

  @SubscribeMessage('subscribe')
  handleSubscribe(client: Socket, symbol: string) {
    client.join(`signals:${symbol}`);
    client.emit('subscribed', { symbol });
    console.log(`Client ${client.id} subscribed to ${symbol}`);
  }

  @SubscribeMessage('unsubscribe')
  handleUnsubscribe(client: Socket, symbol: string) {
    client.leave(`signals:${symbol}`);
    client.emit('unsubscribed', { symbol });
  }
}
```

---

## Frontend Integration

```typescript
// src/hooks/useMLSignals.ts
import { useEffect, useState, useCallback } from 'react';
import { io, Socket } from 'socket.io-client';
import { useAuth } from './useAuth';

interface MLSignal {
  symbol: string;
  horizon: number;
  signal: {
    type: 'buy' | 'sell' | 'hold';
    confidence: number;
  };
  priceRange?: {
    current: number;
    predictedHigh: number;
    predictedLow: number;
  };
  timestamp: string;
}

export function useMLSignals(symbol: string) {
  const { token } = useAuth();
  const [signal, setSignal] = useState<MLSignal | null>(null);
  const [connected, setConnected] = useState(false);
  const [socket, setSocket] = useState<Socket | null>(null);

  useEffect(() => {
    const newSocket = io('/signals', {
      auth: { token },
      transports: ['websocket'],
    });

    newSocket.on('connect', () => {
      setConnected(true);
      newSocket.emit('subscribe', symbol);
    });

    newSocket.on('disconnect', () => {
      setConnected(false);
    });

    newSocket.on('signal', (data: MLSignal) => {
      if (data.symbol === symbol) {
        setSignal(data);
      }
    });

    setSocket(newSocket);

    return () => {
      newSocket.emit('unsubscribe', symbol);
      newSocket.disconnect();
    };
  }, [symbol, token]);

  return { signal, connected };
}

// Usage in component
function TradingChart({ symbol }) {
  const { signal, connected } = useMLSignals(symbol);

  return (
    <div>
      <ConnectionStatus connected={connected} />
      {signal && <SignalIndicator signal={signal} />}
      <Chart symbol={symbol} />
    </div>
  );
}
```

---

## Error Handling

```typescript
// src/middleware/ml-error-handler.ts
import { Request, Response, NextFunction } from 'express';
import { AxiosError } from 'axios';

export function mlErrorHandler(
  error: Error,
  req: Request,
  res: Response,
  next: NextFunction
) {
  // Handle Axios errors from ML Engine
  if (error instanceof AxiosError) {
    if (error.code === 'ECONNREFUSED') {
      return res.status(503).json({
        success: false,
        error: {
          code: 'ML_ENGINE_UNAVAILABLE',
          message: 'ML Engine is temporarily unavailable',
        },
      });
    }

    if (error.response) {
      return res.status(error.response.status).json({
        success: false,
        error: error.response.data?.error || {
          code: 'ML_ENGINE_ERROR',
          message: error.message,
        },
      });
    }
  }

  next(error);
}
```

---

## Health Check Integration

```typescript
// src/services/health/ml-health.service.ts
import { Injectable } from '@nestjs/common';
import { MLClientService } from '../ml/ml-client.service';

@Injectable()
export class MLHealthService {
  constructor(private mlClient: MLClientService) {}

  async check(): Promise<HealthCheckResult> {
    const startTime = Date.now();

    try {
      const healthy = await this.mlClient.healthCheck();
      const latency = Date.now() - startTime;

      return {
        name: 'ml-engine',
        status: healthy ? 'healthy' : 'unhealthy',
        latencyMs: latency,
      };
    } catch (error) {
      return {
        name: 'ml-engine',
        status: 'unhealthy',
        error: error.message,
        latencyMs: Date.now() - startTime,
      };
    }
  }
}
```

---

## Referencias

- [ET-ML-001: Arquitectura](./ET-ML-001-arquitectura.md)
- [ET-ML-004: FastAPI Endpoints](./ET-ML-004-api.md)
- [Socket.io Documentation](https://socket.io/docs/)

---

**Autor:** Requirements-Analyst
**Fecha:** 2025-12-05