speak-observability

name	speak-observability
description	Set up comprehensive observability for Speak integrations with metrics, traces, and alerts. Use when implementing monitoring for Speak operations, setting up dashboards, or configuring alerting for language learning feature health. Trigger with phrases like "speak monitoring", "speak metrics", "speak observability", "monitor speak", "speak alerts", "speak tracing".
allowed-tools	Read, Write, Edit
version	1.0.0
license	MIT
author	Jeremy Longshore <[e-mail]>

Speak Observability

Overview

Set up comprehensive observability for Speak language learning integrations.

Prerequisites

• Prometheus or compatible metrics backend
• OpenTelemetry SDK installed
• Grafana or similar dashboarding tool
• AlertManager configured

Key Metrics for Language Learning

Business Metrics

Metric	Type	Description
speak_lessons_started_total	Counter	Lessons initiated
speak_lessons_completed_total	Counter	Lessons completed
speak_lessons_abandoned_total	Counter	Lessons abandoned mid-session
speak_pronunciation_score	Histogram	Pronunciation scores distribution
speak_active_sessions	Gauge	Currently active lesson sessions
speak_daily_active_learners	Gauge	Unique learners today

Technical Metrics

Metric	Type	Description
speak_api_requests_total	Counter	Total API requests
speak_api_duration_seconds	Histogram	Request latency
speak_api_errors_total	Counter	Error count by type
speak_speech_recognition_duration_seconds	Histogram	Audio processing time
speak_rate_limit_remaining	Gauge	Rate limit headroom

Prometheus Metrics Implementation

import { Registry, Counter, Histogram, Gauge } from 'prom-client';

const registry = new Registry();

// Business metrics
const lessonsStarted = new Counter({
  name: 'speak_lessons_started_total',
  help: 'Total lessons started',
  labelNames: ['language', 'topic', 'difficulty'],
  registers: [registry],
});

const lessonsCompleted = new Counter({
  name: 'speak_lessons_completed_total',
  help: 'Total lessons completed',
  labelNames: ['language', 'topic', 'difficulty'],
  registers: [registry],
});

const pronunciationScore = new Histogram({
  name: 'speak_pronunciation_score',
  help: 'Pronunciation scores distribution',
  labelNames: ['language'],
  buckets: [40, 50, 60, 70, 80, 85, 90, , ],
  : [registry],
});

 activeSessions =  ({
  : ,
  : ,
  : [],
  : [registry],
});


 apiRequests =  ({
  : ,
  : ,
  : [, , ],
  : [registry],
});

 apiDuration =  ({
  : ,
  : ,
  : [, ],
  : [, , , , , , , ],
  : [registry],
});

 speechRecognitionDuration =  ({
  : ,
  : ,
  : [],
  : [, , , , , ],
  : [registry],
});

Instrumented Speak Client

async function instrumentedSpeakRequest<T>(
  method: string,
  endpoint: string,
  operation: () => Promise<T>
): Promise<T> {
  const timer = apiDuration.startTimer({ method, endpoint });

  try {
    const result = await operation();
    apiRequests.inc({ method, endpoint, status: 'success' });
    return result;
  } catch (error: any) {
    apiRequests.inc({ method, endpoint, status: 'error' });
    throw error;
  } finally {
    timer();
  }
}

// Lesson tracking
async function trackLessonStart(session: LessonSession): Promise<void> {
  lessonsStarted.inc({
    language: session.language,
    topic: session.topic,
    difficulty: session.difficulty,
  });
  activeSessions.inc({ language: session.language });
}

async function (): <> {
   labels = {
    : session.,
    : session.,
    : session.,
  };

   (summary.) {
    lessonsCompleted.(labels);
  }  {
    lessonsAbandoned.({
      ...labels,
      : summary. || ,
    });
  }

  activeSessions.({ : session. });

  
   (summary.) {
    pronunciationScore.(
      { : session. },
      summary.
    );
  }
}

Distributed Tracing

OpenTelemetry Setup

import { trace, SpanStatusCode, context, propagation } from '@opentelemetry/api';

const tracer = trace.getTracer('speak-service');

async function tracedSpeakCall<T>(
  operationName: string,
  attributes: Record<string, string>,
  operation: () => Promise<T>
): Promise<T> {
  return tracer.startActiveSpan(`speak.${operationName}`, async (span) => {
    span.setAttributes(attributes);

    try {
      const result = await operation();
      span.setStatus({ code: SpanStatusCode.OK });
      return result;
    } catch (error: any) {
      span.setStatus({ code: SpanStatusCode.ERROR, message: error.message });
      span.recordException(error);
      throw error;
    } finally {
      span.end();
    }
  });
}

// Trace a complete lesson session
async  (): <> {
   (
    ,
    {
      : userId,
      : config.,
      : config.,
    },
     () => {
       session =  speakService.(config);

      
       (!session.) {
         (
          ,
          { : session. },
           () => {
             prompt =  session.();
             response =  ();
             session.(response);
          }
        );
      }

       session.();
    }
  );
}

Structured Logging

import pino from 'pino';

const logger = pino({
  name: 'speak-service',
  level: process.env.LOG_LEVEL || 'info',
  formatters: {
    level: (label) => ({ level: label }),
  },
});

interface SpeakLogContext {
  service: 'speak';
  operation: string;
  sessionId?: string;
  userId?: string;
  language?: string;
  duration_ms?: number;
  pronunciationScore?: number;
  error?: any;
}

function logSpeakOperation(
  level: 'info' | 'warn' | 'error',
  operation: string,
  context: Partial<SpeakLogContext>
): void {
  const logContext: SpeakLogContext = {
    service: 'speak',
    operation,
    ...context,
  };

  logger[level](logContext, `Speak `);
}


(, , {
  : session.,
  : session.,
  : session.,
  : summary.,
  : summary.,
});

Alert Configuration

Prometheus AlertManager Rules

# speak_alerts.yaml
groups:
  - name: speak_alerts
    rules:
      # High error rate
      - alert: SpeakHighErrorRate
        expr: |
          rate(speak_api_errors_total[5m]) /
          rate(speak_api_requests_total[5m]) > 0.05
        for: 5m
        labels:
          severity: warning
          service: speak
        annotations:
          summary: "Speak API error rate > 5%"
          description: "Error rate is {{ $value | humanizePercentage }}"

      # Speech recognition slow
      - alert: SpeakSpeechRecognitionSlow
        expr: |
          histogram_quantile(0.95,
            rate(speak_speech_recognition_duration_seconds_bucket[5m])
          ) > 5
        for: 5m
        labels:
          severity: warning
          service: speak
        annotations:
          summary: "Speech recognition P95 latency > 5s"

      # Low lesson completion rate
      - alert: SpeakLowCompletionRate
        expr: |
          rate(speak_lessons_completed_total[1h]) /
          rate(speak_lessons_started_total[1h]) < 0.5
         
        
           
           
        
           

      
        
           
         
        
           
           
        
           

      
        
                  
        
           
           
        
           

Grafana Dashboard

{
  "title": "Speak Language Learning",
  "panels": [
    {
      "title": "Active Lesson Sessions",
      "type": "stat",
      "targets": [{
        "expr": "sum(speak_active_sessions)"
      }]
    },
    {
      "title": "Lessons Started vs Completed",
      "type": "graph",
      "targets": [
        { "expr": "rate(speak_lessons_started_total[5m])", "legendFormat": "Started" },
        { "expr": "rate(speak_lessons_completed_total[5m])", "legendFormat"  
      
    
    
       
       
       
         
      
    
    
       
       
       
             
             
             
      
    
    
       
       
       
         
      
    
    
       
       
       
         
      
    
  

Output

• Business and technical metrics
• Distributed tracing configured
• Structured logging implemented
• Alert rules deployed
• Grafana dashboard ready

Error Handling

Issue	Cause	Solution
Missing metrics	No instrumentation	Wrap client calls
Trace gaps	Missing propagation	Check context headers
Alert storms	Wrong thresholds	Tune alert rules
High cardinality	Too many labels	Reduce label values

Examples

Quick Metrics Endpoint

app.get('/metrics', async (req, res) => {
  res.set('Content-Type', registry.contentType);
  res.send(await registry.metrics());
});

Resources

• Prometheus Best Practices
• OpenTelemetry Documentation
• Speak Observability Guide

Next Steps

For incident response, see speak-incident-runbook.