# SLA Definitions for Performance Monitoring
# Production Configuration
#
# Each SLA definition specifies performance targets and baseline percentiles
# Used to track and alert on performance degradation
#
# SLA Target: <X>ms = P50 should be below X ms
# Baseline P95: Performance threshold for 95th percentile
# Baseline P99: Performance threshold for 99th percentile
#
# See docs/PERFORMANCE_MONITORING_GUIDE.md for detailed configuration

slas:
  # Agent Startup SLA
  # Target: Initialize new agent in <2s
  # Used for: Agent spawning, initialization
  agent_startup:
    target: 2000           # ms - median response time target
    p95: 2000              # 95th percentile - 95% of startups should be <2s
    p99: 5000              # 99th percentile - 99% of startups should be <5s
    enabled: true
    alert_threshold: 1.2   # Alert if performance degrades >20%

  # Query Execution SLA
  # Target: Execute database queries in <5s
  # Used for: Database queries, API calls
  query_execution:
    target: 5000           # ms - median response time target
    p95: 5000              # 95th percentile - 95% of queries should be <5s
    p99: 10000             # 99th percentile - 99% of queries should be <10s
    enabled: true
    alert_threshold: 1.2   # Alert if performance degrades >20%

  # Skill Execution SLA
  # Target: Complete skill execution in <30s
  # Used for: Skill runs, custom operations
  skill_execution:
    target: 30000          # ms - median response time target
    p95: 30000             # 95th percentile - 95% of skills should complete in <30s
    p99: 60000             # 99th percentile - 99% of skills should complete in <60s
    enabled: true
    alert_threshold: 1.2   # Alert if performance degrades >20%

  # Transaction Commit SLA
  # Target: Commit database transaction in <5s
  # Used for: Database transactions, writes
  transaction_commit:
    target: 5000           # ms - median response time target
    p95: 5000              # 95th percentile - 95% of commits should be <5s
    p99: 10000             # 99th percentile - 99% of commits should be <10s
    enabled: true
    alert_threshold: 1.2   # Alert if performance degrades >20%

  # Log Shipment SLA
  # Target: Ship logs in <1s
  # Used for: Log aggregation, metrics export
  log_shipment:
    target: 1000           # ms - median shipment time
    p95: 1000              # 95th percentile
    p99: 2000              # 99th percentile
    enabled: true
    alert_threshold: 1.2   # Alert if performance degrades >20%

  # Checkpoint Save SLA
  # Target: Save checkpoint in <5s
  # Used for: State management, distributed coordination
  checkpoint_save:
    target: 5000           # ms - median save time
    p95: 5000              # 95th percentile
    p99: 10000             # 99th percentile
    enabled: true
    alert_threshold: 1.2   # Alert if performance degrades >20%

  # Cache Lookup SLA
  # Target: Cache lookup in <100ms
  # Used for: In-memory cache, rapid access patterns
  cache_lookup:
    target: 100            # ms - median lookup time
    p95: 100               # 95th percentile
    p99: 500               # 99th percentile
    enabled: true
    alert_threshold: 1.2   # Alert if performance degrades >20%

# Global SLA Configuration
global:
  # Monthly SLA violation threshold (target <5%)
  # This means we aim for >95% compliance monthly
  monthly_violation_target: 0.05

  # Degradation detection parameters
  degradation_window_minutes: 5      # Evaluate last 5 minutes of data
  degradation_threshold_percent: 20  # Alert if >20% degradation from baseline
  degradation_consecutive_samples: 5 # Require 5+ samples to declare degradation

  # Alert configuration
  alert_resolution_hours: 24         # Auto-resolve alerts after 24h if not re-triggered
  alert_batch_window_minutes: 5      # Batch similar alerts within 5 minute window
  alert_max_frequency_per_hour: 10   # Max 10 alerts per operation per hour

  # Retention policy
  metrics_retention_days: 90          # Keep detailed metrics for 90 days
  alerts_retention_days: 365          # Keep alert history for 1 year
  baseline_calculation_days: 30       # Default baseline window (30 days of history)
  baseline_cache_ttl_minutes: 5       # Cache baseline calculations for 5 minutes

  # Performance targets for monitoring service itself
  metric_recording_overhead_ms: 10    # Recording should add <10ms overhead
  baseline_calculation_target_ms: 5000 # Calculation should complete in <5s
  sla_check_target_ms: 50             # SLA checks should complete in <50ms

# Alerting thresholds
alerting:
  # SLA violation alert conditions
  violation_alert:
    enabled: true
    level: warning                    # Alert level for SLA violations
    threshold: 1                       # Alert on every violation

  # Degradation alert conditions
  degradation_alert:
    enabled: true
    level: warning                    # Alert level for degradation
    threshold: 1.2                    # Alert if >20% degradation

  # Critical threshold alerts
  critical_alert:
    enabled: true
    level: critical                   # Alert level for critical threshold
    threshold: 1.5                    # Alert if >50% degradation

# Notification channels
notifications:
  # Email notifications for critical alerts
  email:
    enabled: true
    critical_only: false
    recipients: []                    # Configure in environment or secrets

  # Slack notifications for performance issues
  slack:
    enabled: true
    critical_only: true
    channel: '#performance-alerts'
    webhook_url: "${SLACK_WEBHOOK_PERFORMANCE}"  # Set via environment variable

  # PagerDuty for critical incidents
  pagerduty:
    enabled: false
    critical_only: true
    api_key: "${PAGERDUTY_API_KEY}"   # Set via environment variable