[P1] Document SLOs/SLIs for each service

[wonderwomancode]

Problem

No defined SLOs (Service Level Objectives). Can't measure if services are "healthy enough" or need improvement.

Acceptance Criteria

• SLOs defined for all 4 core services
• SLIs (indicators) measurable with current/planned tooling
• Error budgets calculated monthly
• SLO dashboard accessible to team
• Alert when error budget < 20% remaining

Proposed SLOs

service-cloud-api (GraphQL)

SLI	Target	Measurement
Availability	99.9% (43 min/month downtime)	Health check success rate
Latency P95	< 200ms	Proxy histogram
Latency P99	< 500ms	Proxy histogram
Error Rate	< 0.1%	5xx responses / total
Throughput	Handle 100 RPS	Load test baseline

service-auth

SLI	Target	Measurement
Availability	99.95% (22 min/month)	Health check
Login Success	> 99%	Success / attempts
Token Issuance	< 50ms P95	Response time
Error Rate	< 0.05%	Failed auth / total

infrastructure-proxy (Pingap)

SLI	Target	Measurement
Availability	99.99% (4 min/month)	TCP health check
Latency P99	< 50ms added	Proxy metrics
TLS Handshake	< 100ms	Proxy metrics
Error Rate	< 0.01%	5xx / total

service-secrets (Infisical)

SLI	Target	Measurement
Availability	99.9%	Health check
Secret Fetch	< 100ms P95	API response
Encryption	100% at rest	Audit

Error Budget Calculation

Monthly Error Budget = (1 - SLO) × 43,200 minutes

Example: 99.9% availability
Budget = 0.1% × 43,200 = 43.2 minutes of downtime allowed

Implementation

1. Create SLO documentation

File: SLO.md

# AlternateFutures Service Level Objectives

## service-cloud-api
- Availability: 99.9% (measured by synthetic health checks)
- Latency P95: 200ms (measured by proxy histogram)
...

2. Create SLO tracking workflow

File: .github/workflows/slo-report.yml

name: Weekly SLO Report

on:
  schedule:
    - cron: '0 9 * * 1'  # Monday 9am UTC

jobs:
  report:
    runs-on: ubuntu-latest
    steps:
      - name: Calculate SLIs
        run: |
          # Query Prometheus/metrics for last 7 days
          # Calculate availability, latency percentiles
          
      - name: Generate report
        run: |
          # Compare to SLOs, calculate error budget remaining
          
      - name: Post to Discord
        run: |
          # Send weekly SLO summary

3. SLO Dashboard (Future)

Once Grafana deployed:

• Create SLO dashboard panel
• Show current vs target
• Error budget burn rate

Required Before Implementation

• Prometheus metrics enabled (Build self-healing backup service for platform databases #1 in infrastructure-proxy)
• Health checks on all services (#108 in service-cloud-api)
• Centralized alerting ([P1] Create centralized alerting system #2 in this repo)

Testing

# Verify SLIs are measurable
curl https://api.alternatefutures.ai/health  # Availability
curl http://proxy:3018/metrics | grep latency  # Latency

Definition of Done

• SLO.md created with all service objectives
• SLIs are measurable (not aspirational)
• Team reviewed and agreed on targets
• Error budget tracking implemented
• Weekly SLO report workflow running