Skip to main content

Infrastructure Overview

FLYPILOT has production-ready infrastructure that all agents must use. This infrastructure prevents wheel reinventing, enables cross-session learning, and eliminates API errors.

What We Have​

As of 2026-02-03, we have 6 production systems:

1. Decision Management ✅​

Document all architectural decisions as markdown files. Enables cross-session learning.

# Search before deciding
grep -ril "topic" docs/planning/decisions/

# Record a decision
cat > docs/planning/decisions/$(date +%Y-%m-%d)-my-decision.md << 'EOF'
# Decision: [Title]
**Date:** YYYY-MM-DD
**Impact:** high
**Category:** architecture
**Department:** fp

## Decision
[What was decided.]

## Rationale
[Why.]
EOF

Why mandatory: Prevents 30-60 min/session of redoing work by reviewing past decisions first.

2. Sync Operations ✅​

Get data from Asana/Shortcut/Linear without writing custom API clients.

sync.cli sync --profile quick     # 7 days, incremental
sync.cli sync --profile omg       # OMG only

Features:

  • 10 pre-configured sync profiles for all use cases
  • Incremental sync (90-95% fewer API calls)
  • Rate limiting built-in
  • Tested and production-ready

3. Rate Limiting ✅​

Automatic API throttling with 80% safety margin. Zero errors.

You don't do anything - it's automatic. The sync system handles all rate limiting.

Rate limits (per hour):

  • Asana: 1500 requests
  • Shortcut: 1000 requests
  • Linear: 1000 requests

4. Memory & Research Notes ✅​

Store research findings, technical discoveries, and important insights.

  • AI platform memory — Claude Code and Gemini manage project memory natively within sessions.
  • Persistent notes — Write findings to docs/planning/research-notes.md for cross-session, cross-platform access.
# Add a finding
echo "- $(date +%Y-%m-%d): Finding — tags: topic" >> docs/planning/research-notes.md

# Search findings
grep -i "keyword" docs/planning/research-notes.md

5. Session Continuity ✅​

Continue work across machines and sessions via handoff files and AI platform context.

  • AI platforms (Claude Code, Gemini) handle session context natively.
  • Handoff files (handoffs/) provide structured continuity for ongoing tasks across sessions or between agents.
# Check for active handoffs
ls handoffs/

# Create a handoff
cat > handoffs/session-$(date +%Y-%m-%d).md << 'EOF'
## Context / Next Steps
[What to pick up and where to go]
EOF

6. WPEngine Monitoring ✅​

228 production sites monitored every 2 hours.

Alerts to Slack #tech-alerts for:

  • SSL certificates expiring
  • Backup failures
  • Disk space issues
  • Performance degradation

Status: Production (Mac Studio cron deployed)

7. Sync Profiles ✅​

10 pre-configured sync commands for every use case:

  • quick - 7 days, fast updates
  • full - 365 days, complete backfill
  • omg - OMG department only
  • lm - Learned Media department only
  • fp - FLYPILOT department only
  • tasks-only - No time entries
  • time-only - No tasks
  • content-backfill - Comments/attachments
  • historical - 2 years of data
  • standard - Default (30 days)

The Mandatory 5-Step Workflow​

BEFORE EVERY SESSION:

  1. grep -ril "topic" docs/planning/decisions/ - Don't reinvent wheels
  2. Check handoffs/ for ongoing context

DURING WORK: 3. Use sync profiles (not custom API clients) 4. Record decisions (as markdown files in docs/planning/decisions/) 5. Document research (in docs/planning/research-notes.md or AI platform memory)

END OF SESSION: 6. Write a handoff file if work is ongoing

Time investment: 2-3 min per session Time saved: 30-60 min per session (not redoing work) ROI: 10-20x

Why This Matters​

Before (Old Way)​

❌ Every orchestrator writes custom API clients ❌ Decisions buried in Slack/PRs ❌ Research lost when session ends ❌ Rate limit errors break automation ❌ No cross-orchestrator learning

After (New Way)​

✅ Sync profiles handle all API calls (tested, rate-limited) ✅ Decisions searchable as markdown files in git ✅ Research preserved in notes files and AI platform memory ✅ Zero rate limit errors (80% safety margin) ✅ Every session builds on past decisions

Impact: Infrastructure that compounds instead of duplicates.

Success Metrics​

After 1 week:

  • ✅ 10+ decisions recorded as markdown files
  • ✅ Research notes being populated
  • ✅ 0 rate limit errors
  • ✅ 0 infrastructure bypasses
  • ✅ Workflow feels automatic

After 1 month:

  • ✅ 30+ decisions (knowledge base growing)
  • ✅ Research notes library is useful
  • ✅ All active orchestrators follow the workflow
  • ✅ Infrastructure is muscle memory

Ready to start? Continue to the Getting Started Guide for Mike's hands-on learning path.