Step 6: Architecture Decision — Scheduling
When to Use Each Approach
| Trigger Type | Approach | Example |
|---|---|---|
| Time-based (daily, hourly) | LaunchAgent (macOS) or cron | Nightly data collection |
| Event-based (new email, webhook) | n8n or webhook listener | Email arrives -> classify |
| User-initiated | Slash command | /weekly-review |
| Continuous | Always-running daemon | Message bridge, file watcher |
Hybrid Pattern (Recommended)
LaunchAgent (schedule)
└── Python script (collect data)
└── Output to JSONL/markdown
└── Claude agent (reason over data)
└── Output: email draft, SMS, reportWhy hybrid? Scripts are cheap and reliable for data collection. Agent reasoning is expensive but powerful for classification and generation. Combine both.
Decision Template
For your workflow analysis document:
| Trigger Type | Approach | Examples |
|---|---|---|
| Time-based | LaunchAgent / cron | Your examples |
| Event-based | Webhook / n8n | Your examples |
| User-initiated | Slash command | Your examples |
Decision: LaunchAgents / Claude cron / Hybrid / Other
Rationale: …
Last updated on