@cronicorn/mcp-server 1.6.1 → 1.7.1

package/dist/docs/use-cases.md

@@ -0,0 +1,208 @@
+---
+id: use-cases
+title: Use Cases & Examples
+description: Real-world scenarios showing how Cronicorn adapts to your application needs
+tags:
+  - user
+  - examples
+sidebar_position: 3
+mcp:
+  uri: file:///docs/use-cases.md
+  mimeType: text/markdown
+  priority: 0.8
+---
+
+# Use Cases & Examples
+
+Cronicorn's adaptive scheduling shines in scenarios where static cron jobs fall short. Here are real-world examples across different industries.
+
+## E-Commerce: Flash Sale Monitoring
+
+You're running a Black Friday sale and traffic spikes 5× normal volume. Your checkout system needs close monitoring, but you don't want to manually adjust schedules every hour.
+
+**Setup:**
+
+| Endpoint | Baseline | Purpose |
+|----------|----------|---------|
+| Traffic Monitor | Every 5 min | Track visitors and page load times |
+| Order Processor Health | Every 3 min | Monitor checkout performance |
+| Inventory Sync Check | Every 10 min | Ensure stock accuracy |
+
+**What Happens:**
+
+- **9am (sale starts)**: Traffic surges detected. AI tightens monitoring to every 30 seconds for the next hour.
+- **10am (struggling)**: Slow pages detected. AI activates cache warm-up (one-shot) and increases monitoring.
+- **11am (recovering)**: Performance stabilizes. AI gradually relaxes back to baseline over 30 minutes.
+- **2pm (stable)**: All endpoints return to normal 3-10 minute intervals.
+
+**Result**: Your team gets real-time visibility during the surge without manual schedule changes. The system backs off automatically when things stabilize.
+
+
+## Marketing: Content Publishing & Engagement Tracking
+
+You publish blog posts and want to monitor engagement closely for the first 24 hours, then back off to daily checks.
+
+**Setup:**
+
+| Endpoint | Baseline | Behavior |
+|----------|----------|----------|
+| Publish Post | Cron: "0 9 * * MON" | Publishes every Monday at 9am |
+| Track Analytics | Every 30 min | Page views, time on page |
+| Check Email Opens | Every hour | Newsletter engagement |
+| Promote Top Posts | Paused | Activates when engagement crosses threshold |
+
+**What Happens:**
+
+- **Monday 9am**: Post publishes on schedule
+- **9am-6pm**: AI tightens analytics tracking to every 5 minutes (fresh content gets attention)
+- **Tuesday-Friday**: AI relaxes to hourly checks (engagement is stable)
+- **Friday**: High engagement detected, AI activates promotional boost endpoint
+- **Next Monday**: Analytics reset to baseline 30 min intervals
+
+**Result**: You get real-time engagement data when it matters most, without paying for excessive API calls once traffic stabilizes.
+
+## DevOps: Infrastructure Health & Auto-Remediation
+
+Your SaaS platform runs background monitoring that should attempt automatic fixes before paging engineers at 2am.
+
+**Setup:**
+
+- **Health checks** run continuously: API latency, error rates, database connections, queue depth
+- **Investigation endpoints** start paused: slow query logs, memory profilers, distributed traces
+- **Remediation actions** fire as one-shots: restart unhealthy pods, flush caches, scale workers
+- **Alerts** escalate intelligently: Slack ops → Slack support → PagerDuty on-call
+
+**What Happens:**
+
+1. Database latency increases above threshold
+2. AI activates slow query investigation (was paused)
+3. Investigation identifies problematic query
+4. System fires one-shot remediation to kill long-running queries
+5. If fixed: Alert sent to Slack, investigation pauses again
+6. If not fixed after 3 attempts: Page on-call engineer
+
+**Result**: Most issues self-heal without waking anyone up. Engineers only get paged when automation can't resolve the problem.
+
+## Data Engineering: ETL Pipeline Coordination
+
+You run nightly data syncs from Salesforce, transform the data, and load it into your warehouse. The transformation step should only run after extraction succeeds.
+
+**Setup:**
+
+| Endpoint | Schedule | Notes |
+|----------|----------|-------|
+| Extract from Salesforce | Cron: "0 2 * * *" | Daily at 2am |
+| Transform Records | Paused | Activates only after extraction |
+| Load to Warehouse | Paused | Activates after transformation |
+| Check Pipeline Lag | Every 2 min | Monitors data freshness |
+
+**What Happens:**
+
+- **2:00am**: Salesforce extraction starts
+- **2:15am**: Extraction completes successfully. AI activates transformation endpoint for immediate one-shot run.
+- **2:20am**: Transformation finishes. AI activates warehouse load for one-shot run.
+- **2:25am**: Full pipeline complete. Investigation endpoints pause until tomorrow.
+- **If extraction fails**: Transformation and loading stay paused. Alert fires.
+
+**Result**: You coordinate complex dependencies without writing custom orchestration code. Each step waits for upstream success.
+
+## SaaS Operations: Usage Monitoring & Billing
+
+You need to track customer API usage hourly, send quota warnings at specific thresholds, and generate invoices monthly.
+
+**Setup:**
+
+- **Usage tracking**: Aggregate API calls every hour
+- **Quota monitoring**: Check for overages every 10 minutes
+- **Warning emails**: One-shot sends at 80%, 90%, 100% of quota
+- **Invoice generation**: Cron runs on 1st of each month
+- **Payment reminders**: Escalating sequence (3 days → 7 days → 14 days → 30 days overdue)
+
+**What Happens:**
+
+- **Throughout month**: Usage tracked hourly, quota checked every 10 min
+- **Customer approaching limit**: AI fires one-shot warning emails at each threshold
+- **1st of month at midnight**: Invoices generated for all accounts
+- **Payment not received**: Reminder sequence starts automatically (paused between emails)
+- **Payment received**: AI pauses further reminders
+
+**Result**: Billing runs like clockwork with minimal manual intervention. Customers get timely warnings, and dunning sequences are fully automated.
+
+## Web Scraping: Competitive Price Monitoring
+
+You track competitor pricing across multiple sites, but need to respect their rate limits and adapt when they change.
+
+**Setup:**
+
+| Endpoint | Baseline | Constraints |
+|----------|----------|-------------|
+| Scrape Prices | Every 4 hours | Min: 30 min, Max: 12 hours |
+| Validate Data | After each scrape | Checks for blocks/CAPTCHAs |
+| Check Proxy Health | Every 5 min | Monitors rotating proxies |
+| Alert Price Drop | Paused | Fires when competitor drops price |
+
+**What Happens:**
+
+- **Normal operation**: Scraping every 4 hours
+- **Rate limit warning detected**: AI backs off to every 6 hours for next day
+- **Competitor sale detected**: AI tightens to every 30 minutes (respects min interval)
+- **CAPTCHA encountered**: Validation fails, AI pauses scraping for 2 hours
+- **Proxy recovery**: Health check passes, AI resumes scraping
+
+**Result**: You adapt to site changes automatically. Scrapers respect rate limits, pause when blocked, and intensify during competitor sales.
+
+## Common Patterns
+
+### Adaptive Monitoring Frequency
+
+Start with a baseline, tighten when issues detected, relax when stable:
+
+```
+Baseline: Every 5 minutes
+During issues: Every 30 seconds (respects min interval)
+When stable: Every 15 minutes (respects max interval)
+```
+
+### Conditional Endpoint Activation
+
+Keep expensive operations paused until needed:
+
+```
+Investigation endpoints: Paused by default
+Activate: When health checks fail
+Deactivate: When issues resolve or timeout expires
+```
+
+### One-Shot Actions with Cooldowns
+
+Fire immediate actions that shouldn't repeat too quickly:
+
+```
+Cache warm-up: Fire once now, don't repeat for 15 minutes
+Alert escalation: Send once, pause for 1 hour before next
+Scale workers: Trigger immediately, cooldown for 20 minutes
+```
+
+### Calendar-Based + Adaptive Hybrid
+
+Use cron for time-sensitive tasks, intervals for everything else:
+
+```
+Daily reports: Cron "0 9 * * *" (always 9am)
+Monitoring: Interval 300000ms (adapts between min/max)
+Monthly billing: Cron "0 0 1 * *" (1st of month)
+```
+
+## When to Use Cronicorn
+
+**Great fit:**
+- You adjust cron schedules manually based on metrics
+- You need backoff logic after failures
+- You want tight monitoring during critical periods
+- You coordinate multiple dependent endpoints
+- You respect rate limits that vary by usage
+
+**Probably overkill:**
+- Simple daily backup at 2am (basic cron is fine)
+- Completely static schedules that never change
+- Sub-second precision required (use a real-time system)