lindy-cost-tuning skill

---
name: lindy-cost-tuning
description: |
  Optimize Lindy AI costs and manage usage efficiently.
  Use when reducing costs, analyzing usage patterns,
  or optimizing budget allocation.
  Trigger with phrases like "lindy cost", "lindy billing",
  "reduce lindy spend", "lindy budget".
allowed-tools: Read, Write, Edit
version: 1.0.0
license: MIT
author: Jeremy Longshore <[email protected]>
---

# Lindy Cost Tuning

## Overview
Optimize Lindy AI costs while maintaining service quality.

## Prerequisites
- Access to Lindy billing dashboard
- Usage data available
- Understanding of pricing tiers

## Instructions

### Step 1: Analyze Current Usage
```typescript
import { Lindy } from '@lindy-ai/sdk';

async function analyzeUsage() {
  const lindy = new Lindy({ apiKey: process.env.LINDY_API_KEY });

  const usage = await lindy.usage.monthly({
    startDate: '2025-01-01',
    endDate: '2025-01-31',
  });

  const analysis = {
    totalRuns: usage.agentRuns.total,
    totalCost: usage.billing.total,
    costPerRun: usage.billing.total / usage.agentRuns.total,
    topAgents: usage.byAgent
      .sort((a: any, b: any) => b.cost - a.cost)
      .slice(0, 5),
    peakHours: usage.byHour
      .sort((a: any, b: any) => b.runs - a.runs)
      .slice(0, 5),
  };

  console.log('Usage Analysis:', analysis);
  return analysis;
}
```

### Step 2: Implement Usage Budgets
```typescript
interface Budget {
  monthly: number;
  daily: number;
  perAgent: number;
}

const budget: Budget = {
  monthly: 500, // $500/month
  daily: 20,    // $20/day
  perAgent: 50, // $50/agent/month
};

async function checkBudget(): Promise<boolean> {
  const lindy = new Lindy({ apiKey: process.env.LINDY_API_KEY });

  const usage = await lindy.usage.current();

  if (usage.billing.monthly >= budget.monthly) {
    console.error('Monthly budget exceeded!');
    await sendAlert('Budget exceeded', { usage, budget });
    return false;
  }

  if (usage.billing.today >= budget.daily) {
    console.warn('Daily budget exceeded');
    return false;
  }

  return true;
}
```

### Step 3: Optimize Agent Costs
```typescript
// Cost-optimized agent configuration
const optimizedAgent = {
  name: 'Cost-Efficient Agent',
  instructions: 'Be brief. Answer in 1-2 sentences.',
  config: {
    model: 'gpt-3.5-turbo', // Cheaper model
    maxTokens: 100,         // Limit output
    temperature: 0.3,       // Less creative = fewer tokens
  },
};

// Route simple queries to cheaper agents
async function routeQuery(input: string) {
  const isSimple = input.length < 100 && !input.includes('analyze');

  const agentId = isSimple
    ? 'agt_cheap_simple'
    : 'agt_expensive_complex';

  return lindy.agents.run(agentId, { input });
}
```

### Step 4: Implement Caching to Reduce Calls
```typescript
import NodeCache from 'node-cache';

const cache = new NodeCache({ stdTTL: 3600 }); // 1 hour cache

async function cachedRun(agentId: string, input: string) {
  const cacheKey = `${agentId}:${input}`;

  // Check cache first (free!)
  const cached = cache.get(cacheKey);
  if (cached) {
    console.log('Cache hit - $0');
    return cached;
  }

  // Only call API if cache miss
  const result = await lindy.agents.run(agentId, { input });
  cache.set(cacheKey, result);

  console.log('Cache miss - API call made');
  return result;
}
```

### Step 5: Set Up Cost Alerts
```typescript
async function setupCostAlerts() {
  const lindy = new Lindy({ apiKey: process.env.LINDY_API_KEY });

  // Alert at 80% of budget
  await lindy.billing.alerts.create({
    threshold: 400, // $400 of $500 budget
    type: 'monthly',
    channels: ['email', 'slack'],
    message: 'Approaching monthly budget limit',
  });

  // Daily anomaly detection
  await lindy.billing.alerts.create({
    threshold: 50, // 50% above average
    type: 'anomaly',
    channels: ['slack'],
    message: 'Unusual spending detected',
  });
}
```

## Cost Optimization Checklist
```markdown
[ ] Usage analysis completed
[ ] Budget limits defined
[ ] Cost alerts configured
[ ] Caching implemented
[ ] Cheaper models for simple tasks
[ ] Max tokens configured
[ ] Unused agents identified and disabled
[ ] Peak usage patterns analyzed
```

## Output
- Usage analysis report
- Budget enforcement
- Cost-optimized agents
- Caching for reduced API calls
- Alert system

## Error Handling
| Issue | Cause | Solution |
|-------|-------|----------|
| Budget exceeded | High usage | Throttle or pause |
| Cost spike | Anomaly | Investigate and alert |
| Cache ineffective | Low hit rate | Tune TTL |

## Examples

### Monthly Cost Report
```typescript
async function generateCostReport() {
  const usage = await analyzeUsage();

  const report = `
# Lindy Cost Report - ${new Date().toISOString().slice(0, 7)}

## Summary
- Total Runs: ${usage.totalRuns}
- Total Cost: $${usage.totalCost.toFixed(2)}
- Cost per Run: $${usage.costPerRun.toFixed(4)}

## Top Agents by Cost
${usage.topAgents.map((a: any) => `- ${a.name}: $${a.cost.toFixed(2)}`).join('\n')}

## Recommendations
${usage.costPerRun > 0.05 ? '- Consider cheaper models for simple tasks' : ''}
${usage.cacheHitRate < 0.3 ? '- Improve caching strategy' : ''}
  `;

  return report;
}
```

## Resources
- [Lindy Pricing](https://lindy.ai/pricing)
- [Usage Dashboard](https://app.lindy.ai/usage)
- [Cost Optimization Guide](https://docs.lindy.ai/cost)

## Next Steps
Proceed to `lindy-reference-architecture` for architecture patterns.