perplexity-rate-limits

# Perplexity Rate Limits

## Overview
Handle Perplexity Sonar API rate limits. Perplexity uses a leaky bucket algorithm: burst capacity is available, with tokens refilling continuously at your assigned rate. Rate limits are based on requests per minute (RPM).

## Rate Limit Tiers

| Tier | RPM | Notes |
|------|-----|-------|
| Free / Starter | 50 | Default for new API keys |
| Search API | ~3 req/sec | Per-endpoint limit |
| Higher tiers | Contact sales | Custom limits available |

Rate limits apply per API key, not per model. Using `sonar-pro` counts against the same RPM as `sonar`.

## Prerequisites
- `PERPLEXITY_API_KEY` set
- Understanding of HTTP 429 responses

## Instructions

### Step 1: Exponential Backoff with Jitter
```typescript
async function withExponentialBackoff<T>(
  operation: () => Promise<T>,
  config = { maxRetries: 5, baseDelayMs: 1000, maxDelayMs: 30000, jitterMs: 500 }
): Promise<T> {
  for (let attempt = 0; attempt <= config.maxRetries; attempt++) {
    try {
      return await operation();
    } catch (error: any) {
      if (attempt === config.maxRetries) throw error;

      const status = error.status || error.response?.status;
      // Only retry on 429 (rate limit) and 5xx (server errors)
      if (status && status !== 429 && status < 500) throw error;

      const exponentialDelay = config.baseDelayMs * Math.pow(2, attempt);
      const jitter = Math.random() * config.jitterMs;
      const delay = Math.min(exponentialDelay + jitter, config.maxDelayMs);

      console.warn(`[Perplexity] ${status || "error"} — retry ${attempt + 1}/${config.maxRetries} in ${delay.toFixed(0)}ms`);
      await new Promise((r) => setTimeout(r, delay));
    }
  }
  throw new Error("Unreachable");
}

// Usage
const result = await withExponentialBackoff(() =>
  perplexity.chat.completions.create({
    model: "sonar",
    messages: [{ role: "user", content: "test query" }],
  })
);
```

### Step 2: Queue-Based Rate Limiting
```typescript
import PQueue from "p-queue";

// 50 RPM = ~0.83 req/sec. Set intervalCap=1, interval=1200ms for safety.
const perplexityQueue = new PQueue({
  concurrency: 3,
  interval: 1200,
  intervalCap: 1,
});

async function queuedSearch(query: string, model = "sonar") {
  return perplexityQueue.add(() =>
    withExponentialBackoff(() =>
      perplexity.chat.completions.create({
        model,
        messages: [{ role: "user", content: query }],
      })
    )
  );
}

// Batch queries are automatically rate-limited
const queries = ["query 1", "query 2", "query 3", "query 4", "query 5"];
const results = await Promise.all(queries.map((q) => queuedSearch(q)));
```

### Step 3: Token Bucket Implementation (No Dependencies)
```typescript
class TokenBucket {
  private tokens: number;
  private lastRefill: number;

  constructor(
    private maxTokens: number = 50,
    private refillRate: number = 50 / 60  // 50 per minute = 0.83/sec
  ) {
    this.tokens = maxTokens;
    this.lastRefill = Date.now();
  }

  async acquire(): Promise<void> {
    this.refill();
    if (this.tokens >= 1) {
      this.tokens -= 1;
      return;
    }
    // Wait until a token is available
    const waitMs = (1 / this.refillRate) * 1000;
    await new Promise((r) => setTimeout(r, waitMs));
    this.refill();
    this.tokens -= 1;
  }

  private refill() {
    const now = Date.now();
    const elapsed = (now - this.lastRefill) / 1000;
    this.tokens = Math.min(this.maxTokens, this.tokens + elapsed * this.refillRate);
    this.lastRefill = now;
  }

  get available(): number {
    this.refill();
    return Math.floor(this.tokens);
  }
}

const bucket = new TokenBucket(50, 50 / 60);

async function rateLimitedSearch(query: string) {
  await bucket.acquire();
  return perplexity.chat.completions.create({
    model: "sonar",
    messages: [{ role: "user", content: query }],
  });
}
```

### Step 4: Python Rate Limiting
```python
import time, asyncio
from collections import deque

class RateLimiter:
    def __init__(self, rpm: int = 50):
        self.rpm = rpm
        self.window = deque()

    def wait_if_needed(self):
        now = time.time()
        # Remove timestamps older than 60 seconds
        while self.window and self.window[0] < now - 60:
            self.window.popleft()
        if len(self.window) >= self.rpm:
            sleep_time = 60 - (now - self.window[0])
            time.sleep(max(0, sleep_time))
        self.window.append(time.time())

limiter = RateLimiter(rpm=50)

def rate_limited_search(client, query: str, model: str = "sonar"):
    limiter.wait_if_needed()
    return client.chat.completions.create(
        model=model,
        messages=[{"role": "user", "content": query}],
    )
```

## Error Handling
| Signal | Meaning | Action |
|--------|---------|--------|
| HTTP 429 | RPM exceeded | Backoff and retry |
| `Retry-After` header | Seconds until reset | Honor this value exactly |
| Repeated 429s | Sustained overload | Reduce concurrency or add queue |
| 429 on burst | Bucket empty | Space requests 1.2s apart |

## Output
- Automatic retry with exponential backoff and jitter
- Queue-based rate limiting for batch operations
- Token bucket for fine-grained control
- Python rate limiter for synchronous code

## Resources
- [Perplexity Rate Limits](https://docs.perplexity.ai/guides/rate-limits)
- [p-queue Documentation](https://github.com/sindresorhus/p-queue)

## Next Steps
For security configuration, see `perplexity-security-basics`.