perplexity-architecture-variants

# Perplexity Architecture Variants ## Overview Three validated architectures for Perplexity Sonar API at different scales. Each builds on the previous, adding caching and orchestration as volume grows. ## Decision Matrix | Factor | Direct Widget | Cached Layer | Research Pipeline | |--------|--------------|--------------|-------------------| | Volume | <500/day | 500-5K/day | 5K+/day | | Latency (p50) | 2-5s | 50ms (cached) / 2-5s (miss) | 10-30s | | Model | `sonar` | `sonar` + cache | `sonar` + `sonar-pro` | | Monthly Cost | <$150 | $50-$300 | $300+ | | Complexity | Minimal | Moderate | High | ## Instructions ### Variant 1: Direct Search Widget (<500 queries/day) Best for: Adding AI search to an existing app. No cache needed at this scale. ```typescript // Simple endpoint — add to any Express/Next.js app import OpenAI from "openai"; const perplexity = new OpenAI({ apiKey: process.env.PERPLEXITY_API_KEY!, baseURL: "https://api.perplexity.ai", }); app.post("/api/search", async (req, res) => { try { const response = await perplexity.chat.completions.create({ model: "sonar", messages: [{ role: "user", content: req.body.query }], max_tokens: 1024, }); res.json({ answer: response.choices[0].message.content, citations: (response as any).citations || [], }); } catch (err: any) { if (err.status === 429) { res.status(429).json({ error: "Rate limited. Try again shortly." }); } else { res.status(500).jso...