AI SaaS Gross Margin & Unit Economics Simulator (2026)

import { Redis } from '@upstash/redis';
import { Anthropic } from '@anthropic-ai/sdk';
import { GoogleGenerativeAI } from '@google/generative-ai';

const redis = Redis.fromEnv();
const anthropic = new Anthropic();
const gemini = new GoogleGenerativeAI(process.env.GEMINI_API_KEY!);

export async function routeUserQuery(userId: string, prompt: string, isComplex: boolean) {
  // 1. Check user monthly token consumption from Redis
  const currentTokens = await redis.get<number>(`tokens:${userId}`) || 0;
  const PLAN_TOKEN_LIMIT = 2_500_000; // 2.5M tokens on $49 Pro Plan

  // 2. Enforce model downgrade if user exceeded 80% of quota or prompt is simple
  if (currentTokens > PLAN_TOKEN_LIMIT * 0.8 || !isComplex) {
    // Route to Gemini 2.0 Flash: $0.10 / 1M tokens (98% gross margin)
    const model = gemini.getGenerativeModel({ model: 'gemini-2.0-flash' });
    const res = await model.generateContent(prompt);
    await redis.incrby(`tokens:${userId}`, 1500);
    return res.response.text();
  }

  // 3. Route to Claude 3.7 Sonnet with Prompt Caching for complex reasoning
  const response = await anthropic.messages.create({
    model: 'claude-3-7-sonnet-20250219',
    max_tokens: 1024,
    system: [
      {
        type: 'text',
        text: SYSTEM_PROMPT_STATIC_KNOWLEDGE,
        cache_control: { type: 'ephemeral' } // 90% discount on cache hits!
      }
    ],
    messages: [{ role: 'user', content: prompt }]
  });

  await redis.incrby(`tokens:${userId}`, response.usage.input_tokens + response.usage.output_tokens);
  return response.content[0].text;
}