Per-request cost estimation is computed and available for monitoring

Raw token counts are an abstraction that most product decisions do not operate on. Knowing a request consumed 1,200 tokens tells you nothing actionable; knowing it cost $0.014 does. Dollar cost estimation unlocks concrete decisions: which feature is too expensive to offer on the free plan, which user is consuming 10x the cohort average, whether a model swap will save $300/month, and at what traffic level a cost-per-user budget is breached. NIST AI RMF MEASURE 2.5 requires quantitative monitoring of AI operational costs; dollar estimates are that quantification.

Add a cost calculator utility in src/lib/ai/cost-calculator.ts that converts token counts to dollar amounts using model pricing constants, then call it immediately after logging token usage.

// src/lib/ai/cost-calculator.ts
const PRICE_PER_1M_TOKENS = {
  "gpt-4o":         { input: 2.50,  output: 10.00 },
  "gpt-4o-mini":    { input: 0.15,  output: 0.60  },
  "claude-3-5-sonnet-20241022": { input: 3.00, output: 15.00 },
} as const;

export function estimateCost(
  model: keyof typeof PRICE_PER_1M_TOKENS,
  promptTokens: number,
  completionTokens: number
): number {
  const p = PRICE_PER_1M_TOKENS[model];
  return (promptTokens * p.input + completionTokens * p.output) / 1_000_000;
}

Store cost_usd alongside token counts in ai_request_logs. Verify by querying the table and confirming the values are plausible for the model used.

• ID: ai-token-optimization.caching-cost.cost-estimation-per-request

• Severity: low

• What to look for: Look for logic that converts token counts into dollar amounts using model pricing rates. This could be a utility function (calculateCost, estimateCost), a derived column in the request logs table, or an admin dashboard metric. Check if pricing constants are defined anywhere in the codebase (e.g., GPT4O_INPUT_PRICE_PER_1M = 2.50). Count all instances found and enumerate each.

• Pass criteria: The system computes or stores a dollar cost estimate alongside token counts, making it possible to answer "how much did this session cost?" or "what is our cost per feature?". At least 1 implementation must be confirmed.

• Fail criteria: Only raw token counts are tracked (or nothing is tracked), with no financial cost derivation. Dollar costs are not available without manual calculation using external pricing tables.

• Skip (N/A) when: The project is a hobby project or personal tool where cost management is not a concern. Also skip if the project uses a flat-rate API plan where per-request cost is meaningless. Signal: No production deployment signals (no vercel.json, netlify.toml), single developer, no user accounts.

• Detail on fail: "No cost estimation per request — token counts logged but financial impact invisible"

• Remediation: Token counts are abstract. Dollar amounts drive product decisions: pricing plans, feature cost analysis, user quotas, and abuse detection.

  // src/lib/ai/cost-calculator.ts
  const PRICE_PER_1M_TOKENS = {
    "gpt-4o":         { input: 2.50,  output: 10.00 },
    "gpt-4o-mini":    { input: 0.15,  output: 0.60  },
    "claude-3-5-sonnet-20241022": { input: 3.00, output: 15.00 },
  } as const;
  
  export function estimateCost(
    model: keyof typeof PRICE_PER_1M_TOKENS,
    promptTokens: number,
    completionTokens: number
  ): number {
    const prices = PRICE_PER_1M_TOKENS[model];
    return (promptTokens * prices.input + completionTokens * prices.output) / 1_000_000;
  }
  

  Then include the estimate in logs:

  const costUsd = estimateCost("gpt-4o", result.usage.promptTokens, result.usage.completionTokens);
  await db.insert("ai_request_logs").values({ ..., cost_usd: costUsd });
  

  Verify by querying the logs table for cost_usd values and confirming they are plausible for the model used.