ai-request-logger

Lightweight utility for structured logging of LLM API requests.

Why

LLM API calls are often hard to observe in production.

ai-request-logger provides structured logs for every AI request so you can monitor:

• latency
• token usage
• prompt size
• estimated cost

This helps teams detect prompt growth, debug latency issues, and track AI spending.

License

MIT

Install

npm install ai-request-logger

or

pnpm add ai-request-logger

or

bun add ai-request-logger

Usage

import { withAIRequestLogger } from "ai-request-logger";

const response = await withAIRequestLogger({
  provider: "openai",
  model: "gpt-4o-mini",
  prompt,
  logger: (log) => console.log(log),
  call: async () => client.responses.create({
    model: "gpt-4o-mini",
    input: prompt,
  }),
});

Works with any LLM provider

This library is provider-agnostic and works with:

• OpenAI
• Anthropic
• Gemini
• AWS Bedrock
• Azure OpenAI
• Cohere
• Custom APIs

Log fields

• provider
• model
• latencyMs
• promptLength
• contextLength
• requestId
• promptTokens
• completionTokens
• totalTokens
• estimatedCost
• inputCost
• outputCost

Example log

{
  "provider": "openai",
  "model": "gpt-4o-mini",
  "latencyMs": 421,
  "promptLength": 120,
  "promptTokens": 30,
  "completionTokens": 50,
  "totalTokens": 80,
  "estimatedCost": 0.00003
}

API

withAIRequestLogger({
  provider: string,
  model: string,
  prompt?: string,
  context?: string[],
  requestId?: string,
  pricingOverrides?: PricingTable,
  onStart?: () => void,
  logger?: (log: AIRequestLog) => void,
  onComplete?: (log: AIRequestLog, response: ResponseType) => void,
  call: () => Promise<ResponseType>
}): Promise<ResponseType>

Notes

• If no logger is provided, logs default to console.log.
• Cost estimation is best-effort and only returns values for known models.

Related packages

Part of a small AI developer toolkit:

• token-budget-guard – enforce token budgets for LLM calls
• llm-retry-guard – safe retry wrapper for LLM APIs
• ai-request-logger – structured logging for AI requests