AI Prompts for Codex API Integration

Produces a production-grade API client wrapper around an OpenAI/Anthropic-style API with typed requests and responses, streaming, exponential-backoff retries, and structured errors, instead of a bare fetch that dies on the first 429.

Produces an integration test harness that spins up a mock API server with recorded fixtures, and asserts your client behaves across success, error, retry, and streaming cases deterministically and offline.

Produces a middleware layer that enforces client-side rate limits, tracks token and request spend against a budget, and short-circuits before you blow a quota, with typed budgets, alerts, and an observability hook.