Loading feed…
SitePoint0 Hot0 comments17 min read
Rubberduck: Emulating OpenAI Locally for Reliable AI Development
Rubberduck is a local HTTP server that emulates the OpenAI REST API, returning deterministic, configurable responses. It solves common pain points for teams building AI-powered features: unpredictable API costs, rate limits, flaky tests due to non-deterministic LLM responses, and network dependencies in CI/CD pipelines. The guide covers installation as an npm dev dependency, configuring response fixtures (including latency simulation and error scenarios like 429s), swapping between local and live APIs via environment variables, writing reliable Vitest tests against emulated responses, and integrating Rubberduck into CI pipelines using start-server-and-test. Advanced topics include streaming SSE emulation, multi-model fixture scoping, and embedding endpoint support for RAG pipelines. The post also outlines limitations: Rubberduck cannot do prompt engineering or evaluate output quality, and fixtures must be manually updated when OpenAI changes its API schema.
Source: https://www.sitepoint.com/rubberduck-local-ai-emulation. 8sync News only summarizes and links out; content copyright belongs to the authors and original sources.
