what is production-architecture-explorer fr?

Production Architecture Explorer is an interactive web page that maps out the seven layers a real AI system needs once it moves past a simple demo and into production use. The idea behind it is that most problems with AI systems in the real world, like runaway costs, missing rollback plans, or agents that take unsafe actions, come from gaps in these surrounding layers rather than from the AI model itself being wrong. The seven layers you can explore are the gateway and router, which handles rate limits, choosing which model to use, spending caps, and switching providers if one fails, prompt management, for versioning and testing prompts the way you would version code, the retrieval pipeline, which finds and feeds relevant information into a request, agent orchestration, for multi step tasks where the AI uses tools and follows a chain of actions, guardrails and safety, which filter what goes in and out and enforce rules, observability, for tracing and debugging what happened in production, and evaluation and feedback, for testing quality and improving over time. Clicking on any layer expands it to show what it actually does, what kinds of failures it guards against, and two versions of how you might build it: a simple starting point versus a fuller enterprise setup. The tool also links each layer to a 25 day AI engineering curriculum, so you can see which lessons connect to which part of the system, and there is an animation that shows a single request traveling through all seven layers in order, with notes at each step. It is built with Next.js, React, TypeScript, and Tailwind CSS for styling, with Framer Motion powering the animations. To run it yourself, you install the dependencies and start it with a single development server command, then open it in your browser. The project has no listed license information in its README, and it is a young, unstarred project intended mainly as a learning and reference tool rather than a production service itself.