wtf is meta-model-cookbook?

Meta Model API Cookbook is a collection of example code showing how to build with Meta's Model API, which serves an AI model called Muse Spark. The main selling point is compatibility: since the API is built to work like the OpenAI and Anthropic APIs, existing code written for those services can often be pointed at Meta's servers with only small changes, like swapping the base web address and API key. Muse Spark itself can read very large amounts of text at once, and a preview version is currently free to try. The repository is organized as a set of copy paste starting points, split into three sections. The first section covers basic building blocks: making a first API call, streaming responses as they are generated, calling external tools, getting reliably structured JSON output, reusing prompts to save cost, controlling how much the model reasons before answering, sending images, working with very long text inputs, handling errors safely, and grounding answers in live web search results. The second section shows patterns for building AI agents, programs that can plan, take actions, and check their own work, including a basic agent loop, managing a growing conversation over many turns, and making careful in place file edits. The third section is a set of full example applications: reading data out of chart images, fixing bugs from a screenshot of an error, controlling smart glasses hands free, generating slide decks, building websites and browser games with a coding agent that checks its own results in a real browser, running model generated code safely in a sandbox, coordinating a team of several specialized AI agents on one shared task board, and controlling a computer's desktop directly from screenshots. Each recipe is provided as a self contained Python file or notebook meant to be copied and adapted, and the license terms are included separately in the repository.