Building and deploying AI-powered apps with GitHub Spark (docs.github.com)

GitHub announced Spark (public preview), a low-code-to-code platform that generates, iterates, and deploys full‑stack AI apps from natural-language prompts, mockups or markdown. Users (requires a GitHub account with Copilot Pro+) describe an app and Spark scaffolds a TypeScript/React project, auto-generates prompts powered by GitHub Models, provisions key‑value storage, and wires in GitHub authentication. You can refine behavior via natural-language iteration, visual editors, or direct code edits (live preview + Copilot completions), open a synced Codespace for advanced development, and deploy with one click to a managed runtime that includes cloud storage and LLM inference. For AI/ML practitioners this lowers the barrier for prototyping and productizing generative features: Spark handles prompt generation, model selection, API integration and runtime hosting, while exposing editable prompts, styles (Tailwind/CSS), assets, and a two‑way Git repo sync for collaboration. Important caveats: it uses an opinionated React/TypeScript stack (external libs may be incompatible), data persistence is auto‑provisioned unless explicitly disabled, and a Copilot setting meant to block suggestions matching public code may not work as expected with Spark. Spark speeds iteration and deployment of ML-enabled apps but introduces governance, dependency and data‑sharing considerations you should audit before publishing.