google-gemini/cookbook — reverse-engineered prompt

Build me a Gemini API cookbook repo that feels like a hands on learning hub, mostly notebook based, with a small set of JavaScript quickstarts too. I want simple get started guides for beginners, plus practical examples people can run, copy, and tweak for real use cases.

Please organize it into quickstarts and examples. Cover the main Gemini workflows in a friendly way, including basic prompting, thinking mode, managed agents, image generation, music and TTS, video generation, live multimodal experiences, webhooks for async jobs, file search for using your own data, grounding with outside sources like Google Maps, and examples that show different inference tier tradeoffs. Keep each notebook focused, runnable, and easy to open in Colab.

Add clear setup instructions for API keys, short explanations of what each guide teaches, and small demo outputs so someone can learn by experimenting instead of reading a lot of theory. If anything has changed recently, look up the current Gemini docs online and use the latest recommended approach.