google-gemini/genai-processors ? reverse-engineered prompt

Build me a lightweight Python library for working with Gemini style AI pipelines that feels easy to use and easy to extend.

I want a simple way to pass in text, images, audio, or mixed content, process it with small reusable components, and get results back either all at once or as a stream. The main idea should be a Processor class that can be chained together or run in parallel, with a clean async API based on asyncio. Please include a friendly content model, support for streaming parts, and a few built in processors for common patterns like normal model calls and realtime streaming. It should be simple enough that someone can write their own processor by subclassing or using a decorator.

Add a few small examples that show how to echo content, build a basic research style pipeline, and do a realtime audio or text stream. Make sure the library is well organized, documented, and works cleanly with Python 3.10 or newer. If you need to check current Gemini docs, look them up online.