genai-processors

Why we included this project

Gemini developers who keep writing the same async glue around every model call will get more mileage out of this library. Everything flows through one content model, and a Processor is just a small unit that turns an incoming stream into outgoing parts, so you can chain units together or run several in parallel to build bigger pipelines and agent behaviors. asyncio is the default rather than an afterthought, so concurrent calls and streaming output need no extra plumbing. The included processors cover ordinary turn-based model calls and the Live API for real-time audio-in/audio-out interaction, and you can write custom processors as plain functions or by subclassing the base classes. It is a lightweight alternative to the heavier orchestration frameworks, and teams already invested in Gemini should find the composition model handy for research agents and live apps.