An open-source framework for AI apps, built and used in production by Google (genkit.dev)

Google has open-sourced Genkit, a full-stack framework for building production AI apps that the company already uses internally. Genkit offers a unified SDK and APIs to integrate hundreds of models from providers like Google (Gemini/Vertex AI), OpenAI, Anthropic, Ollama and more, letting teams pick best‑of‑breed models without rewriting app logic. It supports multimodal generation (text and images), type‑safe structured outputs, tool calling and agentic workflows, persisted chat, RAG-style retrieval, prompt templating and multi-step AI workflows — plus client helpers for web and mobile (Next.js, React, Angular, iOS, Android). The technical intent is pragmatic: consistent SDKs for JavaScript/TypeScript, Go and Python, a local CLI and Developer UI for rapid iteration (input testing, model comparisons, execution traces), and production monitoring for throughput, latency and error metrics. Genkit targets server-side deployment anywhere (Firebase Cloud Functions, Cloud Run, third‑party hosts) while providing client integration helpers. For AI/ML teams this reduces vendor lock‑in, accelerates experimentation across models, and closes the gap between prototypes and production by combining developer tooling, observability and multi‑provider model orchestration in a single, open framework.