Firebase Genkit with Ollama Support

Google announced the release of Firebase Genkit at Google I/O 2024, providing developers with an open-source framework to build, deploy, and monitor production-ready AI-powered applications. The framework includes native support for Ollama, enabling developers to run Google's open-source Gemma model on local machines for development and testing.

Local AI Development with Ollama and Gemma

Firebase Genkit integrates with Ollama to allow the execution of Google's Gemma model locally. This integration allows developers to iterate on AI features without relying exclusively on cloud-based APIs during the development phase. The setup is supported across MacOS, Windows, Linux, and via Docker containers.

Technical Implementation and Setup

To integrate Ollama with Firebase Genkit, developers must first install the Genkit CLI and pull the Gemma model via Ollama. The initialization process involves the following steps:

1. Installation: Install the Genkit CLI globally using npm i -g genkit.
2. Model Acquisition: Download the Gemma model using the command ollama pull gemma.
3. Project Initialization: Create a new Node.js project, run npm init, and execute genkit init to initialize the framework.
4. Provider Selection: During the genkit init process, developers can select Ollama as the model provider.

Once initialized, Genkit runs on localhost:4000, providing a developer interface for testing and monitoring AI workflows.

Ecosystem and Optimizations

Firebase Genkit is available as an open-source project on GitHub. For developers using NVIDIA hardware, NVIDIA has provided optimizations specifically for Firebase Genkit to run on RTX GPUs, further enhancing local performance for generative AI development.