Key features
GenAI models
Unified API for generation across AI models built by Google (Gemini, Gemma) and third party providers. Supports multimodal input, multimedia content generation, and custom options.
Generate structured output as strongly-typed objects with custom schemas for easy integration into your app.
Define custom tools for your AI models to fetch data, display UI, write to a database, and more.
Write well structured prompts with all relevant metadata encapsulated in a single dotprompt file. Supports handlebars templating, history, multimedia, and more.
Retrieval
Build context-aware AI features by indexing your data and dynamically retrieving relevant information from your database. Genkit provides flexible, light-weight abstractions for indexers and retrievers that work with any database provider.
Evaluation
Evaluate your end-to-end AI workflow using a variety of pre-built and custom evaluators. As easy as:
genkit eval:flow myAiWorkflow --input testQuestions.json
Extensibility with plugins
Access pre-built components and integrations for models, vector stores, tools, evaluators, observability, and more through Genkit’s open ecosystem of plugins built by Google and the community. For a list of existing plugins from Google and the community, explore the #genkit-plugin keyword on npm.
You can also use this extensibility to easily define custom components whenever existing plugins don’t fit your needs.
For more information:
Deployment
Deploy your AI feature with a single command through the Firebase or Google Cloud CLI to:
- Cloud Functions for Firebase (Node.js only)
- Firebase App Hosting as a Next.js app (Early preview, Node.js only)
- Google Cloud Run (Node.js or Go)
You can also deploy to any container platform where your chosen runtime is supported.
Observability and monitoring
Genkit is fully instrumented with OpenTelemetry and provides hooks to export telemetry data. Easily log traces and telemetry to Google Cloud using pre-built plugins or set up with a custom provider for full end-to-end observability and monitoring in production.