Genkit is a framework for building AI-powered applications. It provides open source libraries for Node.js and Go, along with tools to help you debug and iterate quickly.
Learn more in our documentation for Node.js and Go.
Genkit is a versatile framework, which you can use to build many different types of AI applications. Common use cases include:
-
Intelligent agents: Create agents that understand user requests and perform tasks autonomously, such as personalized travel planning or itinerary generation.
- Example: Compass Travel Planning App
-
Data transformation: Convert unstructured data, like natural language, into structured formats (e.g., objects, SQL queries, tables) for integration into your app or data pipeline.
-
Retrieval-augmented generation: Create apps that provide accurate and contextually relevant responses by grounding generation with your own data sources, such as chatbots or question answering systems.
Genkit is built for developers seeking to add generative AI to their apps with Node.js or Go, and can run anywhere these runtimes are supported. It's designed around a plugin architecture that can work with any generative model API or vector database, with many integrations already available.
While developed by the Firebase team, Genkit can be used independently of Firebase or Google Cloud services.
Note
Genkit for Go is in alpha, so we only recommend it for prototyping.
-
Unified generation API: Generate text, media, structured objects, and tool calls from any generative model using a single, adaptable API.
-
Vector database support: Add retrieval-augmented generation (RAG) to your apps with simple indexing and retrieval APIs that work across vector database providers.
-
Enhanced prompt engineering: Define rich prompt templates, model configurations, input/output schemas, and tools all within a single, runnable .prompt file.
-
AI workflows: Organize your AI app logic into Flows - functions designed for observability, streaming, integration with Genkit devtools, and easy deployment as API endpoints.
-
Built-in streaming: Stream content from your Genkit API endpoints to your client app to create snappy user experiences.
Genkit provides a CLI and a local UI to streamline your AI development workflow.
The Genkit CLI is the quickest way to start a new Genkit project. It also includes commands for running and evaluating your Genkit functions (flows).
- Install:
npm i -g genkit - Initialize a new project:
genkit init
The Genkit developer UI is a local interface for testing, debugging, and iterating on your AI application.
Key features:
- Run: Execute and experiment with Genkit flows, prompts, queries, and more in dedicated playgrounds.
- Inspect: Analyze detailed traces of past executions, including step-by-step breakdowns of complex flows.
- Evaluate: Review the results of evaluations run against your flows, including performance metrics and links to relevant traces.
Extend Genkit with plugins for specific AI models, vector databases, and platform integrations from providers like Google and OpenAI.
- Node.js plugins: Explore on npm
- Go plugins: Explore on pkg.go.dev
Create and share your own plugins:
- Write Node.js plugins: Plugin Authoring Guide
- Write Go plugins: (Documentation coming soon)
Find excellent examples of community-built plugins for OpenAI, Anthropic, Cohere, and more in this repository.
Want to try Genkit without a local setup? Explore it on Project IDX, Google's AI-assisted workspace for full-stack app development in the cloud.
-
Join the community: Stay updated, ask questions, and share your work with other Genkit users on our Discord server.
-
Provide feedback: Report issues or suggest new features using our GitHub issue tracker.
-
Engage in discussions: Participate in conversations about Genkit on our GitHub Discussions forum.
Contributions to Genkit are welcome and highly appreciated! See our Contribution Guide to get started.
Genkit is built by Firebase with contributions from the Open Source Community.