I’ve been thinking a lot about the best way to deploy GenAI backends with @Genkit
The idea I kept coming back to was simple: what if deploying AI tools, prompts, and flows could feel as native to Kubernetes as deploying any other backend?
So I started building Genkit Operator 🚀
With it, you can describe your Genkit app using Kubernetes resources like Prompt, Model, PluginConfig, and Flow, and the operator takes care of exposing them as HTTP endpoints.
Apply YAML. Get a production-ready Genkit endpoint.
It supports multiple providers like OpenAI, Anthropic, Google AI, Vertex AI, AWS Bedrock, and Ollama, and it is designed to work nicely with GitOps tools like Argo CD and Flux.
Still early, but I’m excited about the direction: making GenAI backends more declarative, portable, and Kubernetes-native.
Check it out here: https://t.co/c67xdmoAIw
Vercel AI SDK wraps the model. Genkit wraps the pipeline. This dev puts both middleware systems side by side — hooks, built-ins, composition, streaming, and a decision matrix to help you pick the right one.
{ author: @Xavidop + @GoogleDevExpert }
https://t.co/dbRTZJS81N
Would love to hear your thoughts particularly on the developer tooling side. We're going to be cooking up a lot in the coming weeks/months.
https://t.co/dLSLCWfrnA
🚀 I just released genkitx-temporal!
A new Genkit plugin to run Genkit flows as Temporal Workflows, bringing durable execution, retries, timeouts, cancellation, scaling, and Temporal UI visibility to your GenAI apps.
https://t.co/3TZP5yp9pj
#Genkit#Temporal#GenAI#OpenSource #TypeScript
@genkitframework@temporalio@googledevexpert
Go developers, Genkit middleware is now available! 🚀
Enhance your AI workflows with powerful capabilities like intercepting the generation loop, adding custom logic, and handling retries seamlessly.
Check out the documentation to get started ⬇️
🚀 @GenkitFramework Java now has Middleware V2!
Bring fine-grained control to your AI generation pipeline in Java, intercept, transform, and observe every step of generate() and tool calls.
✨ What you can do with Middleware V2:
🔹 Wrap entire generate() calls (logging, caching, retries, guardrails)
🔹 Wrap individual tool executions with full access to ToolRequestPart / ToolResponsePart metadata
🔹 Compose multiple middlewares cleanly
🔹 Short-circuit, mutate requests/responses, or add tracing — all type-safe
Perfect for production-grade GenAI apps where observability, safety, and control matter. ☕️🤖
📖 Docs: https://t.co/9sZ7QTL49w
#GenkitJava #GenAI #Java #LLM #AI #OpenSource #GoogleAI
Updated the Middleware section of "Mastering Genkit: Go Edition" to fully align with the Middleware V2 API! ⚡️
Working with @Xavidop to keep the content fresh. This update details the three powerful hook points in the new API:
🔹 WrapGenerate: Hooks into the tool-call loop to monitor or alter accumulated requests.
🔹 WrapModel: Wraps individual model API calls, perfect for retries, fallbacks, and cost tracking.
🔹 WrapTool: Wraps each tool execution, useful for auditing or human-in-the-loop approvals.
Check out the updated chapter here:
https://t.co/eUyH0vAKcQ
@GenkitFramework@golang
@fender_kn and I have updated Mastering Genkit, Go Edition to use Genkit 1.7.0 and Go 1.26.
We also refreshed the Middleware chapter, which now uses Middleware V2, making the examples more aligned with the latest Genkit APIs and patterns.
If you are building GenAI applications with Go, this book is a great place to start and go deeper into the Genkit ecosystem.
📖 Check it out here:
https://t.co/cuSFi62bVN
@genkitFramework@golang@googledevexpert