Genkit

Debugging "thinking" LLMs just got easier. 🧠🔍 Most reasoning models return thought tokens that are invaluable for debugging but often too verbose for your production application logs. Michael Doyle shared a great workflow for handling this in Genkit: 1. Capture a trace from your app. 2. Load it into the Genkit Dev UI. 3. Re-run the specific call in the Model Playground with thinking tokens enabled. This enables you to inspect the model's inner monologue and refine your prompts without changing a single line of application code. See the full tip here: https://lnkd.in/g_saAYK9

Headed to Google Cloud Next? ☁️🚀 The Genkit team is excited to be part of the action this year! We’ll be hanging out at the Flutter team’s booth, ready to talk all things Generative AI, agentic workflows, and developer experience. Whether you're already building with Genkit or just getting started with AI, stop by to see some live demos and chat with the team. See you there!

Want to build full-stack, agentic apps with Dart? 🎯 In the latest Flutter Community AI Circle, our very own Christopher Gill shows you exactly how to do it using Genkit. He covers everything from initial setup to building robust, autonomous agents that can think and act within your Dart applications. If you're looking to graduate your Flutter and Dart apps from simple prompts to agentic workflows, check out the following session: https://lnkd.in/gA3Uj59i

How do you move from 'it works on my machine' to production-ready AI infrastructure in Flutter? 🚀 Our friends at Very Good Ventures just published an excellent guide on refactoring a Flutter-based shopping assistant to use Genkit. By swapping manual AI integration for Genkit's modular architecture, they unlocked: ✅ Flows: Structured, composable AI operations. ✅ Middleware: Built-in retries for flaky model providers. ✅ Custom Tools: Seamlessly querying product catalogs within the generation loop. ✅ Structured Output: Type-safe JSON responses—no more regex hacks. The best part? They kept the UI completely untouched while significantly improving backend reliability and maintainability. If you're building GenUI features and need a more robust backend, this is a must-read. Check it out: https://lnkd.in/gT_iT-TB https://lnkd.in/gXyVXBDM

Ready to build your first AI agent with Genkit Dart? 🎯 Jamiu Okanlawon has open-sourced a fantastic new workshop that walks you through building a 'Music Store Agent' from the ground up. This is a perfect end-to-end example for Dart and Flutter developers looking to move beyond simple prompts into autonomous agentic workflows. ✨ Implementing tools for your agent to interact with data. 🔄 Creating a robust agentic loop. 📦 Handling structured output for reliable app integration. 🧪 Best practices for testing AI logic. The repository includes a detailed Codelab guide and a starter branch so you can build along with the workshop. Great work, Jamiu! Check out the repo and start building: https://lnkd.in/gVmJuB27

Genkit JS 1.32.0 is here! 🚀 We’ve added support for the latest models to help you build more powerful AI applications: ✨ Support for Gemma 4 in Gemini API ✨ Support for Lyria 3 in Vertex AI ✨ Preview support for Veo 3.1 Lite The update also includes a fix for the Anthropic plugin (no more duplicated maxOutputTokens keys!) and improved telemetry with interrupt metadata in span attributes. Check out the full release notes and upgrade your projects today: https://lnkd.in/g4i3ahEw

Did you know Genkit can run on Azure Functions? ☁️ Xavier Portilla Edo has published a step-by-step guide on deploying AI-powered Genkit flows to Azure Functions with the Azure OpenAI plugin. Highlights from the tutorial: ✨ Define and register Genkit flows as Azure Functions 🔐 Implement protected flows with API key authentication 🌊 Streaming support for real-time AI responses 🛠️ Local development with Azure Functions Core Tools and the Genkit Dev UI This is a great example of Genkit's platform-agnostic design — your AI flows, your cloud. Read the full guide: https://lnkd.in/gknfM2y3

Running Genkit on AWS Lambda with Bedrock! 🚀 Xavier Portilla Edo has published a complete guide to deploying AI-powered Genkit flows on AWS Lambda, powered by Amazon Bedrock. What's covered: ✨ Initialize Genkit with the Bedrock plugin for AWS model access 🌊 Build streaming flows that deliver real-time AI responses via Lambda 🛠️ Local development with Serverless Offline 📦 Deploy and invoke flows using the Genkit Client SDK This tutorial is a great showcase of Genkit's "use any cloud" philosophy — making it easy to adopt Genkit wherever your infrastructure already lives. Read the full guide: https://lnkd.in/gRYGjNQf

Running Genkit on AWS Lambda with Bedrock! 🚀 Xavier Portilla Edo has published a complete guide to deploying AI-powered Genkit flows on AWS Lambda, powered by Amazon Bedrock. What's covered: ✨ Initialize Genkit with the Bedrock plugin for AWS model access 🌊 Build streaming flows that deliver real-time AI responses via Lambda 🛠️ Local development with Serverless Offline 📦 Deploy and invoke flows using the Genkit Client SDK This tutorial is a great showcase of Genkit's "use any cloud" philosophy — making it easy to adopt Genkit wherever your infrastructure already lives. Read the full guide: https://lnkd.in/gRYGjNQf

Hybrid AI is here for Dart developers 🚀 Sasha Denisov just published genkit_flutter_gemma — a new Genkit Dart plugin that wraps flutter_gemma and brings on-device inference into the Genkit ecosystem. Run Gemma, Llama, DeepSeek, Qwen, Phi, FastVLM, and more — locally across Android, iOS, macOS, Windows, Linux, and Web — using the same Genkit plugin API. Pair it with Jhin Lee genkit_llamadart plugin (GGUF models via llama.cpp) and you have a clear path to hybrid AI apps: ☁️ Cloud providers (Gemini, OpenAI) for heavy reasoning 📱 On-device inference for offline, low-latency, or privacy-sensitive tasks 🔄 Switch between them without changing your flows This is what the Genkit plugin system was designed for — and the community is running with it. https://lnkd.in/g5WtzdKV