Skip to main content
Back

Google Unveils ADK: Python Kit for Building Production-Ready Agents

TRY NOW

April 14, 2025 5 min read

Hey,

Welcome to AlphaSignal – the most read newsletter by AI developers. 


We bring you the top 1% of news, papers, models, and repos, all summarized to keep you updated on the latest in AI.

IN TODAY'S SIGNAL

Read time: 5 min 45 sec

🎖️ Top News

📌  CoreWeave

⚡️ Trending Signals

💻  Top Tutorials

🧠  Deep Dive

If you're enjoying AlphaSignal please forward this email to a colleague. 

It helps us keep this content free.

TOP NEWS

Agents

Google releases open-source Agent Development Kit to to build and deploy Gemini agents with tool use

⇧ 6,028 Likes

What's New

Google has released ADK, a production-grade Python framework for building agentic systems with full support for Gemini and third-party models. You can now build, test, and deploy tool-using agents that run on Google infrastructure. The framework supports recursive tool use, workflows, and containerized deployment out of the box.

Multi-Model Compatibility

ADK works with Gemini and external models through LiteLLM.

  • Supports Gemini 1.5 and third-party models like Claude, LLaMA, Mistral, and Jurassic-2

  • Lets you bring your own model via a custom connector

  • Uses Vertex AI Agent Builder for integration with Google’s model serving stack

  • Offers fallback and routing across models using prompt-based logic

Tooling and Extensibility

You can define and manage tools with Python.

  • Built-in tools include google_search and code_exec

  • Supports OpenAPI specs, Python functions, or custom wrappers

  • Agents can call tools and also invoke other agents as tools

  • Tool behavior is introspectable and testable with structured input/output


Workflow Orchestration

ADK includes a workflow engine that structures agent behavior.

  • Supports sequential, parallel, and conditional execution

  • Built with native support for loops and retries

  • Orchestration uses JSON/YAML and can invoke agents recursively

  • Offers detailed step-level logs for observability and debugging


Deployment and Runtime

ADK runs anywhere, with support for container runtimes and Vertex AI Agent Engine.

  • Deploy as containerized HTTP APIs with minimal setup

  • Deploy agents with Vertex AI for enterprise runtime integration

  • Supports web-based interaction with optional UI and CLI

  • Access APIs managed via Apigee for tool integration

  • Connect to over 100 enterprise systems like BigQuery, AlloyDB, NetApp

  • Includes memory management for long-running agents

  • Use Gemini 2.5 Pro Experimental models directly via native model support


Availability
The public repo us available under Apache 2.0.


Community Feedback


Natalie VRBeliever

"Workflow orchestration AND built-in evaluation? ADK checks all the boxes for slick AI agent development. Love the multi-agent focus!"

Shawn Chauhan

"This is a game changer for anyone looking to work with multi-agent systems—love the low-code approach!"

Verysell AI

"Google’s Agent Development Kit (ADK) is setting the stage for powerful, flexible, and scalable AI agents! The ability to integrate multiple agents in modular architectures, combined with a rich tool ecosystem and robust developer experience, opens up a lot of possibilities for creating dynamic, production-ready AI systems."

Try now

Unlock Faster Training with Smarter GPU Optimization

AI labs and enterprises need to optimize GPU clusters for faster time to market. CoreWeave, the AI Hyperscaler, is hosting a webinar to share how they achieve superior speed, efficiency, and reliability in their GPU infrastructure.

 Learn advanced optimization techniques for:

  • Faster model loading

  • Efficient checkpointing

  • Scalable fleet management


See how CoreWeave quickly identifies and remediates job failures, driving industry-leading effective training times for large-scale GPU clusters.

Reduce downtime and improve training efficiency.

Register now

partner with us →

TRENDING SIGNALS

Coding Assistance

GitHub integrates Gemini 2.5 Pro in Copilot Chat, enabling stronger reasoning and complex code generation

⇧ 795 Likes

Reasoning Model

ByteDance unveils a 200B MoE model focused on STEM reasoning, surpassing o3-mini on ARC-AGI, a benchmark for AGI-style tasks

⇧ 846 Likes

Generative AI

ElevenLabs updates voice cloning flow with speaker separation, clip trimming, noise removal, and session resume support

⇧ 1,348 Likes

AI Trends

Stanford publishes AI index 2025 showing small models now match closed ones with just 1.7% gap and 78% business adoption

⇧ 1,578 Likes

AI Funding

Ilya Sutskever’s SSI secures $2B funding at $32B valuation, continues pursuit of safe superintelligence as sole product focus

⇧ 1,485 Like

TOP TUTORIALS

RAG

Build Better RAG Systems with Microsoft’s Crash Course

⇧ 1,238 Likes

Learn how to build efficient RAG systems using indexing, vector search, hybrid methods, and multimodal retrieval. This 6-part course includes hands-on demos, GitHub samples, and optimization strategies for real-world performance. Understand tradeoffs, tune retrieval for accuracy, and apply techniques that cut latency and improve relevance in generative applications.

MCP

Create Custom MCP Servers Using Claude Code

⇧ 581 Likes

Define a custom MCP server using Claude Code by combining tools, memory, context rules, and prompts. This tutorial shows how to configure behavior using tool_choice, input_schema, and step logic. Build agentic workflows that maintain state, call external APIs, and follow strict control flows in a YAML file.

LLMs

How to Teach LLMs to Follow Instructions

⇧ 1,193 Stars

Learn how to fine-tune a decoder-only LLM to follow free-form instructions like answering questions or rewriting text. Reuse the pretraining loss and loop only the data format changes. The tutorial shows prompt/response formatting, dynamic padding, loss masking, and clean end-of-text handling using a lightweight dataset for fast, practical experimentation.

DEEP DIVE

LLM 

Pre-Training GPT-4.5

⇧ 6,538 Likes

This podcast dives into the challenges faced during the development of OpenAI's GPT-4.5 model. Sam Altman, Amin Tootoonchian, Alex Paino, and Daniel Selsam explores the technical aspects of scaling models, handling compute bottlenecks, and managing large-scale distributed systems.


You'll gain insights into how scaling from 10,000 to 100,000 GPUs impacted error rates and performance. You'll also learn about the novel Torch sum bug, data efficiency strategies, and the importance of ML-systems co-design.


Key topics covered:

  • Challenges in scaling GPT-4.5 training

  • Strategies for large GPU cluster management

  • Handling early run failures and bug fixes

  • Pre-training as data compression

  • Key metrics tracking and performance monitoring

  • Future directions in model scaling and hardware advancements

Watch now

At Alpha Signal, our mission is to build a sharp, engaged community focused on AI, machine learning, and cutting-edge language models, helping over 200,000 developers stay informed and ahead. We’re passionate about curating the best in AI, from top research and trending technical blogs to expert insights and tailored job opportunities. We keep you connected to the breakthroughs and discussions that matter, so you can stay in the loop without endless searching. We also work closely with partners who value the future of AI, including employers and advertisers who want to reach an audience as passionate about AI as we are.


Our partnerships are based on shared values of ethics, responsibility, and a commitment to building a better world through technology.Privacy is a priority at Alpha Signal. Our Privacy Policy clearly explains how we collect, store, and use your personal and non-personal information. By using our website, you accept these terms, which you can review on our website. This policy applies across all Alpha Signal pages, outlining your rights and how to contact us if you want to adjust the use of your information. We’re based in the United States. By using our site, you agree to be governed by U.S. laws.

Looking to promote your company, product, service, or event to 200,000+ AI developers? Let's work together.

Work with us