Unpacking ADK v1.2.0 & v1.2.1: What’s New in Google’s Agent Development Kit
Introduction
Google’s Agent Development Kit (ADK) continues to evolve, empowering developers to build sophisticated AI agents with greater ease and flexibility. The recent v1.2.0 and v1.2.1 releases introduce several noteworthy enhancements aimed at improving usability, security, and developer experience.
Key Updates in ADK v1.2.0
Agent Engine Deployment
The ADK CLI now supports deploying agents directly to the agent engine, providing developers with more options for deploying their AI agents.
GCS Artifact Service Integration
An option to use Google Cloud Storage for artifacts in adk web has been added, allowing for scalable and reliable storage solutions.
Parallel Tool Call Handling
Implemented index tracking to manage parallel tool calls using LiteLLM, enhancing the toolkit’s ability to handle concurrent operations.
Enhanced List Operations
Added sortByColumn functionality to list operations, enabling more efficient data sorting and management.
Eval Set Management
Implemented get_eval_case, update_eval_case, and delete_eval_case functionalities for the local eval sets manager, improving evaluation workflows.
Vertex AI Search Tool Configurations
Exposed additional configurations from the latest Google GenAI SDK, providing developers with more control over the Vertex AI Search Tool.
Agent Visualization
Introduced new agent visualization features, aiding in the understanding and debugging of agent workflows.
Langchain Tool Support
Added support for Langchain’s StructuredTool and tools with run_manager in _run arguments without args_schema, broadening the range of compatible tools.
Anthropic Model Updates
Updated support for Anthropic models, ensuring compatibility with the latest advancements.
BigQuery Credentials
Set BigQuery scope by default in BigQuery credentials, simplifying authentication processes.
Jira Agent Sample
Added sample code for a Jira agent connecting to Jira Cloud, providing a practical example for integrating with external services.
Enhancements in ADK v1.2.1
Comprehensive CLI Help Text
The ADK CLI now displays full help text upon receiving invalid arguments, assisting developers in troubleshooting command usage errors.
Secure Web Binding
The adk web command now binds to 127.0.0.1 by default, enhancing security during local development by restricting access to the local machine.
In-Memory Runner Flexibility
InMemoryRunner now accepts BaseAgent in its constructor, offering greater flexibility in agent testing scenarios.
Documentation Improvements
Numerous docstring enhancements have been made throughout the codebase, improving clarity and understanding for developers.
UI and Bug Fixes
Various user interface tweaks and bug fixes have been implemented, contributing to a more stable and user-friendly toolkit.
Updated Samples
Refreshed contributing samples to assist contributors in validating their implementations effectively, fostering a more robust development environment.
Final note
ADK v1.2.0 and v1.2.1 reflect Google’s commitment to continuous improvement, focusing on developer-centric enhancements that streamline the process of building, testing, and deploying AI agents. These updates not only improve the toolkit’s usability but also lay the groundwork for more secure and efficient agent development workflows.
For a comprehensive overview of the changes, refer to the release notes:
