Released on October 25, 2025, adk-docker-base 1.17.0 provides a streamlined update to your AI agent development environment. This release centers on a significant dependency bump of the core google-adk library to version 1.17.0, unlocking a suite of powerful new capabilities for agent control, tool integration, and developer productivity.

What’s new

This update brings several high-impact features from the Google Agent Development Kit directly into the base image:

Advanced Agent Control & Observability

  • Session Rewinding: You can now rewind an agent’s session to a previous state, allowing for more flexible debugging and interaction flow.
  • Parallel Agent Resumption: Improved support for managing parallel agents, specifically when multiple branches are paused awaiting tool confirmation.
  • Granular UI Control: New per-agent speech configurations in the UI and better observability controls to manage LLM request/response spans.

Expanded Tooling & Integration

  • Vertex AI Code Execution: Integration with the Vertex AI Code Execution Sandbox API allows your agents to execute generated code in a secure, managed environment.
  • Enhanced MCP Support: The Model Context Protocol (MCP) integration now supports dynamic per-request headers and optional confirmation requirements for tools.
  • Search Optimizations: New options to bypass multi-tool limits for both Google Search and Vertex AI Search tools.

Developer Experience & Stability

  • Evaluation Tooling: The ADK CLI now includes native support for creating and managing evaluation sets.
  • Critical Fixes: This release resolves several stability issues, including LangChain 1.0.0 compatibility and MySQL data truncation bugs.
  • Performance Gains: Optimized context cache creation to reduce overhead during requests.

Why it matters

For developers building complex AI agents, these updates translate to higher reliability and greater flexibility. The addition of the Vertex AI Code Execution Sandbox removes a significant hurdle for agents that need to perform real-world computation securely. Meanwhile, session rewinding and parallel agent management provide a more sophisticated way to handle non-linear conversations and complex multi-step tasks.

The inclusion of native evaluation set creation in the CLI also streamlines the “build-test-refine” loop, making it easier to quantify agent performance improvements over time.

Getting Started

Upgrading to 1.17.0 is straightforward. Simply update the base image reference in your Dockerfile:

# Update from:
FROM ghcr.io/unitvectory-labs/adk-docker-base:1.16.0

# To:
FROM ghcr.io/unitvectory-labs/adk-docker-base:1.17.0

We recommend reviewing the google-adk v1.17.0 release notes for details on API migrations, such as the shift from invocation_context to callback_context.

This post was AI-generated using the model unsloth/gemma-4-31B-it-GGUF:UD-Q5_K_XL. It is based on the release of adk-docker-base 1.17.0 (2025-10-25), generated on 2026-04-09. Created by release-storyteller.