adk-python

v1.34.0 Feature

This release adds 4 notable features for engineering teams evaluating rollout.

Published 16d AI Agents & Assistants

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

agent agentic agentic-ai agents agents-sdk ai

+10 more

ai-agents aiagentframework genai genai-chatbot llm llms multi-agent multi-agent-systems multi-agents multi-agents-collaboration

Affected surfaces

crypto_tls auth

Summary

AI summary

Broad release touches Bug Fixes, 1.34.0, Performance Improvements, and Code Refactoring.

Changes in this release

Type	Severity	Summary	CVE
Feature
Feature	Low	Add persistent task stores support for A2A. Add persistent task stores support for A2A. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Feature	Low	General Gemini Live API support added to ADK evaluate. General Gemini Live API support added to ADK evaluate. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Feature	Low	MTLS support added to Google Cloud Telemetry exporter. MTLS support added to Google Cloud Telemetry exporter. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Feature	Low	A2aAgentExecutor factory support added to to_a2a() function. A2aAgentExecutor factory support added to to_a2a() function. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Feature	Low	Support for non-ADK produced input-required events added. Support for non-ADK produced input-required events added. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Feature	Low	Config option to include tool calls/responses in conversation history added. Config option to include tool calls/responses in conversation history added. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Feature	Low	Gemini auto review and invoke workflows added under CI. Gemini auto review and invoke workflows added under CI. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Feature	Low	GCPSkillRegistry implemented in ADK. GCPSkillRegistry implemented in ADK. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Feature	Low	Skill Registry implemented in ADK. Skill Registry implemented in ADK. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Feature	Low	Agent Skill description validation made more informative. Agent Skill description validation made more informative. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Feature	Low	Data retrieval handling simplified for ask_data_agent and ask_data_insights tools. Data retrieval handling simplified for ask_data_agent and ask_data_insights tools. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Feature	Low	OAuth PKCE support added to McpToolset. OAuth PKCE support added to McpToolset. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix
Bugfix	High	Persist refreshed OAuth2 credentials to store for auth. Persist refreshed OAuth2 credentials to store for auth. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Medium	Fix visibility of output_key state delta in callbacks for agents. Fix visibility of output_key state delta in callbacks for agents. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Medium	Map negative thinking_budget to adaptive thinking for Anthropic. Map negative thinking_budget to adaptive thinking for Anthropic. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Medium	Remove unneeded OAuth flows for auth. Remove unneeded OAuth flows for auth. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Medium	Handle none config in per_turn_user_simulator_quality for evaluation. Handle none config in per_turn_user_simulator_quality for evaluation. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Avoid pre-serializing dict values in Interactions API to prevent double-escaping. Avoid pre-serializing dict values in Interactions API to prevent double-escaping. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Enforce CacheMetadata active-state invariant for cache. Enforce CacheMetadata active-state invariant for cache. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Handle fingerprint-only metadata in performance analyzer for cache. Handle fingerprint-only metadata in performance analyzer for cache. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Catch OSError when importing AnthropicLlm. Catch OSError when importing AnthropicLlm. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Fallback to project id if credentials don't contain quota project. Fallback to project id if credentials don't contain quota project. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Fix missing dynamically loaded tools in SkillToolset during same invocation. Fix missing dynamically loaded tools in SkillToolset during same invocation. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Ensure sub live agent doesn't inherit session resumption handle from parent live agent for live. Ensure sub live agent doesn't inherit session resumption handle from parent live agent for live. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Preserve string content in Anthropic tool_result blocks for models. Preserve string content in Anthropic tool_result blocks for models. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Preserve tool_use IDs for Anthropic models on session resume for models. Preserve tool_use IDs for Anthropic models on session resume for models. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Treat empty GenerateContentResponse without prompt feedback as successful for models. Treat empty GenerateContentResponse without prompt feedback as successful for models. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Only serialize llm_response to json if it will be included in the trace. Only serialize llm_response to json if it will be included in the trace. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Preserve live_session_id in function call handling. Preserve live_session_id in function call handling. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—
Bugfix	Low	Prevent compaction of events involved in Human-in-the-Loop interactions. Prevent compaction of events involved in Human-in-the-Loop interactions. Source: granite4.1:8b-q6_K@2026-05-19 Confidence: high	—

Full changelog

1.34.0 (2026-05-18)

Features

a2a: add support for persistent task stores (cd78d87)
add general support for Gemini Live API in ADK evaluate (790c9be)
Add mTLS support to Google Cloud Telemetry exporter (cfe8d2c)
add support for A2aAgentExecutor factory in to_a2a() function (115124c)
add support for non-ADK produced input-required events (6e53472)
Added config option to include tool calls/responses in conversation history passed to user simulator (baf7efb)
ci: add Gemini auto review and invoke workflows (fd8b492)
Implement GCPSkillRegistry in ADK (88ebd42)
Implement Skill Registry in ADK (380d261)
Make Agent Skill description validation more informative (9f38973)
Simplify data retrieved handling of ask_data_agent tool and ask_data_insights tool (48f1b30)
Support OAuth PKCE in McpToolset (e7316dc)

Bug Fixes

agents: fix visibility of output_key state delta in callbacks (0524797)
anthropic: map negative thinking_budget to adaptive thinking (03b915b)
auth: persist refreshed OAuth2 credentials to store (218ea76), closes #5329
auth: remove unneeded OAuth flows (c35a579)
avoid pre-serializing dict values in Interactions API to prevent double-escaping (85f397d)
cache: enforce CacheMetadata active-state invariant (76b9f0b)
cache: handle fingerprint-only metadata in performance analyzer (9c5de58)
Catch OSError when importing AnthropicLlm (91cb5c6)
evaluation: handle none config in per_turn_user_simulator_quality (eed9bd3), closes #5677
fallback to project id if crendetials don't contain quota project (e377cb5)
Fix missing dynamically loaded tools in SkillToolset during the same invocation (f9097cb)
live: ensure sub live agent doesn't inherit session resumption handle from parent live agent to avoid interrupting the conversation (8dd9147)
models: preserve string content in Anthropic tool_result blocks (9a1e75f), closes #5358
models: preserve tool_use IDs for Anthropic models on session resume (327c45f), closes #5074
models: treat empty GenerateContentResponse without prompt feedback as successful (0cb9ae9)
only serialize llm_response to json if it will be included in the trace (1284493)
Preserve live_session_id in function call handling (07a9a01)
Prevent compaction of events involved in Human-in-the-Loop interactions (bb2efb6)
raise eagerly on importing AgentRegistry if a2a-sdk is missing (33cf6cb)
small: Convert events to the A2A format while respecting user vs agent role (59f7347)
tools: preserve code_execution_result and executable_code in AgentTool (7e61b51), closes #5481
tools: Prevent AnyIO CancelScope task boundary violations during MCP session creation failure (4309159)
Update model name in hello_world agent (192f19d)
Update model to gemini-3-flash-preview in hello word agent sample (6d89d21)
Update model to gemini-3-flash-preview in session state agent sample (2d423e8)
use tool_responses role for gemma4 models in LiteLLM integration (3d07960), closes #5650

Performance Improvements

lazy-load service registries and split apps.app to cut cold start ~8% (bd062ec)
models: guard debug log evaluation with isEnabledFor (57d8fc7)
utils: cache find_context_parameter introspection (ec54bd4)

Code Refactoring

Make the "a2a_metadata" string a constant that can be depended on by extension developers (0821f2d)

View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track adk-python

Get notified when new releases ship.

About adk-python

An open-source, code-first Python toolkit for building, evaluating, and deploying sophisticated AI agents with flexibility and control.

All releases →