Changelog

1.1.1

Features

Add BigQuery first-party tools. See here for more details.

1.1.0

Features

Extract agent loading logic from fast_api.py to a separate AgentLoader class and support more agent definition folder/file structure.
Added audio play in web UI.
Added input transcription support for live/streaming.
Added support for storing eval run history locally in adk eval cli.
Image artifacts can now be clicked directly in chat message to view.
Left side panel can now be resized.

Bug Fixes

Avoid duplicating log in stderr.
Align event filtering and ordering logic.
Add handling for None param.annotation.
Fixed several minor bugs regarding eval tab in web UI.

Miscellaneous Chores

Updates mypy config in pyproject.toml.
Add google search agent in samples.
Update filtered schema parameters for Gemini API.
Adds autoformat.sh for formatting codebase.

1.0.0

⚠ BREAKING CHANGES

Evaluation dataset schema is finalized with strong-type pydantic models. (previously saved eval file needs re-generation, for both adk eval cli and the eval tab in adk web UI).
BuiltInCodeExecutor (in code_executors package) replaces BuiltInCodeExecutionTool (previously in tools package).
All methods in services are now async, including session service, artifact service and memory service.
- list_events and close_session methods are removed from session service.
agent.py file structure with MCP tools are now easier and simpler (now vs before). Old format is not working anymore.
Memory schema and MemoryService is redesigned.
Mark various class attributes as private in the classes in the tools package.
Disabled session state injection if instruction provider is used. (so that you can have {var_name} in the instruction, which is required for code snippets)
Toolbox integration is revamped: tools/toolbox_tool.py → tools/toolbox_toolset.py.
Removes the experimental remote_agent.py. We'll redesign it and bring it back.

Features

Dev UI:
- A brand new trace view for overall agent invocation.
- A revamped evaluation tab and comparison view for checking eval results.
Introduced BaseToolset to allow dynamically add/remove tools for agents.
- Revamped MCPToolset with the new BaseToolset interface.
- Revamped GoogleApiTool, GoogleApiToolset and ApplicationIntegrationToolset with the new BaseToolset interface.
- Resigned agent.py file structure when needing MCPToolset.
- Added ToolboxToolset.
Redesigned strong-typed agent evaluation schema.
- Allows users to create more cohesive eval sets.
- Allows evals to be extended for non-text modality.
- Allows for a structured interaction with the uber eval system.
Redesigned Memory schema and MemoryService interfaces.
Added token usage to LlmResponse.
Allowed specifying --adk_version in adk deploy cloud_run cli. Default is the current version.

Bug Fixes

Fixed adk deploy cloud_run failing bug.
Fixed logs not being printed due to google-auth library.

Miscellaneous Chores

Display full help text when adk cli receives invalid arguments.
adk web now binds 127.0.0.1 by default, instead of 0.0.0.0.
InMemoryRunner now takes BaseAgent in constructor.
Various docstring improvements.
Various UI tweaks.
Various bug fixes.
Update various contributing/samples for contributors to validate the implementation.

0.5.0

⚠ BREAKING CHANGES

Updated artifact and memory service interface to be async. Agents that interact with these services through callbacks or tools will now need to adjust their invocation methods to be async (using await), or ensure calls are wrapped in an asynchronous executor like asyncio.run(). Any service that extends the base interface must also be updated.

Features

Introduced the ability to chain model callbacks.
Added support for async agent and model callbacks.
Added input transcription support for live/streaming.
Captured all agent code error and display on UI.
Set param required tag to False by default in openapi_tool.
Updated evaluation functions to be asynchronous.

Bug Fixes

Ensured a unique ID is generated for every event.
Fixed the issue when openapi_specparser has parameter.required as None.
Updated the 'type' value on the items/properties nested structures for Anthropic models to adhere to JSON schema.
Fix litellm error issues.

Miscellaneous Chores

Regenerated API docs.
Created a developer folder and added samples.
Updated the contributing guide.
Docstring improvements, typo fixings, GitHub action to enforce code styles on formatting and imports, etc.

0.4.0

⚠ BREAKING CHANGES

Set the max size of strings in database columns. MySQL mandates that all VARCHAR-type fields must specify their lengths.
Extract content encode/decode logic to a shared util, resolve issues with JSON serialization, and update key length for DB table to avoid key too long issue in mysql.
Enhance FunctionTool to verify if the model is providing all the mandatory arguments.

Features

Update ADK setup guide to improve onboarding experience.
feat: add ordering to recent events in database session service.
feat(llm_flows): support async before/after tool callbacks.
feat: Added --replay and --resume options to adk run cli. Check adk run --help for more details.
Created a new Integration Connector Tool (underlying of the ApplicationIntegrationToolSet) so that we do not force LLM to provide default value.

Bug Fixes

Don't send content with empty text to LLM.
Fix google search reading undefined for renderedContent.

Miscellaneous Chores

Docstring improvements, typo fixings, github action to enfore code styles on formatting and imports, etc.

0.3.0

⚠ BREAKING CHANGES

Auth: expose access_token and refresh_token at top level of auth credentials, instead of a dict (commit).

Features

Added support for running agents with MCPToolset easily on adk web.
Added custom_metadata field to LlmResponse, which can be used to tag LlmResponse via after_model_callback.
Added --session_db_url to adk deploy cloud_run option.
Many Dev UI improvements:
- Better google search result rendering.
- Show websocket close reason in Dev UI.
- Better error message showing for audio/video.

Bug Fixes

Fixed MCP tool json schema parsing issue.
Fixed issues in DatabaseSessionService that leads to crash.
Fixed functions.py.
Fixed skip_summarization behavior in AgentTool.

Miscellaneous Chores

README.md improvements.
Various code improvements.
Various typo fixes.
Bump min version of google-genai to 1.11.0.

0.2.0

⚠ BREAKING CHANGES

Fix typo in method name in Event: has_trailing_code_exeuction_result --> has_trailing_code_execution_result.

Features

adk CLI:
- Introduce adk create cli tool to help creating agents.
- Adds --verbosity option to adk deploy cloud_run to show detailed cloud run deploy logging.
Improve the initialization error message for DatabaseSessionService.
Lazy loading for Google 1P tools to minimize the initial latency.
Support emitting state-change-only events from planners.
Lots of Dev UI updates, including:
- Show planner thoughts and actions in the Dev UI.
- Support MCP tools in Dev UI. (NOTE: agent.py interface is temp solution and is subject to change)
- Auto-select the only app if only one app is available.
- Show grounding links generated by Google Search Tool.
.env file is reloaded on every agent run.

Bug Fixes

LiteLlm: arg parsing error and python 3.9 compatibility.
DatabaseSessionService: adds the missing fields; fixes event with empty content not being persisted.
Google API Discovery response parsing issue.
load_memory_tool rendering issue in Dev UI.
Markdown text overflows in Dev UI.

Miscellaneous Chores

Adds unit tests in Github action.
Improves test coverage.
Various typo fixes.

0.1.0

Features

Initial release of the Agent Development Kit (ADK).
Multi-agent, agent-as-workflow, and custom agent support
Tool authentication support
Rich tool support, e.g. built-in tools, google-cloud tools, third-party tools, and MCP tools
Rich callback support
Built-in code execution capability
Asynchronous runtime and execution
Session, and memory support
Built-in evaluation support
Development UI that makes local development easy
Deploy to Google Cloud Run, Agent Engine
(Experimental) Live(Bidi) audio/video agent support and Compositional Function Calling(CFC) support