Kathy Wu

This commit **implements a caching mechanism** for **performance improvement** within the **`google.adk.tools.base_toolset` module**. Specifically, the `BaseToolset.get_tools_with_prefix()` method now stores and reuses its results for calls made within the same invocation. This **refactoring** significantly **reduces redundant network calls**, thereby **decreasing latency** and improving the overall efficiency of tool retrieval. New unit tests in `tests/unittests/tools/test_base_toolset.py` have been added to thoroughly verify this caching behavior.

This commit **enhances the skill name validation logic** within the **Google ADK skills module** to support `snake_case` skill names in addition to the standard `kebab-case`. This change addresses a limitation where Python's import system does not allow `kebab-case` for module names, thus enabling **importing within skill scripts**. A new **feature flag**, `SNAKE_CASE_SKILL_NAME`, was introduced in `_feature_registry.py` to control this allowance, and the `validate_name` function in `src/google/adk/skills/models.py` was updated to respect it. This provides **skill developers** with more flexibility for naming conventions, crucial for Python-based skill development.

This commit introduces a **new Slack integration** for the **ADK (Agent Development Kit)**, enabling agents to operate directly within Slack. It implements the core `SlackRunner` class within the `google.adk.integrations.slack` module, which is responsible for handling Slack message events and managing the Slack application lifecycle. A new sample agent in `contributing/samples/slack_agent/agent.py` demonstrates its usage, complemented by comprehensive documentation and unit tests for the `SlackRunner`. This **new capability** significantly expands the deployment options for ADK agents, allowing them to interact with users in a widely used communication platform.

This commit implements a **fix** within the **Agent Registry** integration, specifically updating the `get_remote_a2a_agent` function in `src/google/adk/integrations/agent_registry/agent_registry.py`. The change modifies the agent retrieval logic to prioritize and utilize the **full agent card** when available, ensuring that more detailed agent information is consistently used. This **bug fix** improves the accuracy and completeness of agent data provided by the registry, enhancing the quality of information for remote agents. A new test case has been added to `tests/unittets/integrations/agent_registry/test_agent_registry.py` to validate this enhanced processing. The update ensures that downstream components relying on agent information receive the most comprehensive data possible.

This commit performs a **refactoring** of import statements within the **`google.adk.tools.environment_simulation`** module. It **reorders and relocates imports** for `FeatureName` and `experimental` in both `environment_simulation_engine.py` and `tool_spec_mock_strategy.py`. This is a **maintenance** task focused on improving code hygiene and consistency, ensuring that necessary modules are imported correctly and in a standardized manner. The change has no functional impact on the environment simulation's behavior, primarily affecting internal code organization.

This commit introduces a **new feature** to support execution timeouts within the **ADK code execution framework**. It adds a `timeout_seconds` field to the base class `BaseCodeExecutor` and implements the core timeout logic in `UnsafeLocalCodeExecutor` using Python's `multiprocessing` module, specifically within the `_execute_in_process` and `execute_code` methods. This enhancement allows for preventing indefinitely running or hung code executions, significantly improving the **stability and resource management** of local code execution. A new unit test in `test_unsafe_local_code_executor.py` verifies the correct functioning of this timeout mechanism.

This commit **refactors** the project's directory structure by **moving the `LangchainTool` and `LangchainToolConfig` classes** from `src/google/adk/tools` to a new, dedicated `src/google/adk/integrations/langchain` package. This **code reorganization** improves the logical grouping of **integrations** within the `google.adk` library. The old `langchain_tool.py` file is now deprecated and re-exports the moved symbols, ensuring backward compatibility while guiding users to update their import paths. This change enhances the **module structure** and **maintainability** of the `langchain` integration.

This commit performs a significant **refactoring** by **relocating** the `CrewaiTool` implementation and its configuration. The `CrewaiTool` and `CrewaiToolConfig` classes are moved from `src/google/adk/tools` to the new `src/google/adk/integrations/crewai` package, improving **module organization** for **integrations**. A new `__init__.py` file now exposes these components within the `integrations/crewai` package. The original `crewai_tool.py` now re-exports the tool from its new location, providing a deprecation warning to guide users to the updated import path. This **maintenance** task ensures better structural clarity and consistency within the project, with corresponding unit tests also updated.

This commit introduces a **new feature** to the **`google.adk.tools`** module, enabling more flexible **tool management** by allowing `SkillToolset` to include other `BaseToolset` instances in its `additional_tools` field. This enhancement permits the **resolution of tools from nested toolsets**, significantly improving the modularity and organization of tool configurations. The `__init__` and `get_tools` methods of `SkillToolset` were updated to support this new capability. This change allows for more complex and reusable tool structures, with expanded unit tests confirming the correct dynamic tool resolution logic for nested scenarios.

This commit introduces a **new capability** to the `google.adk.skills` module by adding the `_list_skills_in_dir` utility function. This function enables the **programmatic discovery and listing of skills** present within a specified local directory, enhancing skill management capabilities. The new utility is exported via `src/google/adk/skills/__init__.py` as `list_skills_in_dir` for easier access. Comprehensive **unit tests** have been added in `tests/unittests/skills/test__utils.py` to ensure its correctness across various valid and invalid skill scenarios. This feature provides a foundational tool for dynamically loading and interacting with skills from the filesystem.

This commit introduces a **new capability** by adding **support for ADK tools** within the **`SkillToolset`**, allowing skills to declare and utilize external tools. Specifically, the **`Frontmatter` model** in `src/google/adk/skills/models.py` is updated to validate a new `adk_additional_tools` metadata field, and the **`SkillToolset`** in `src/google/adk/tools/skill_toolset.py` gains logic to dynamically resolve these tools based on skill metadata. This enhancement provides greater flexibility for skill definitions, enabling users to integrate custom ADK tools, as demonstrated by the updated `skills_agent` sample which now includes a `GetTimezoneTool`.

This commit performs a significant **refactoring** of the **Spanner integration module**, relocating all related tools and components from `src/google/adk/tools/spanner` to the more semantically appropriate `src/google/adk/integrations/spanner` directory. This **maintenance** effort improves the logical organization of the codebase by consolidating integration-specific functionalities under a dedicated `integrations` namespace. To ensure backward compatibility and a smooth transition, the original `src/google/adk/tools/spanner/__init__.py` file now issues a **deprecation warning** and re-exports modules from their new location. All associated sample code, documentation, and integration tests have been updated to reflect these new module paths, ensuring the continued functionality of the **Spanner tools** while guiding users towards the new, standardized structure.

This commit **refactors** the project structure by **moving the `ApiRegistry` class** and its associated logic from `src/google/adk/tools/api_registry.py` to a new, dedicated **`src/google/adk/integrations/api_registry`** module. This reorganization aims to better categorize integration-related components within the `google.adk` package. The original `api_registry.py` file now issues a **deprecation warning** to guide users to the new import path, while sample agents and unit tests have been updated to reflect this change. This ensures a smoother transition for consumers of the `ApiRegistry` functionality and improves overall code maintainability.

This commit **introduces a new `get_humidity.py` script** to the **sample skills agent's weather skill**, providing a concrete example of how to use the `RunSkillScript` tool. This **new capability** simulates fetching live humidity data for a specified location, enhancing the agent's demonstration features. The `agent.py` configuration is updated to enable script execution via `UnsafeLocalCodeExecutor` within the `SkillToolset`. Additionally, the `SKILL.md` documentation for the weather skill is updated to include instructions for utilizing this new script, improving the clarity and completeness of the sample.

This commit **adds a new README file** to the `src/google/adk/integrations` directory, providing essential **documentation** for this specific subsystem. The README clarifies the **purpose and scope of the integrations module**, detailing what types of code belong there and outlining contribution guidelines. This **maintenance** improvement enhances developer understanding and consistency, streamlining future contributions and ensuring proper usage of the `integrations` folder within the `google/adk` project.

This commit provides a **compatibility fix** and **refactoring** for the **`google.adk.skills` package**, specifically addressing the `DEFAULT_SKILL_SYSTEM_INSTRUCTION` constant. It **re-exports** this constant from both `src/google/adk/skills/__init__.py` and `src/google/adk/skills/prompt.py` to prevent breaking existing user code. A `__getattr__` function is introduced in these modules to issue a **deprecation warning** when the constant is accessed via these paths, guiding users towards its intended new location. This ensures backward compatibility while gracefully signaling future API changes.

This commit delivers a **bug fix** for the **sample skills agent**, correcting an incorrect reference to the **weather skill**. Specifically, it updates the loading path in `contributing/samples/skills_agent/agent.py` from `weather_skill` to `weather-skill`. This ensures the sample agent can properly locate and integrate the weather skill, making the example functional and preventing loading failures.

This commit **updates build and export configurations** to properly integrate the newly introduced **`agent_registry`** module within the **ADK (Agent Development Kit)**. This **maintenance update** enables the `AgentRegistry` client library, which provides functionality to interact with the Google Cloud Agent Registry service and create ADK components. Specifically, it adjusts **Copybara** and **BUILD file configurations** to recognize and export the `src/google/adk/integrations/agent_registry` folder. This ensures the new **Agent Registry feature** is correctly built and available for use, supporting the creation of ADK components.

This commit introduces a **new feature** by updating the **Agent Registry** to directly instantiate `AgentCard` objects from information embedded within the `get agents` endpoint response, anticipating an upcoming API modification. This adaptation ensures the `Agent Registry` can correctly process the new `AgentCard` data structure, which includes **new protocol enums**, as the `get agents` endpoint transitions from returning a URL to full `AgentCard` data. Consequently, the **unit tests** for the `Agent Registry` are updated to reflect these changes, ensuring robust validation of the new data handling and direct embedding of agent card details.

This commit introduces a **new feature** to the **Agent Development Kit (ADK)** by providing a **client library for the Agent Registry API**. This new capability allows users to programmatically discover, look up, and connect to agents and MCP servers cataloged in the registry. It includes a **new sample agent** (`contributing/samples/agent_registry_agent/agent.py`) demonstrating its usage, along with comprehensive **unit tests** (`tests/unittests/integrations/agent_registry/test_agent_registry.py`) to ensure the client's reliability. This integration significantly enhances the ADK's ability to manage and interact with a network of agents.