Open
Conversation
…viderRegistry within CredentialManager The flow for integrating a new auth method will be as follows. The ADK framework contributor will 1. extend the `AuthScheme` to create their own within `adk/auth/auth_scheme.py` 2. implement `BaseAuthProvider` within their dedicated directory in `adk/integrations/auth` 3. do the static registration of the new scheme and provider with AuthProviderRegistry of CredentialManager. PiperOrigin-RevId: 881775983
PiperOrigin-RevId: 881782296
The _update_type_string function now recursively processes "properties" at any level of the schema, ensuring that all "type" fields within nested objects are correctly lowercased. This improves handling of complex Co-authored-by: George Weale <gweale@google.com> PiperOrigin-RevId: 882050939
PiperOrigin-RevId: 882138997
Co-authored-by: Yifan Wang <wanyif@google.com> PiperOrigin-RevId: 882192675
This change adds logic to extract and re-embed the `thought_signature` field associated with function calls in Gemini models when converting between LiteLLM's ChatCompletionMessageToolCall and ADK's types.Part Close #4650 Co-authored-by: George Weale <gweale@google.com> PiperOrigin-RevId: 882212223
Co-authored-by: Kathy Wu <wukathy@google.com> PiperOrigin-RevId: 882212923
The `can_use_output_schema_with_tools` function now checks if a model is a LiteLlm instance by inspecting its type's Method Resolution Order, rather than directly importing `LiteLlm` Co-authored-by: George Weale <gweale@google.com> PiperOrigin-RevId: 882253446
PiperOrigin-RevId: 882270600
V4 still uses deprecated Node.js 20. Co-authored-by: Liang Wu <wuliang@google.com> PiperOrigin-RevId: 882275971
Co-authored-by: Liang Wu <wuliang@google.com> PiperOrigin-RevId: 882293566
Co-authored-by: Liang Wu <18244712+wuliang229@users.noreply.github.com>
Co-authored-by: Guoyi Lou <guoyilou@google.com> PiperOrigin-RevId: 882787811
This change introduces a new interceptor that adds the 'https://google.github.io/adk-docs/a2a/a2a-extension/' extension to the request headers in the A2A client from the RemoteAgent side. To send this extension along with requests, the RemoteAgent has to be instantiated with the `use_legacy` flag set to False. The AgentExecutor will default to the new implementation when this extension is requested by the client, but this behavior can be disabled via the `use_legacy` flag. The 'force_new' flag on the agent_executor side can be used to bypass the presence of the extension, and always activate the new version of the agent_executor. PiperOrigin-RevId: 883021792
Co-authored-by: Xuan Yang <xygoogle@google.com> PiperOrigin-RevId: 883246168
…creation due to missing version field Co-authored-by: Achuth Narayan Rajagopal <achuthr@google.com> PiperOrigin-RevId: 883336463
Co-authored-by: Xiang (Sean) Zhou <seanzhougoogle@google.com> PiperOrigin-RevId: 883401159
Merge #4718 ### Link to Issue or Description of Change **1. Link to an existing issue (if applicable):** - Closes: N/A - Related: N/A **2. Or, if no issue exists, describe the change:** **Problem** ADK’s MCP integration currently does not expose the MCP sampling callback capability. This prevents agent-side LLM sampling handlers from being used when interacting with MCP servers that support sampling. The MCP Python SDK supports sampling callbacks, but these parameters are not propagated through the ADK MCP integration layers. **Solution** Add sampling callback support by propagating the parameters through the MCP stack: - Add `sampling_callback` and `sampling_capabilities` parameters to `McpToolset` - Forward them to `MCPSessionManager` - Forward them to `SessionContext` - Pass them into `ClientSession` initialization This enables agent-side sampling handling when interacting with MCP servers. --- ### Testing Plan **Unit Tests** - [x] I have added or updated unit tests for my change. - [x] All unit tests pass locally. Added `test_mcp_sampling_callback.py` to verify that the sampling callback is correctly invoked. Example result: pytest tests/unittests/tools/mcp_tool/test_mcp_sampling_callback.py 1 passed **Manual End-to-End (E2E) Tests** Manual testing was performed using a FastMCP sampling example server where the sampling callback was invoked from the agent side and returned the expected response. --- ### Checklist - [x] I have read the CONTRIBUTING.md document. - [x] I have performed a self-review of my own code. - [x] I have commented my code where necessary. - [x] I have added tests proving the feature works. - [x] Unit tests pass locally. - [x] I have manually tested the change end-to-end. --- ### Additional context This change aligns ADK MCP support with the sampling capabilities available in the MCP Python SDK and enables agent implementations to handle sampling requests via a callback. Co-authored-by: Kathy Wu <wukathy@google.com> COPYBARA_INTEGRATE_REVIEW=#4718 from Piyushmrya:fix-mcp-sampling-callback 18f477f PiperOrigin-RevId: 883401178
Co-authored-by: Kathy Wu <wukathy@google.com> PiperOrigin-RevId: 883401885
Co-authored-by: Kathy Wu <wukathy@google.com> PiperOrigin-RevId: 883403479
Closes issue #4805 Co-authored-by: Liang Wu <wuliang@google.com> PiperOrigin-RevId: 883403628
…-2-preview The gemini-embedding-2-preview model requires the Vertex AI :embedContent endpoint instead of the legacy :predict endpoint used by older models (text-embedding-004, text-embedding-005). In google-genai <1.64.0, embed_content() unconditionally routed to :predict on Vertex AI, which returns FAILED_PRECONDITION for this model. v1.64.0 (googleapis/python-genai@af40cc6) introduced model-aware dispatch in embed_content(): models with "gemini" in the name are routed to :embedContent via t_is_vertex_embed_content_model(), while older text-embedding-* models continue to use :predict. This version also enforces a single-content-per-call limit for the embedContent API, which is why FilesRetrieval sets embed_batch_size=1. Co-authored-by: Xiang (Sean) Zhou <seanzhougoogle@google.com> PiperOrigin-RevId: 883689438
…provision an Agent Engine if neither agent_engine_resource_name nor sandbox_resource_name is provided The AgentEngineSandboxCodeExecutor now has three initialization modes: 1. Create both an Agent Engine and sandbox if neither resource name is provided. 2. Creating a new sandbox within a provided agent_engine_resource_name. 3. Using a provided sandbox_resource_name. PiperOrigin-RevId: 884088248
… execute_sql to support fine grained access controls PiperOrigin-RevId: 884166439
The metric takes into account all the turns of the multi-turn conversation. The class delegates the responsibility to Vertex Gen AI Eval SDK. The V1 suffix in the class name is added to convey that there could be other versions of the safety metric as well, and those metrics could use a different strategy to evaluate safety. Co-authored-by: Ankur Sharma <ankusharma@google.com> PiperOrigin-RevId: 884504910
Tool use: The class delegates the responsibility to Vertex Gen AI Eval SDK. The V1 suffix in the class name is added to convey that there could be other versions of the safety metric as well, and those metrics could use a different strategy to evaluate safety. Task trajectory: this metric is different from `Multi-Turn Overall Task Success`, in the sense that task success only concerns itself with the goal of whether the success was achieved or not. How that was achieved is not its concern. This metric on the other hand does care about the path that agent took to achieve the goal. Co-authored-by: Ankur Sharma <ankusharma@google.com> PiperOrigin-RevId: 884525532
Co-authored-by: Kathy Wu <wukathy@google.com> PiperOrigin-RevId: 884557773
When using OAuth2Session with `client_secret_post`, Authlib automatically includes the client_id and client_secret in the request body. Explicitly passing `client_id` again results in a duplicate parameter in the token exchange request, which can cause issues with some OAuth providers. Close #4782 Co-authored-by: George Weale <gweale@google.com> PiperOrigin-RevId: 884574091
PiperOrigin-RevId: 889912199
PiperOrigin-RevId: 889935432
- Migrate `tools/bigquery` to `integrations/bigquery` to match standard ADK layout. - Update unit tests to fully assert integration migration. Co-authored-by: Haiyuan Cao <haiyuan@google.com> PiperOrigin-RevId: 889963735
PiperOrigin-RevId: 889992570
… default Co-authored-by: Sasha Sobran <asobran@google.com> PiperOrigin-RevId: 890025323
…definitions The list_agents method in AgentLoader no longer attempts to determine the agent language for each directory Co-authored-by: George Weale <gweale@google.com> PiperOrigin-RevId: 890078066
…r session management Co-authored-by: Xuan Yang <xygoogle@google.com> PiperOrigin-RevId: 890127075
The workflow now accepts a `release_branch` input, checks out the specified release branch, and updates the version in `src/google/adk/version.py` Co-authored-by: George Weale <gweale@google.com> PiperOrigin-RevId: 890134954
…ponses Close: #4609 Co-authored-by: Xuan Yang <xygoogle@google.com> PiperOrigin-RevId: 890144644
When an MCP server (or any toolset) is unavailable or raises an exception during get_tools(), the error previously propagated uncaught through _convert_tool_union_to_tools() and canonical_tools(), crashing the entire agent silently with no log output. This change wraps the toolset.get_tools_with_prefix() call in a try-except that catches exceptions, logs a warning with the toolset class name and error details, and returns an empty tool list. This allows the agent to continue operating with tools from other working toolsets. The fix preserves the existing retry behavior — McpToolset's @retry_on_errors decorator still attempts reconnection before the error reaches this handler. On subsequent agent invocations, get_tools() is called again, naturally retrying the connection. Fixes: #3341 Co-authored-by: Xiang (Sean) Zhou <seanzhougoogle@google.com> PiperOrigin-RevId: 890512146
The deep_merge_dicts function will now extend lists when merging dictionaries, instead of overwriting them. Co-authored-by: Vishwa Murugan <vishwamurugan@google.com> PiperOrigin-RevId: 890531331
… ExpressMode and GCP projects Otherwise we get errors of the form: `Project/location and API key are mutually exclusive in the client initializer` (https://github.com/googleapis/python-genai/blob/1ccad7b70ae1068cb5f3e866ad7ca4d42aa55e1e/google/genai/_api_client.py#L590-L591). Co-authored-by: Yeesian Ng <ysian@google.com> PiperOrigin-RevId: 890588927
PiperOrigin-RevId: 890595289
This PR updates the `toolbox-adk` package version in the `pyproject.toml` and also updates the MCP Toolbox server version to the latest in the docs. PiperOrigin-RevId: 890633573
…ecution This change introduces a new EnvironmentToolset within the ADK, providing tools to interact with a BaseEnvironment. The toolset includes Execute, ReadFile, EditFile, WriteFile. System instructions are injected to guide the LLM on workspace usage and file manipulation. Co-authored-by: Liang Wu <wuliang@google.com> PiperOrigin-RevId: 890672569
…le I/O locally This change introduces the LocalEnvironment class, which allows agents to execute shell commands and perform file operations using asyncio subprocesses on the local machine. It includes support for a working directory and custom environment variables. A sample agent configuration is also added to demonstrate its usage. Co-authored-by: Liang Wu <wuliang@google.com> PiperOrigin-RevId: 890685517
Co-authored-by: Shangjie Chen <deanchen@google.com> PiperOrigin-RevId: 890694124
Merge #4842 ### Link to Issue or Description of Change **1. Link to an existing issue** Closes #4841 **2. Changes** This change aligns SSE transport behavior with Streamable HTTP transport for MCP tools, enabling users to customize the httpx.AsyncClient for SSE-based MCP connections (e.g., proxy, auth, TLS config). - Update `SseConnectionParams` model to include httpx_client_factory parameter (matching StreamableHTTPConnectionParams pattern) - Pass `httpx_client_factory` from `SseConnectionParams` to underlying `sse_client()` call - Add `model_config` to `SseConnectionParams` to allow arbitrary types for the factory function - Update docstrings to document the new `httpx_client_factory attribute` - Add unit test for new senarios ### Testing Plan **Unit Tests:** - [x] I have added or updated unit tests for my change. - `tests/unittests/tools/mcp_tool/test_mcp_session_manager.py` - [x] All unit tests pass locally. - All test cases have passed the test. **Manual End-to-End (E2E) Tests:** 1. Set custom httpx_client_factory parameter for SseConnectionParams ```Python from typing import Any import httpx from google.adk.agents.llm_agent import Agent from google.adk.models import LiteLlm from google.adk.tools.mcp_tool import SseConnectionParams, McpToolset # Cutom httpx client factory for MCP SSE transport def custom_httpx_client_factory( headers: dict[str, str] | None = None, timeout: httpx.Timeout | None = None, auth: httpx.Auth | None = None, ) -> httpx.AsyncClient: # Print message for debugging print(' === Custom httpx_client_factory running! ===') kwargs: dict[str, Any] = { "follow_redirects": True, } if timeout is None: kwargs["timeout"] = httpx.Timeout(30, read=300) else: kwargs["timeout"] = timeout if headers is not None: kwargs["headers"] = headers if auth is not None: kwargs["auth"] = auth return httpx.AsyncClient(**kwargs) connection_params = SseConnectionParams( url='http://127.0.0.1:9000/sse', httpx_client_factory=custom_httpx_client_factory, # Custom httpx client factory ) tool = McpToolset(connection_params=connection_params) root_agent = Agent( model=LiteLlm(model="ollama_chat/qwen3.5:35b-a3b-q4_K_M"), name='root_agent', description='A helpful assistant for user questions.', instruction='Answer user questions to the best of your knowledge', tools=[tool], ) ``` 2. Run the Agent, and display the debug log in console ``` 2026-03-15 23:54:14,304 - INFO - agent_loader.py:130 - Found root_agent in my_agent.agent INFO: 127.0.0.1:55642 - "POST /run_sse HTTP/1.1" 200 OK === Custom httpx_client_factory running! === 2026-03-15 23:54:14,345 - INFO - _client.py:1740 - HTTP Request: GET http://127.0.0.1:9000/sse "HTTP/1.1 200 OK" 2026-03-15 23:54:14,349 - INFO - _client.py:1740 - HTTP Request: POST http://127.0.0.1:9000/messages/?session_id=b5750c6e214048c28110dff11c412ecb "HTTP/1.1 202 Accepted" ``` ### Checklist - [x] I have read the [CONTRIBUTING.md](https://github.com/google/adk-python/blob/main/CONTRIBUTING.md) document. - [x] I have performed a self-review of my own code. - [x] I have commented my code, particularly in hard-to-understand areas. - [x] I have added tests that prove my fix is effective or that my feature works. - [x] New and existing unit tests pass locally with my changes. - [x] I have manually tested my changes end-to-end. - [x] Any dependent changes have been merged and published in downstream modules. ### Additional context Co-authored-by: Kathy Wu <wukathy@google.com> COPYBARA_INTEGRATE_REVIEW=#4842 from ushelp:main e3125fb PiperOrigin-RevId: 890708694
- Add 1P BQ Skills `bigquery-ai-ml`, add mandatory reference routing and reference file for `AI.*` syntax. - Include dedicated reference documents for various BigQuery ML features. - Update unit tests to fully assert skill validations. Co-authored-by: Haiyuan Cao <haiyuan@google.com> PiperOrigin-RevId: 890749300
Revert global list extension in deep_merge_dicts and instead explicitly handle list extension only for render_ui_widgets in merge_parallel_function_response_events. Co-authored-by: Vishwa Murugan <vishwamurugan@google.com> PiperOrigin-RevId: 891147765
Close: #4671 Co-authored-by: Xuan Yang <xygoogle@google.com> PiperOrigin-RevId: 891827874
…n intermediate data PiperOrigin-RevId: 891846041
This reduces latency by avoiding redudant network calls Co-authored-by: Kathy Wu <wukathy@google.com> PiperOrigin-RevId: 891931454
Moved `before_model_callback` inside the `call_llm` span. Wrapped both `after_model_callback` and error callbacks with `trace.use_span()` to bind callbacks to the correct overarching span. Added regression tests to verify span ID consistency. Co-authored-by: Haiyuan Cao <haiyuan@google.com> PiperOrigin-RevId: 891952170
Co-authored-by: Liang Wu <wuliang@google.com> PiperOrigin-RevId: 891960308
Fixes an issue where the Agent Analytics plugin could log plain-text OAuth credentials and access tokens to BigQuery. Sensitive keys are now redacted. Co-authored-by: Haiyuan Cao <haiyuan@google.com> PiperOrigin-RevId: 891963630
…into an isolated snapshot sequence before starting loop _list_sessions_impl intermittently would throw `RuntimeError: dictionary changed size during iteration` Co-authored-by: Ankur Sharma <ankusharma@google.com> PiperOrigin-RevId: 892454902
PiperOrigin-RevId: 892466615
… list events PiperOrigin-RevId: 892504394
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
20 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Automated sync of v1 changes from main into v2. The oncall is responsible for reviewing and merging this PR. Resolve conflicts in favor of the v2 implementation.