Set role='model' in the Content object generated by create_request_input_event()
to ensure proper role attribution when returning function call events to the client.
Co-authored-by: Shangjie Chen <deanchen@google.com>
PiperOrigin-RevId: 929538818
Merge https://github.com/google/adk-python/pull/6031
> **Stacked on #6030** (`fix/experimental-typing`). This PR targets that branch; please review/merge #6030 first, after which this will be retargeted to `main`.
## Summary
Adds `E2BEnvironment`, a `BaseEnvironment` backed by an [E2B](https://e2b.dev) sandbox. It gives agents a persistent remote workspace for shell execution, file CRUD, and on-demand installs (`pip`/`apt`) without touching the host machine.
- The sandbox TTL is bounded to cap credit usage and is extended on each operation; an expired idle sandbox is transparently recreated.
- Lazy-imports the SDK behind a new `e2b` extra, so the base package stays lean.
- Includes a data-analysis sample that downloads a public (GCS-hosted) dataset and analyzes it inside the sandbox.
## Usage
```python
from google.adk.integrations.e2b import E2BEnvironment
from google.adk.tools.environment import EnvironmentToolset
toolset = EnvironmentToolset(environment=E2BEnvironment())
```
## Test plan
- [x] `pytest tests/unittests/integrations/e2b/` (14 passed)
- [x] `pyright src/google/adk/integrations/e2b/_e2b_environment.py` — 0 errors
- [x] Sample agent loads (`contributing/samples/environment_and_skills/e2b_environment`)
- [ ] Manual run against a live E2B sandbox (requires `E2B_API_KEY`)
Co-authored-by: Wei Sun (Jack) <weisun@google.com>
COPYBARA_INTEGRATE_REVIEW=https://github.com/google/adk-python/pull/6031 from google:feat/e2b f2b5584e28c71e7110e9d98e1d0839f575183368
PiperOrigin-RevId: 929443164
Standard REST, SSE, and WebSocket endpoints now actively monitor client disconnections. When a connection drop is intercepted, the executing worker task tree is aborted mid-run to prevent background task leaks, API token burn, and resource wastage.
Includes:
1. REST `/run` connection drop monitor blocking on raw ASGI receive events with 0% CPU consumption.
2. Graceful `499 Client Closed Request` status response on REST disconnections to suppress server-side exceptions.
3. Unit and integration tests validating REST task cancellations and generator finalizations.
4. Comprehensive architectural design documents and user guide under `/docs`.
5. An invocation-focused cooperative counting agent sandbox under `/contributing/samples/core/abort`.
Change-Id: I50fcba7a8bb56cf28b0a3bb8fa66b85732062a67
Update ADK to support the new Interactions API. See: https://ai.google.dev/gemini-api/docs/interactions-breaking-changes-may-2026
Key changes include:
* Updated `interactions_utils.py` to parse new v2 interaction events such as
`InteractionCreatedEvent`, `InteractionCompletedEvent`, and explicit step
types (`ModelOutputStep`, `FunctionCallStep`, etc.).
* Transitioned from "turns" to "steps" terminology, including renaming
`convert_contents_to_turns` to `convert_contents_to_steps`.
* Updated `contributing/samples/interactions_api`
Breaking Changes in interactions_utils.py:
- Renamed "turn" functions to "step" equivalents:
- `convert_content_to_turn` -> `_convert_content_to_step`
- `convert_contents_to_turns` -> `_convert_contents_to_steps`
- `convert_interaction_output_to_part` -> `_convert_interaction_step_to_parts`
- Updated `build_interactions_request_log` to use required `input_steps` instead of `input_turns`.
- Deprecated `convert_part_to_interaction_content`
Change-Id: I962724321342ba0192251896ce51832b214d60d9
Add RequestInputTool to allow single LLM agents to pause, request user input, and resume seamlessly mid-loop without workflow graphs.
Also add comprehensive unit tests, integration tests, and a new official sample with automated session replay tests.
Change-Id: Ifb50cc2e644aeb5ffb2be797255608d0d5a37234
FunctionTool._preprocess_args only converted dict args to a Pydantic
model for single-model and Optional[Model] annotations. A
Union[ModelA, ModelB] parameter was left as a raw dict, so
isinstance checks inside the tool failed with "Unexpected entity
type: <class 'dict'>"
Use pydantic.TypeAdapter to validate against the full Union so
pydantic picks the matching member. None and instances of any
declared union member pass through unchanged; instances of
unrelated BaseModels fall back to the existing graceful-failure
warning path.
Close#5799
Change-Id: Ie69f8efc8395162eac375a0eaad0c77ed2097cec
This change allows AgentEngineSandboxComputer to create new sandboxes
using either a specified sandbox template or a sandbox snapshot. The
environment variables VMAAS_SANDBOX_TEMPLATE_NAME and
VMAAS_SANDBOX_SNAPSHOT_NAME are introduced to configure this behavior.
Co-authored-by: Emily Feng <emilyfeng@google.com>
Change-Id: Iebdd980a16966ba765cacbce6d63d0d5b691650a
Introduces a generate_chart tool to the Data Agent sample, leveraging
Altair and vl-convert to render Vega-Lite specifications into charts.
Co-authored-by: Han Cao <huanc@google.com>
Change-Id: I5765487406d511e650091f5dc884102c43568fd4
Workflows failed because the gemini-2.5-pro model is not available for our projects and the CONTRIBUTING.md relative path resolved incorrectly after relocation. Updated the model to gemini-3.5-flash and adjusted the path resolver depth.
Change-Id: Iffc8a2b0044f3d73b0cb0b364db220d2bd055683
JSON test files generated by the agent test runner were missing trailing newlines, causing the pre-commit end-of-file-fixer hook to fail. Added an explicit newline write on serialization.
Change-Id: Idaffc4f2c23fa213f595ddef2188b0050aeb431d
Generated test JSON files for parallel_functions and pydantic_argument samples to verify their behavior.
Also fixed a test replay issue by ignoring timestamps in state (in agent_test_runner.py) to prevent false failures due to dynamic data.
Note: Some sample agents could not have tests generated successfully due to:
1. agent_tool_with_grounding_metadata: Fails on replay because AgentTool hides sub-agent events, causing mock LLM to get out of sync.
2. hello_world_stream_fc_args: Fails on rebuild because streaming function calls are not supported in the Unary API used by the test recorder.
3. output_schema_with_tools: Has external dependencies (Wikipedia/Google Search), making it unsuitable for automated regression tests without proper mocking.
Change-Id: I69df7daf49d175905dcbd85cf70914e460717915
This CL implements the class in the integrations folder, used specifically for the Skill Registry API.
Co-authored-by: Kathy Wu <wukathy@google.com>
PiperOrigin-RevId: 915627057
`gemini-1.*` and `gemini-2.0*` models are respectively deprecated and scheduled for shutdown on June 1, 2026. `gemini-2.5*` models are their successors.
No regressions in unit tests:
```
========================================================================================== 5583 passed, 2237 warnings in 84.91s (0:01:24) ===========================================================================================
```
PiperOrigin-RevId: 907663315
## Overview
This PR updates internal tutorials and demonstration guidance following the renaming of the database connectivity tools (`genai-toolbox` -> `mcp-toolbox`) and aligns resource downloads with the newly scoped artifact destinations.
## Changes
* Removed references to legacy `googleapis.github.io/genai-toolbox` pages.
* Updated download commands to pull binaries directly from the rebranded bucket.
* Bumped the embedded curl instruction default payload target to the newly stable `v1.1.0` release tier.
PiperOrigin-RevId: 900207615
Merge https://github.com/google/adk-python/pull/5218
### Link to Issue or Description of Change
**1. Link to an existing issue (if applicable):**
(None, this is a new sample)
**2. Or, if no issue exists, describe the change:**
**Problem:**
There was no sample demonstrating how to use `LocalEnvironment` with `EnvironmentToolset` for manual skill activation (discovering and loading skills by reading files) instead of using the pre-configured `SkillToolset`.
**Solution:**
Created a new sample `local_environment_skill` that demonstrates this pattern:
- Agent is instructed to find skills in `skills/` folder using the `find` command.
- Agent loads skills by reading `SKILL.md` using the `ReadFile` tool.
- Added a weather skill sample and reference data to demonstrate usage.
### Testing Plan
**Unit Tests:**
*(This is a sample/demonstration, so no new unit tests were added for the framework itself.)*
**Manual End-to-End (E2E) Tests:**
1. Ran `adk web contributing/samples`.
2. Interacted with the agent in the Web UI.
3. Prompt: "Can you check the weather in Sunnyvale?"
4. Verified that the agent successfully used `find` to locate the skill, read `SKILL.md`, read `weather_info.md`, and answered correctly.
### 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 manually tested my changes end-to-end.
COPYBARA_INTEGRATE_REVIEW=https://github.com/google/adk-python/pull/5218 from google:samples/local_env_skill 39404309e8b9237a279ecd6ed14bb87eb84098f0
PiperOrigin-RevId: 897229752
We recently overhauled our MCP Toolbox documentation and moved it to a new domain. This PR propagates that URL change here.
PiperOrigin-RevId: 896118946
We changed skills to use kebab case and created weather-skill, but forgot to get rid of weather_skill in the sample
Co-authored-by: Kathy Wu <wukathy@google.com>
PiperOrigin-RevId: 893597141
The `receive` method now accumulates function calls from multiple `LiveServerMessage` instances. These accumulated tool calls are yielded as a single `LlmResponse` containing all function call parts only when a turn_complete message is received.
Without the change, the tool_1's response is sent to the model as soon as it's generated, triggering a second call for tool_2. Upon receiving two consecutive tool_2's responses, the model utters the same message twice.
Fixes issue #4902
Co-authored-by: Liang Wu <wuliang@google.com>
PiperOrigin-RevId: 893197482
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
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
This change introduces a new `SpannerAdminToolset` with tools for managing Google Cloud Spanner resources. The toolset includes functions to list and get details of Spanner instances and instance configurations, and to create, list databases. The new toolset is marked as experimental. Unit tests for the new admin tools are also added
PiperOrigin-RevId: 885116111
Shangjie Chen