* ci: pin actions to SHAs and harden workflows with zizmor auto-fixes
* ci: set least-privilege GITHUB_TOKEN permissions on workflows
* ci: move PR/issue context into env vars to prevent template injection
* docs(integrations): add Latitude observability integration
Add a Latitude integration page under docs/integrations with an
observability catalog tag, a square card logo, and a complete,
runnable ADK example using Latitude's Python telemetry SDK.
* Add evaluation tag, env-var credentials, and clean up copy/links
---------
Co-authored-by: Kristopher Overholt <koverholt@google.com>
* docs(integrations): add Agent Threat Rules (ATR) guardrail plugin
Adds docs/integrations/atr-guardrail.md (+ icon) for the adk-atr-guardrail
package -- an in-process, deterministic security guardrail that enforces the
open-source Agent Threat Rules ruleset at three fail-closed Runner callbacks
(before_run / before_model / before_tool). Tagged 'resilience'. Created per
maintainer guidance on google/adk-python#6130 to ship as a standalone package
listed in the integrations catalog. Verified locally with mkdocs build.
* Use gemini-flash-latest, switch to the App plugin pattern with ADK >= 2.0, and remove the resilience catalog tag
---------
Co-authored-by: Kristopher Overholt <koverholt@google.com>
* Add MLflow scorers integration page
Adds a new integration page at docs/integrations/mlflow-scorers.md
covering MLflow's five Google ADK scorers (ToolTrajectory, ResponseMatch,
ResponseEvaluation, Safety, Hallucination) for agent evaluation. The
integration wraps ADK's TrajectoryEvaluator, RougeEvaluator,
FinalResponseMatchV2Evaluator, SafetyEvaluatorV1, and HallucinationsV1Evaluator
behind MLflow's scorer interface, so ADK users can evaluate agents inside
mlflow.genai.evaluate() runs without leaving the ADK ecosystem.
Complements the existing MLflow Tracing and MLflow AI Gateway integration
pages by covering evaluation, the third leg of the MLflow stack for ADK.
Signed-off-by: debu-sinha <debusinha2009@gmail.com>
* Trigger CLA re-check
Signed-off-by: debu-sinha <debusinha2009@gmail.com>
* Use version-agnostic Gemini aliases and refine copy
* Update category tags for other pages
---------
Signed-off-by: debu-sinha <debusinha2009@gmail.com>
Co-authored-by: Kristopher Overholt <koverholt@google.com>
* Add Aerospike integration to ADK catalog
Document adk-aerospike session, memory, and artifact services with tabbed
Python examples, URI configuration, storage layout, and catalog assets.
* Fix adk web CLI flags and minor formatting edits
---------
Co-authored-by: Kristopher Overholt <koverholt@google.com>
* UpdateContext.md
* Update index.md
* Update index.md
noticed that the hierarchy of bullet points was not correct
* Update context and callbackcontext.md
* Enhance BigQuery Agent Analytics with reliability fixes
Added reliability and observability fixes for BigQuery Agent Analytics, including cross-region storage write routing, dropped-event tracking, and improved span management in Cloud Trace.
* Fix markdown formatting in BQAA doc
Resolve the code-fence/content-tab indentation issues flagged in review:
- Full example: restore the === "Python" content tab and indent the whole
Python code block uniformly to 8 spaces so it is a proper sibling of the
=== "Java" tab (previously the block was half-de-indented with an orphaned
opening fence next to a lone Java tab).
- Public methods: move the Python list inside the === "Python" tab (it had
been pasted at column 0, outside the tab) and remove the duplicated list
that ended up wrongly nested under the "Async context manager" bullet.
* Revise BigQuery Agent Analytics documentation
Updated version requirements and added details about ADK 2.0 multi-agent workflow features and event types. Enhanced descriptions for internal span tracking and storage behavior.
* docs: Document GCS first party toolset
* Update gcs.md for clarity and style considerations
* Update gcs.md
---------
Co-authored-by: Joe Fernandez <931947+joefernandez@users.noreply.github.com>
* Update index.md
1. Updated the evaluation page, added the 4th evaluation "adk conformance", 2. removed numbering in headers as suggested, 3. improved the "notes" and "warnings" that had an incorrect formatting.
* Update index.md
* Update index.md
worked on all your comments
* Update index.md
* Add context graph section to BigQuery Agent Analytics docs
Added a section on the context graph, detailing its purpose, structure, and operational references for materialization and deployment.
* Add files via upload
---------
Co-authored-by: Joe Fernandez <931947+joefernandez@users.noreply.github.com>
* docs(gke): add Workload Identity IAM binding for adk deploy gke
The automated deployment path (`adk deploy gke`) generates a manifest
that uses the `default` Kubernetes service account in the `default`
namespace, but the docs only documented the Workload Identity IAM
binding for the manual path (which uses a custom `adk-agent-sa`).
Without binding `roles/aiplatform.user` to the `default` service
account, agent pods start but model requests fail with
`403 PERMISSION_DENIED`.
Add a "Configure Workload Identity for Agent Platform" step under
Option 2 with the correct binding for the `default` service account,
and cross-reference it from the 403 troubleshooting entry.
Fixes#1254
* Update gke.md
---------
Co-authored-by: Joe Fernandez <931947+joefernandez@users.noreply.github.com>
* Update compaction.md
Added the explanation for token compaction, and a header section for sliding window section, added the section: What is compaction?
* Update compaction.md
Worked on changes
* Update compaction.md
* Update compaction.md
worked on other 2 comments
* Update compaction.md
* docs: Document concurrency and locking in DatabaseSessionService
* Concurrencyandlocking.md
instead of a note, created a subsection
* Concurrencyandlocking.md
---------
Co-authored-by: Zyan <zyanya@google.com>
* Update version for ADK Kotlin to 0.2.0
* Update version for ADK Kotlin examples to 0.2.0
* Update install steps on multi-tool-agent page
* Update ADK Kotlin API reference docs to 0.2.0
* Remove runner parameter from AdkWebServer constructor
* docs(integrations): refresh e2a — hosted endpoint + 7 missing tools
The e2a integration page has drifted from what the e2a MCP server
actually ships. Two gaps this PR closes:
1. **Hosted endpoint not documented.** e2a runs a hosted Streamable
HTTP MCP at https://mcp.e2a.dev/mcp that accepts an API key in
the Authorization header. The page currently only shows the
stdio path (npx -y @e2a/mcp-server), which doesn't work for ADK
agents deployed to Cloud Run since Cloud Run can't host stdio
MCP child processes. Adds a "Hosted MCP Server" tab to both the
Python and TypeScript code examples using
StreamableHTTPConnectionParams.
2. **7 tools missing from the table.** The MCP server now registers
18 tools but the page lists 11. Adds:
Identity: update_agent, delete_agent
Messages: get_attachment_data
Domains: list_domains, register_domain, verify_domain,
delete_domain (new section)
Also clarifies a few descriptions:
- `whoami` notes E2A_AGENT_EMAIL is only required for multi-agent
accounts (the hosted endpoint auto-resolves the sole agent at
session init).
- `delete_agent` and `delete_domain` flag the `confirm: true`
required arg.
- `get_attachment_data` uses 0-based `attachment_index`, not an
attachment ID.
- `E2A_BASE_URL` clarifies the var has no effect when connecting
to the public hosted endpoint.
Cross-linked the cloud-mode webhook example (already shipped in
the e2a repo, not previously referenced from this page) and the
hosted endpoint URL under Additional resources.
No frontmatter / catalog metadata changes — title, description,
icon, and tags (`["mcp"]`) all unchanged.
* fix: import McpToolset from deep path (google-adk 2.0 compat)
google-adk 2.0 stopped re-exporting McpToolset from
`google.adk.tools.mcp_tool`'s package __init__ — only the deep path
`google.adk.tools.mcp_tool.mcp_toolset.McpToolset` still works.
Caught when I actually ran the example against a fresh
google-adk 2.0 install:
ImportError: cannot import name 'McpToolset' from
'google.adk.tools.mcp_tool'
($VENV/site-packages/google/adk/tools/mcp_tool/__init__.py)
Applies to both the stdio code block and the new hosted code block.
StdioConnectionParams and StreamableHTTPConnectionParams already use
deep paths and continue to work — only McpToolset was broken.
* Revert McpToolset import to standard path
The claim that google-adk 2.0 stopped re-exporting McpToolset from
google.adk.tools.mcp_tool is incorrect. The __init__.py at
src/google/adk/tools/mcp_tool/__init__.py explicitly re-exports
McpToolset (and MCPToolset) via `from .mcp_toolset import McpToolset`.
It has done so in every release, including v2.0.0.
* Rename tab to Remote MCP server for consistency with all other integration pages
* Remove link to raw MCP endpoint
* Match API key variable / format with all other integration pages
* Formatting and consistency edits
---------
Co-authored-by: Kristopher Overholt <koverholt@google.com>
* Updated the python code sample for routing.
* Simplified the routing sample
* Update routes.md
add more indent to "if" statement return for readability
---------
Co-authored-by: Joe Fernandez <joefernandez@users.noreply.github.com>
Kristopher Overholt