Found docs updates needed from ADK python release v1.25.1 to v1.26.0

[adk-bot]

To facilitate the release of ADK Python v1.26.0, the following documentation updates are recommended based on the changes compared to v1.25.1.

Compare v1.25.1...v1.26.0

---

1. Update Context documentation to reflect the new unified Context class

Doc file: docs/context/index.md

Current state:

Describes InvocationContext, ReadonlyContext, CallbackContext, and ToolContext as distinct classes.

Proposed Change:

Update to introduce the new unified Context class. Explain that CallbackContext is now an alias for Context. Update examples to use Context features like add_memory directly.

Reasoning:
The CallbackContext class has been refactored to be an alias for the new unified Context class, which provides a superset of features (artifacts, credentials, memory, state).

Reference: src/google/adk/agents/context.py

2. Document the new AgentRegistry integration class

Doc file: docs/integrations/api-registry.md

Current state:

Describes google.adk.tools.api_registry.ApiRegistry for MCP tools.

Proposed Change:

Introduce the new google.adk.integrations.agent_registry.AgentRegistry class. Explain it supports retrieving both MCP toolsets (get_mcp_toolset) and remote A2A agents (get_remote_a2a_agent). Clarify if this replaces the old ApiRegistry tool.

Reasoning:
A new AgentRegistry client was added in src/google/adk/integrations/agent_registry which seems to be a more comprehensive client for the Agent Registry service.

Reference: src/google/adk/integrations/agent_registry/agent_registry.py

3. Update Apigee docs to reflect OpenAI-compatible API support

Doc file: docs/agents/models/apigee.md

Current state:

The ApigeeLLM wrapper is currently designed for use with Vertex AI and the Gemini API (generateContent).

Proposed Change:

Update to state that ApigeeLlm now supports OpenAI-compatible APIs (like /chat/completions). Add examples of how to configure it for OpenAI models using apigee/openai/... model strings or the api_type parameter.

Reasoning:
The ApigeeLlm class was significantly updated to support OpenAI-compatible APIs via Apigee proxies, including a new CompletionsHTTPClient.

Reference: src/google/adk/models/apigee_llm.py

4. Update SkillToolset documentation

Doc file: docs/skills/index.md

Current state:

Not mentioned.

Proposed Change:

Update docs to reflect changes in SkillToolset. Mention DEFAULT_SKILL_SYSTEM_INSTRUCTION and how skills are processed (XML format). Also update frontmatter validation rules (name pattern, length).

Reasoning:
SkillToolset and prompt.py have been updated to standardize skill instructions and add validation.

Reference: src/google/adk/tools/skill_toolset.py

5. Update Spanner tool docs to reflect async functions

Doc file: docs/integrations/spanner.md

Current state:

Not mentioned.

Proposed Change:

Update documentation to note that execute_sql and similarity_search in google.adk.tools.spanner are now async functions and must be awaited.

Reasoning:
The Spanner tool functions were converted to async to avoid blocking the event loop. This is a breaking change for synchronous usage.

Reference: src/google/adk/tools/spanner/query_tool.py

6. Document new add_memory method for direct memory management

Doc file: docs/sessions/memory.md

Current state:

The doc describes add_session_to_memory as the primary way to ingest information.

Proposed Change:

Add documentation for the new add_memory method which allows explicit memory items to be added directly. Explain how VertexAiMemoryBankService supports this via memories.create or memories.generate (with consolidation) based on configuration.

Reasoning:
BaseMemoryService and VertexAiMemoryBankService now support an add_memory method for direct memory ingestion, offering more control than just add_session_to_memory.

Reference: src/google/adk/memory/vertex_ai_memory_bank_service.py

7. Document token-threshold compaction in SingleFlow

Doc file: docs/context/compaction.md

Current state:

Not mentioned.

Proposed Change:

Document the new token-threshold compaction feature. Explain how SingleFlow now includes a CompactionRequestProcessor that triggers compaction based on EventsCompactionConfig.

Reasoning:
Compaction is now integrated into SingleFlow via CompactionRequestProcessor, allowing automatic event compaction based on token thresholds.

Reference: src/google/adk/flows/llm_flows/compaction.py

8. Document ID Token support for Service Account authentication

Doc file: docs/tools-custom/authentication.md

Current state:

Mentions SERVICE_ACCOUNT auth type but only describes access token exchange.

Proposed Change:

Add documentation for ID Token support in Service Account authentication. Explain that setting use_id_token=True allows exchanging service account credentials for an ID token (OIDC), which is required for invoking Cloud Run/Functions services.

Reasoning:
ServiceAccountCredentialExchanger was updated to support ID token generation, enabling authentication for Cloud Run/Functions.

Reference: src/google/adk/tools/openapi_tool/auth/credential_exchangers/service_account_exchanger.py

9. Update LiteLLM docs to mention newly supported providers

Doc file: docs/agents/models/litellm.md

Current state:

Mentions OpenAI, Anthropic, Cohere.

Proposed Change:

Add Azure, AWS Bedrock, Mistral, DeepSeek, and other newly supported providers to the list of examples or supported providers.

Reasoning:
The LiteLlm class updated its supported model regex patterns to include many more providers.

Reference: src/google/adk/models/lite_llm.py

10. Document new SessionNotFoundError exception

Doc file: docs/runtime/index.md

Current state:

Not mentioned.

Proposed Change:

Add documentation for SessionNotFoundError in the error handling section or where runners are discussed. Mention it is raised when a session cannot be found.

Reasoning:
New exception class SessionNotFoundError was added and is used in runners.py. Users might need to handle this exception explicitly.

Reference: src/google/adk/errors/session_not_found_error.py

11. Document new push_config_store parameter in to_a2a function

Doc file: docs/a2a/quickstart-exposing.md

Current state:

The to_a2a function is documented with root_agent and port parameters.

Proposed Change:

Update the to_a2a function documentation to include the new push_config_store parameter. Explain that it allows configuring a custom store for push notifications, defaulting to in-memory if not provided.

Reasoning:
The to_a2a utility function now accepts a push_config_store argument to support A2A push notifications.

Reference: src/google/adk/a2a/utils/agent_to_a2a.py

12. Document --memory_service_uri CLI option

Doc file: docs/runtime/command-line.md

Current state:

The storage options table lists --session_service_uri and --artifact_service_uri.

Proposed Change:

Add --memory_service_uri to the storage options table and the All options table.

Reasoning:
The CLI now supports specifying a custom memory service URI, which is used to configure the MemoryService for the runner.

Reference: src/google/adk/cli/cli.py

13. Document MCP Session Persistence support

Doc file: docs/mcp/index.md

Current state:

Not mentioned.

Proposed Change:

Add a note that MCPSessionManager now supports pickling, which enables session persistence when using MCP tools.

Reasoning:
McpSessionManager added __getstate__ and __setstate__ methods to support pickling, allowing the state of MCP sessions to be saved and restored.

Reference: src/google/adk/tools/mcp_tool/mcp_session_manager.py

14. Update version support note for User Simulation to reflect new Persona features in v1.26.0

Doc file: docs/evaluate/user-sim.md

Current state:

Supported in ADKPython v1.18.0

Proposed Change:

Update the version to v1.26.0 or add a note that Personas were added in v1.26.0.

Reasoning:
The User Simulation feature received a major update in v1.26.0 with the introduction of Personas and Jinja templates. It might be confusing if the doc says v1.18.0 without clarifying when these specific new features were added.

Reference: src/google/adk/evaluation/simulation/pre_built_personas.py