* Tool Client V1 Version

* Langraph integration

* Updates fixing langgraph adapter

* fix build

* fix cspel

* fix cspell

* Add ToolClient integration with Agent Framework and examples for dynamic agent creation

* Made changes to return tools instead of toolclient

* Address comments

---------

Co-authored-by: Lu Sun <lusu@microsoft.com>

…e-sdk-for-python into lusu/agentserver-1110

* fix streaming issue in af

* fix streaming issue in af

* update version to 1.0.0b5

…t Handling

* Refactor Azure AI Tool Client Configuration and Enhance OAuth Consent Handling

- Consolidated the AzureAIToolClientConfiguration class by removing redundant code and improving clarity.
- Introduced OAuth consent handling in the agent's response methods to manage OAuthConsentRequiredError.
- Updated the FoundryCBAgent to configure tools endpoint and agent name from environment variables.
- Enhanced tool client to propagate OAuth consent errors for better handling in the agent.
- Added methods to generate OAuth request IDs and handle OAuth consent requests in the LangGraph response converter.
- Updated sample usage to include tool connection ID from environment variables.
- Incremented version to 1.0.0b5 for the langgraph package.

* Address Pylint and mypy issues

* Updated change logs

…1110

… fields

…44930)

…ader (#44929)

* use git worktree

* [agentserver] Attach package metadata to OpenAIResponse.metadata + header

* [agentserver] Attach package metadata to OpenAIResponse.metadata + header

…ork (#45004)

* [Hosted Agents] Implement managed checkpoints feature for AgentFramework

* misc: use project_endpoint instead of project_id and foundry_endpoint

* misc: use get_project_endpoint

* misc: refine checkpoint init logic

* misc: Add a mutual-exclusion check

* misc: move the check into workflow_agent_adapter

Some MCP tool servers (e.g. HuggingFace) return tool manifests where
certain parameters have an empty string for their JSON Schema 'type'
field. This causes a Pydantic ValidationError when the SDK tries to
deserialise the ListFoundryConnectedToolsResponse, because SchemaType
only accepts the six standard JSON Schema types.

Changes:
- SchemaProperty.type is now Optional[SchemaType] with a model_validator
  that coerces empty type strings to None.
- FoundryToolClient._create_ai_function skips parameters whose type is
  None with a warning log, instead of crashing with an AttributeError.

…45041)

* [Hosted Agents] Implement managed checkpoints feature for Langgraph

* misc: preserves all other compile parameters

* chore: refactor AgentFramework to remove managed_checkpoints flag

* chore: refactor langgraph to remove managed_checkpoints flag

…d langgraph (#45064)

* [Hosted Agents] Add foundry checkpoints samples for agentframework and langgraph

* docs: Add README files for Foundry checkpoint samples

Co-authored-by: Declan <lixiaoli@microsoft.com>

Applies the same guard as the agentframework fix (PR #45051) to the
LangGraph tool resolver. When an MCP tool manifest contains a property
with an empty or unrecognised schema type, the resolver now skips that
field with a warning instead of crashing with AttributeError.

…rovided (#45103)

* implementing

* implemented conversation item converter

* refined with hitl filter

* update hitl helper to filter out hitl messages

* refined conversation repo and store

* create foundry conversation thread repository by default

* refined repo and store

* refine

* remove unused changes

* remove unused changes

* check conversation id

* update version

* fix minors

* integrate conversation history with langgraph

* use asyncOpenAI client directly

* modify unit test

* save items in core package

* add unittests

* add change log

* minor bug fix

* remove some debug logs

---------

Co-authored-by: Declan <lixiaoli@microsoft.com>