feat: support foundry checkpointer

[Alex Wang (a1exwang)]

Summary

1. Adds FoundryCheckpointSaver, an async LangGraph checkpointer backed by Foundry State Store with typed checkpoint and pending-write schemas and User-Agent propagation.

2. Updates checkpointed Responses samples to use FoundryCheckpointSaver when hosted while preserving their existing in-memory or SQLite savers locally. Deployment manifests now use Responses protocol 2.0.0 for Foundry request-context propagation.

Note: disabled user isolation for now to workaround platform bug, will enable after they fix it.

[Copilot]

Pull request overview

Adds a new Foundry-backed LangGraph checkpointer to the langchain_azure_ai hosting surface area, and updates Responses-protocol hosted agent samples/manifests to use it (plus protocol 2.0.0) so checkpointed conversations can survive container replacement in Foundry deployments.

Changes:

• Introduces FoundryCheckpointSaver (async LangGraph BaseCheckpointSaver) backed by Foundry State Store, including pending-write persistence and User-Agent propagation.
• Updates multiple hosted Responses samples to choose FoundryCheckpointSaver when hosted and keep existing in-memory savers locally; updates deployment manifests to Responses protocol 2.0.0.
• Adds/updates unit tests and documentation to cover the new saver and re-export surface.

Reviewed changes

Copilot reviewed 16 out of 16 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
samples/hosting/langgraph-hosted-agents/responses/08_hitl/README.md	Updates HITL sample docs to describe Foundry-backed checkpointing and protocol requirements.
samples/hosting/langgraph-hosted-agents/responses/08_hitl/main.py	Switches hosted runs to FoundryCheckpointSaver and async server startup.
samples/hosting/langgraph-hosted-agents/responses/08_hitl/agent.yaml	Bumps Responses protocol version to 2.0.0.
samples/hosting/langgraph-hosted-agents/responses/08_hitl/agent.manifest.yaml	Bumps Responses protocol version to 2.0.0.
samples/hosting/langgraph-hosted-agents/responses/05_workflows/README.md	Updates docs for hosted vs local checkpointers and async server startup.
samples/hosting/langgraph-hosted-agents/responses/05_workflows/main.py	Switches hosted runs to FoundryCheckpointSaver and async server startup.
samples/hosting/langgraph-hosted-agents/responses/05_workflows/agent.yaml	Bumps Responses protocol version to 2.0.0.
samples/hosting/langgraph-hosted-agents/responses/05_workflows/agent.manifest.yaml	Bumps Responses protocol version to 2.0.0.
samples/hosting/langgraph-hosted-agents/responses/01_basic/main.py	Adds a state-store probe hook into the host server’s request pipeline.
samples/hosting/langgraph-hosted-agents/responses/01_basic/agent.yaml	Bumps Responses protocol version to 2.0.0.
samples/hosting/langgraph-hosted-agents/responses/01_basic/agent.manifest.yaml	Bumps Responses protocol version to 2.0.0 and updates the model id.
libs/azure-ai/tests/unit_tests/agents/hosting/test_types.py	Ensures FoundryCheckpointSaver is re-exported from hosting and has the expected module.
libs/azure-ai/tests/unit_tests/agents/hosting/test_foundry_checkpoint_saver.py	Adds unit tests covering checkpoint/pending-write roundtrips, delete behavior, and UA stamping.
libs/azure-ai/README.md	Documents how to use FoundryCheckpointSaver with ResponsesHostServer.run_async().
libs/azure-ai/langchain_azure_ai/agents/hosting/_foundry_checkpoint_saver.py	Implements the new Foundry-backed LangGraph checkpoint saver.
libs/azure-ai/langchain_azure_ai/agents/hosting/init.py	Adds FoundryCheckpointSaver to lazy re-exports.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.