Added a new sample to show how to bring your own store.

[Sapinder Pal Singh (sapinderpalsingh)]

Summary

Added a Responses protocol hosted-agent sample that demonstrates how to replace all four Foundry hosting storage extension points with custom Redis-backed implementations.

Changes

• Implemented custom stores for:
  • Responses and conversation history
  • Agent session state
  • Workflow checkpoints
  • Function approval requests
• Added deterministic validation tools and a hosted-agent test specification that exercises all four stores, including checkpoint restoration and human approval continuation.
• Added azd and Foundry Toolkit configuration for local development and hosted deployment.

[azure-pipelines]

Azure Pipelines:
There may be pipelines that require an authorized user to comment /azp run to run.

[Ankit Sinha (ankitbko)]

/AzurePipelines run

[azure-pipelines]

Azure Pipelines:
Successfully started running 1 pipeline(s).
1 pipeline(s) were filtered out due to trigger conditions.

[Ankit Sinha (ankitbko)]

/azp run

[azure-pipelines]

Azure Pipelines:
Successfully started running 1 pipeline(s).
1 pipeline(s) were filtered out due to trigger conditions.

[Ankit Sinha (ankitbko)]

/AzurePipelines run

[azure-pipelines]

Azure Pipelines:
Successfully started running 1 pipeline(s).
1 pipeline(s) were filtered out due to trigger conditions.

[Copilot]

Copilot review overview

🔵 Needs a closer look

The hosted-agent persistence scenario and checkpoint restoration validation must be corrected before approval.

Review effort: Lite
Findings: None

Resolved since last review (1)

• Prune expired response IDs from the conversation index

[Copilot]

Copilot review overview

🔵 Needs a closer look

One or more issues must be addressed before approval.

Review effort: Lite
Findings: None

[Copilot]

🟡 Changes recommended

Checkpoint restoration is not exercised, and the requirements-only dependency layout needs documented justification.

2 open findings

• Validate workflow resume and checkpoint restoration · New
• Document exception for requirements-only Python runtime dependencies · New

🧠 Review effort: Lite

---

Give feedback about Copilot approvals in this survey to enter a drawing for a $150 gift card.

[Copilot]

🟡 Changes recommended

The setup omits authentication needed by the deployment hooks, and the tests can pass without proving that the store validation tools ran.

3 open findings

• Assertions do not verify custom store and checkpoint tool invocations · New
• Missing Azure CLI authentication prerequisite for postdeploy hooks · New
• Document exception for requirements-only Python runtime dependencies

1 resolved since last review

• Validate workflow resume and checkpoint restoration

🧠 Review effort: Lite

[Copilot]

🔵 Needs a closer look

The combined persistence contracts, Redis deployment, and identity-access hooks need final human and live-deployment validation.

1 open finding

• Document exception for requirements-only Python runtime dependencies

2 resolved since last review

• Assertions do not verify custom store and checkpoint tool invocations
• Missing Azure CLI authentication prerequisite for postdeploy hooks

🧠 Review effort: Lite

[Copilot]

🔵 Needs a closer look

Custom persistence, cloud provisioning, and identity-based Redis access need final human deployment validation.

1 open finding

• Document exception for requirements-only Python runtime dependencies

🧠 Review effort: Lite

[Copilot]

🔵 Needs a closer look

A continuation bug remains, and the distributed storage deployment and identity hooks need live human validation.

1 open finding

• Explicit previous response is excluded by conversation filter · New

1 resolved since last review

• Document exception for requirements-only Python runtime dependencies

🧠 Review effort: Lite

[Copilot]

🔵 Needs a closer look

An unresolved history-ordering issue remains, and the new persistence and cloud identity setup need human end-to-end validation.

0 open findings

1 resolved since last review

• Explicit previous response is excluded by conversation filter

Previously missed (1)

In code that hasn't changed since last review

Recreated responses retain stale conversation ordering

samples/​python/​hosted-agents/​agent-framework/​responses/​26-custom-store/​src/​agent-framework-agent-custom-store-responses/​redis_response_store.py:154

create_response accepts a deleted response ID, but NX leaves its old conversation score unchanged. If response A is deleted, response B is created, and A is recreated, get_history_item_ids reads A before B even though A is newer, giving the agent history in the wrong order. Give the recreated ID its new sequence score and remove its old conversation membership if the conversation changed.

🧠 Review effort: Lite