Python: Preserve OTel parent context for deferred streams

[Eduard van Valkenburg (eavanvalkenburg)]

Motivation & Context

Streaming hosting paths can defer target execution until after the route handler has returned a streaming response. In that deferred phase, inbound request span context may no longer be current, so agent/chat/tool spans can lose strict parent-child linkage. This change preserves parent context for deferred streaming in shared hosting stream adapters and adds regression coverage for both host-level and Responses-channel paths.

Description & Review Guide

• What are the major changes?
  • Capture current OpenTelemetry context when host-managed streaming is opened.
  • Re-activate the captured context during deferred stream pulls and get_final_response() in _BoundResponseStream.
  • Add host-level regression test proving deferred consumption still observes the expected parent span context.
  • Add Responses-channel integration test proving streaming request handling keeps parent span propagation.
• What is the impact of these changes?
  • Improves trace correctness for hosted streaming scenarios by restoring strict parent-child span linkage when an inbound request span exists.
  • Applies host-wide, so all channels using shared host streaming behavior benefit, while preserving existing stream lifecycle semantics.
• What do you want reviewers to focus on?
  • Context capture/activation boundaries in _BoundResponseStream and whether the new behavior preserves existing close/finalization guarantees.

Related Issue

Fixes #6708

Contribution Checklist

• The code builds clean without any errors or warnings
• All unit tests pass, and I have added new tests where possible
• The PR follows the Contribution Guidelines
• This PR is linked to an issue and there is no other open PR for this issue (see Related Issue above).
• This is not a breaking change. If it is a breaking change, add the breaking change label (or add "[BREAKING]" to the title prefix, before or after any language prefix) — a workflow keeps the label and title prefix in sync automatically.

[github-actions]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/hosting/agent_framework_hosting
_host.py	462	50	89%	90, 145, 160, 188, 258, 268–270, 272–273, 279, 402, 410–411, 414, 416, 423, 429, 432–434, 437, 455, 781–783, 790–794, 800–801, 804, 810–811, 958–959, 966, 1074, 1111, 1282–1285, 1367–1368, 1373–1375
TOTAL	42174	4981	88%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
8305	37 💤	0 ❌	0 🔥	2m 8s ⏱️

[github-actions]

Automated Code Review

Reviewers: 5 | Confidence: 91% | Result: All clear

Reviewed: Correctness, Security Reliability, Test Coverage, Failure Modes, Design Approach

---

Automated review by eavanvalkenburg's agents

[Copilot]

Pull request overview

This PR improves OpenTelemetry trace parent/child linkage for deferred streaming in the Python hosting layer by capturing the active OTel context when a stream is opened and re-activating it during later stream pulls and finalization, with regression tests covering both host-level behavior and the Responses-channel SSE path.

Changes:

• Capture the current OTel context at host-managed stream construction time and re-activate it during deferred iteration and get_final_response() in _BoundResponseStream.
• Add a host-level regression test that simulates deferred consumption (construct stream under a parent span, consume after the span context is detached).
• Add a Responses-channel test asserting streaming invocations observe the request’s parent span context.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
python/packages/hosting/agent_framework_hosting/_host.py	Captures and re-activates OTel context for deferred streaming via _BoundResponseStream.
python/packages/hosting/tests/hosting/test_host.py	Adds a regression test ensuring deferred stream pulls/finalization preserve the captured parent span context.
python/packages/hosting-responses/tests/hosting_responses/test_channel.py	Adds a Responses-channel SSE test to validate parent span context propagation during streaming requests.

[Tao Chen (TaoChenOSU)]

nit: Can you add two screenshots for before and after the fix showing the parent child relationship? It can be any monitoring backend.

[Eduard van Valkenburg (eavanvalkenburg)]

Tao Chen (@TaoChenOSU) This is after, don't have a running from before anymore: