Skip to content

[WIP] Fix Northbound OTel propagation and streaming span lifetime - #4334

Merged
JasonW404 merged 3 commits into
hotfix/v2.8.5from
fix/v285-northbound-otel
Nov 12, 2026
Merged

JasonW404 merged 3 commits into
hotfix/v2.8.5from
fix/v285-northbound-otel

Conversation

@JasonW404

@JasonW404 JasonW404 commented Nov 12, 2026 •

Copy link
Copy Markdown
Member

Northbound requests previously had HTTP tracing disabled, lacked monitoring configuration in deployment, and did not forward W3C trace context to Runtime. Enable the existing instrumentation and propagate traceparent/tracestate alongside the internal JWT so Northbound, Runtime, Agent and tool spans share one trace. Keep the Northbound server span open through completion of the final ASGI send, including failure and disconnect handling.

This PR is split from #4328 and targets hotfix/v2.8.5. It contains only Northbound tracing, Docker/Helm monitoring configuration, and corresponding tests/formal assets. Agent output parsing and tool-result history remain in #4328. No new telemetry dependency or global HTTPX instrumentation is added.

Validation after splitting:

  • 6 propagation/configuration tests and 24 Helm environment-injection tests passed.
  • The subprocess wrapper passed all 8 real ASGI lifecycle cases. Corrected its checkout-root calculation so it tests this branch without relying on an installed SDK.
  • Selected formal asset validation passed. No fresh VM/container/model deployment acceptance was run for this split; product files are byte-identical to the previously validated Northbound changes from fix: 修复 v2.8.5 工具结果历史、文本动作解析、空预留输出上限及北向 OTel #4328.

Deployment: rebuild the shared backend image and recreate affected backend containers with the updated Compose/Helm monitoring configuration. Enable telemetry; explicit URL allowlists must include /nb/ and /agent/internal/northbound/. Span completion means the final server ASGI send returned, not acknowledgement by the client.

@JasonW404
JasonW404 changed the base branch from main to hotfix/v2.8.5 November 12, 2026 06:11
@JasonW404
JasonW404 marked this pull request as ready for review November 12, 2026 06:21
@JasonW404
JasonW404 merged commit 40690dd into hotfix/v2.8.5 Nov 12, 2026
4 checks passed
jeffwu-1999 pushed a commit that referenced this pull request Nov 13, 2026
* chore: bootstrap output protocol and history patch

* fix(agent): converge output parsing and replay runtime-only history

* test(agent): align history and case-insensitive protocol regressions

* fix(agent): retain runtime history echo protection

* fix(agent): defer mixed final answer streaming until execution confirmation

* test: refresh reviewed prompt baselines for protocol patch

* fix(observability): trace northbound requests through complete streaming responses

* test: align northbound monitoring environment precedence

* fix(agent): restore legacy history rendering and deduplicate results

* Revert "test: align northbound monitoring environment precedence"

This reverts commit 7995430.

* Revert "fix(observability): trace northbound requests through complete streaming responses"

This reverts commit 36109ae.

* [WIP] Fix Northbound OTel propagation and streaming span lifetime (#4334)

* fix(observability): trace northbound requests through complete streaming responses

* test: align northbound monitoring environment precedence

* test: keep split northbound validation rooted in its checkout

* chore: set v2.8.5 release version

* fix(agent): deduplicate sandbox string representations in history

* chore: remove unrelated PR4328 formatting and verify legacy trace route

* fix: restore ordinary mixed-answer streaming in CodeAgent

* fix: omit implicit output cap when default reserve is unset
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant