feat(streaming): configurable nginx, controller, and UI inactivity ti…

[yanivmn]

…meouts

Helm chart (helm/kagent/)

• values.yaml:
  • Add ui.streamTimeoutSeconds (default 1800) - client-side chat stream inactivity timeout, propagated to the browser via the UI ConfigMap.
  • Add ui.nginx.proxyReadTimeout / proxySendTimeout (default 1800s) - templated into the UI sidecar nginx config in place of the previously hard-coded per-location values.
  • Add ui.openshiftRoute.annotations with haproxy.router.openshift.io/timeout: 120m required for A2A/SSE streaming through the OpenShift router (default ~60s caused ERR_INCOMPLETE_CHUNKED_ENCODING in the browser).
  • Keep controller.streaming.timeout at 1800s (pre-PR default).
• files/nginx.conf: replace hard-coded proxy_read_timeout / proxy_send_timeout values (600s / 1800s / 3600s) with a single templated value driven by ui.nginx.*.
• templates/openshift-route.yaml: render ui.openshiftRoute.annotations.
• templates/ui-deployment.yaml: expose streamTimeoutSeconds to the UI container.

UI (ui/)

• Add ui/src/app/actions/config.ts to load runtime config (stream timeout).
• ui/src/app/a2a/[namespace]/[agentName]/route.ts: keep upstream connection alive while the stream is open.
• ui/src/components/chat/ChatInterface.tsx: drive the EventSource inactivity timeout from the configured streamTimeoutSeconds and add keep-alive timers to prevent idle browser disconnects.

Authored-by: Yuri Khrakovski

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR aims to make A2A/SSE streaming more robust by extending proxy/browser timeouts and making the client-side stream inactivity timeout configurable at runtime (via Helm-provided env).

Changes:

• Add runtime UI config lookup for stream inactivity timeout and use it in the chat stream consumer.
• Start SSE keep-alives earlier in the A2A proxy route to avoid HAProxy idle timeouts before the first full SSE frame.
• Helm chart updates: introduce configurable UI stream timeout, propagate env var, expose nginx proxy timeouts, and allow OpenShift Route timeout annotations.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
ui/src/components/chat/ChatInterface.tsx	Use runtime-configured stream inactivity timeout instead of a fixed 10-minute constant.
ui/src/app/actions/config.ts	Add server-side runtime config accessor reading timeout from env with a default fallback.
ui/src/app/a2a/[namespace]/[agentName]/route.ts	Reset keep-alive timer as soon as upstream bytes arrive and at stream start.
helm/kagent/values.yaml	Add UI stream timeout + nginx timeout defaults + OpenShift Route timeout annotations.
helm/kagent/templates/ui-deployment.yaml	Wire Helm value into KAGENT_STREAM_TIMEOUT_MS env var for the UI container.
helm/kagent/templates/openshift-route.yaml	Allow user-provided Route annotations for timeouts.
helm/kagent/files/nginx.conf	Parameterize proxy read/send timeouts via Helm values for streaming endpoints.

---

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

[peterj]

thanks for the PR! Can you please address the PR comments?

[yanivmn]

thanks for the PR! Can you please address the PR comments?

Addressed all issues

[yanivmn]

@peterj anything else required for this PR?

[EItanya]

Hey @yanivmn, sorry for letting this fall through, can you resolve conflicts and then we can push it through