[Gabriel Dufresne (GabrielDuf)]

• I have read the contributing guidelines, and I agree with the Code of Conduct.
• Have you checked that there aren't other open pull requests for the same changes?
• Have you tested that the committed code can be executed without errors?
• Have you confirmed that this issue is caused by UniGetUI itself, and not by the package manager or the package involved?

---

Why

cli-headless-e2e fails intermittently on the Windows jobs and passes on re-run, always with:

CLI command failed (1): package search --manager npm --query cowsay --max-results 20
{ "status": "error", "command": "", "message": "The operation was canceled." }

(or the same on package details --manager npm). That message is not npm's: it is the CLI's own client-side timeout. IpcClient.GetRequestTimeout gave every GET /packages/* request 30 s. In green Windows runs npm search takes 1–7 s and npm details 6–15 s; in the failing runs the daemon log shows the npm request starting and never finishing before the job ends. Every failure in the last 8 failed runs was a Windows npm search/details call. Users on a slow network hit the same limit through the CLI.

What changed (IpcClient.cs)

• Read-only package queries (GET /packages/*: search, details, versions, installed, updates) now get 2 minutes instead of 30 s. Package actions keep their 5 minutes.
• When the client's own deadline fires, it now throws a TimeoutException naming the request and the limit (UniGetUI did not answer GET …/packages/search within 120 seconds.) instead of surfacing the bare The operation was canceled., which read like a cancelled package operation. The CLI exit code stays 1.
• TimeoutException was added to IsConnectivityException, so the /status probe and registration liveness checks still treat a timeout as an unreachable endpoint.
• The response body read now also observes the deadline.

Notes for reviewers

• The last Linux cancellation (4 September, package repair --manager npm) is a different case: that POST already has 5 minutes and normally takes ~1 s. With this change a client timeout and a server-reported cancelled operation are now distinguishable in the output.
• There is no unit test for the timeout path: IpcClient can only be constructed against a registered daemon. The 49 existing IPC tests pass, and both dotnet format passes are clean.

🤖 Generated with Claude Code

[Copilot]

🟢 Approval recommended

The scoped changes preserve existing action deadlines and error-handling behavior, with no blocking issues identified.

No open issues 🎉

What changed in this PR

This PR gives CLI package queries more time to complete and makes client-side timeout errors easier to distinguish from cancelled package operations.

Changes:

• Extends read-only package query deadlines from 30 seconds to two minutes, preserving five-minute action deadlines.
• Reports expired deadlines with the request method, path, and timeout duration, and applies the deadline to response-body reads.
• Treats the new TimeoutException as a connectivity failure for endpoint liveness checks.

File	Description
src/​UniGetUI.Interface.IpcApi/​IpcClient.cs	Adjusts query deadlines and improves timeout reporting and classification.

_{Review effort: Balanced}

[randy-but-a-ro]

🤖 Pull request was approved automatically: the AI review is complete and all its review threads are resolved. 🎉

Integration Details

{
	"deliveryId": "47f99090-cee0-11f1-824a-f2ff1b8f13b4",
	"headSha": "2e9beef860b1b3c3cc22cea0ff11fa4762d84f21",
	"reviewer": "copilot-pull-request-reviewer[bot]"
}