proxy: preserve wall-clock duration in metrics

[SuperMarioYL]

Keep request duration from being underreported when upstream timings only cover part of the full request lifecycle.

• compare wall-clock and upstream timing durations
• keep token and throughput values from timings
• add regression coverage for underreported timings

fixes #602

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 9aa228bb-4efc-4d30-b4a4-def9c44fbfbd

📥 Commits

Reviewing files that changed from the base of the PR and between a37b486 and 27d0204.

📒 Files selected for processing (2)

• proxy/metrics_monitor.go
• proxy/metrics_monitor_test.go

---

Walkthrough

The PR modifies duration calculation in the metrics parser to compute wall-clock duration once and compare it against timing-derived duration, selecting the greater value instead of unconditionally using timing data when available.

Changes

Cohort / File(s)	Summary
Duration Calculation Logic proxy/metrics_monitor.go	Modified parseMetrics to compute wallDurationMs once using time.Since(start).Milliseconds(), then update durationMs only if the summed timingsDurationMs exceeds the wall-clock based value, preventing timing data from underreporting actual request duration.
Test Coverage proxy/metrics_monitor_test.go	Added subtest and github.com/tidwall/gjson import to validate parseMetrics behavior when timing-derived duration is lower than wall-clock time, asserting that metrics.DurationMs reflects the greater value.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Possibly related PRs

• proxy: support timings for /infill from llama-server #510: Modifies timing-derived duration parsing and selection logic in proxy/metrics_monitor.go, directly related to this PR's duration calculation changes.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately summarizes the main code change: preserving wall-clock duration in metrics by comparing it with upstream timing duration.
Description check	✅ Passed	The description clearly relates to the changeset, explaining the issue of underreported duration and the solution to compare and preserve wall-clock duration.
Linked Issues check	✅ Passed	The PR addresses issue #602 by computing wall-clock duration and using maximum of wall-clock vs upstream timing duration, ensuring Activity duration reflects actual request lifetime.
Out of Scope Changes check	✅ Passed	All changes are directly related to fixing the duration calculation issue: metrics parsing logic for wall-clock duration preservation and regression test coverage.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[mostlygeek]

Nice thanks for this fix!

[candrews]

@mostlygeek can you please make a release so us users can enjoy this and the other improvement made since the last release?

Thank you!