[wezzcoetzee]

Closes #5948

What Changed

mergeUsage now derives a pricingStatus from the environments that actually contributed buckets, and UsageCoverageNotice renders one line when the model rate table could not be loaded:

Model rates could not be loaded, so 100.0% of responses are excluded from cost. Token counts are unaffected.

No contract change, no new UI surface, no change to any cost calculation. Two tests added in usageMerge.test.ts.

Why

When the LiteLLM rate table fails to fetch, UsageService.ensureRates swallows the failure and every bucket falls through lookupRate to { costUsd: 0, costSource: "unpriced" }. The usage page then renders a confident $0.00 for every model, 0.0% of cost for every provider, and $0.00 cache savings — beside perfectly correct token counts. It is indistinguishable from having spent nothing.

Degrading instead of failing the page is the right call, and this PR does not change it. The bug is that it degrades invisibly.

The server already ships the diagnosis: UsageSummary.pricing.status carries fresh | cached | unavailable, with a doc comment saying it exists "so the UI can be honest about how good the cost figures are", and every bucket carries unpricedRecords. Both cross the wire on every request and no client reads either — grep -rn "pricing" apps/web/src apps/mobile/src returns nothing. This plumbs the existing signal to the existing notice strip.

Two deliberate details:

• The magnitude comes from unpricedShare, not an assumption. Codex buckets can carry providerReported cost even with no rate table, so "everything reads $0.00" would be an overclaim. This also makes the message correct for the partial case: a genuinely new model missing from LiteLLM's table silently undercounts cost today, and gets the same message with a real percentage.
• cached gets no line. Stale rates move cost by rounding, not orders of magnitude; warning about it would be noise.

Reproduced on a machine behind TLS inspection, where the fetch to raw.githubusercontent.com fails. Diagnosing it required reading server.trace.ndjson and noticing usage-model-rates.json was absent from the state dir. That should not be the discovery path.

UI Changes

Before and after, same data, same viewport, same commit base — the only difference is the notice line.

before	after

Follow-up: #5949 tracks the underlying fragility — the rate table has no bundled fallback, so a fresh offline install has no prices at all. Deliberately not addressed here.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes
• I included a video for animation/interaction changes (n/a — static text, no motion)

Note

Show pricing unavailability notice on the usage page when model rates cannot be loaded

• Adds a pricingStatus field to MergedUsage in usageMerge.ts, computed as the weakest rate-table status (unavailable < cached < fresh) across environments that contributed usage buckets; idle environments are excluded.
• Updates UsageCoverageNotice in UsagePage.tsx to display an explanatory message when some responses are unpriced, distinguishing between rates being unavailable (failed to load) and models having no published rate.
• No notice is shown when there are no records or the unpriced share is zero.

^{Macroscope summarized 38d1f68.}

---

Note

Low Risk
UI-only messaging plus merge metadata derived from existing pricing fields; cost math and contracts are unchanged.

Overview
When model rates fail to load or some responses have no published price, the usage page can show $0 cost while token counts look correct. This PR surfaces that gap in the existing coverage notice strip—no new API fields and no change to how cost is calculated.

mergeUsage now exposes pricingStatus on merged totals: the weakest UsageSummary.pricing.status among environments that actually contributed buckets (unavailable < cached < fresh). Idle environments with failed rate tables do not drag down the merged status.

UsageCoverageNotice shows a line when there are records and unpricedShare > 0: either that rates could not be loaded, or that a percentage of responses lack a published rate (with token counts called out as unaffected where relevant). cached pricing does not get its own warning.

^{Reviewed by Cursor Bugbot for commit 38d1f68. Bugbot is set up for automated code reviews on this repo. Configure here.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 76f4b30e-2b9b-4f72-a0d6-837d4ccd7a37

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

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.}

[macroscopeapp]

🟡 Medium usage/UsagePage.tsx:441

When multiple environments contribute records, the pricingStatus === "unavailable" branch falsely reports that the rate table failed to load as the cause of unpriced responses. pricingStatus is a merged environment-level worst status while unpricedShare is aggregated independently, so one environment can be unavailable (contributing only priced records) while a fresh environment contributes unpriced records from an unknown model. In that case the message claims the rate table could not be loaded, but the excluded records actually came from the loaded table lacking a model rate. Consider preserving per-record provenance for unpriced responses, or using a non-causal generic message that does not attribute the exclusion to a rate-table failure for mixed environments.

🤖 Copy this AI Prompt to have your agent fix this:

In file @apps/web/src/components/usage/UsagePage.tsx around line 441:

When multiple environments contribute records, the `pricingStatus === "unavailable"` branch falsely reports that the rate table failed to load as the cause of unpriced responses. `pricingStatus` is a merged environment-level worst status while `unpricedShare` is aggregated independently, so one environment can be `unavailable` (contributing only priced records) while a `fresh` environment contributes unpriced records from an unknown model. In that case the message claims the rate table could not be loaded, but the excluded records actually came from the loaded table lacking a model rate. Consider preserving per-record provenance for unpriced responses, or using a non-causal generic message that does not attribute the exclusion to a rate-table failure for mixed environments.

[wezzcoetzee]

Good catch — fixed in c989e51.

You are right that the two values are aggregated independently: pricingStatus is the weakest status across contributing environments, while unpricedShare sums unpriced records regardless of which environment produced them. Across several devices, the environment whose rate table failed need not be the one that contributed the unpriced records, so asserting the failure caused the exclusion is wrong.

I kept both facts but dropped the causal link rather than adding per-record provenance:

Model rates could not be loaded. 100.0% of responses are excluded from cost. Token counts are unaffected.

Each sentence now stands on its own and is true in the mixed case. Per-record provenance would be the more precise fix, but it means a contract change to carry the reason per bucket, which is more than this PR should take on.

[macroscopeapp]

Sorry, I'm unable to act on this request because you do not have permissions within this repository.

[macroscopeapp]

Approvability

Verdict: Needs human review

1 blocking correctness issue found.

^{You can customize Macroscope's approvability policy. Learn more.}

[wezzcoetzee]

Fixed in #7181