feat(pricing): meter cache writes and long context - #379
Merged
Conversation
WalkthroughAdds tiered pricing, one-hour cache-write rates and usage tracking, centralized metered cost calculation with billing breakdowns, expanded pricing and usage APIs, and dashboard editing, comparison, and reporting support. ChangesPricing contracts and persistence
Usage extraction and cost calculation
Usage analytics and validation
Dashboard pricing and usage experience
Estimated code review effort: 4 (Complex) | ~60 minutes Possibly related issues
Possibly related PRs
Suggested reviewers: 🚥 Pre-merge checks | ✅ 3 | ❌ 2❌ Failed checks (2 warnings)
✅ Passed checks (3 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
✨ Simplify code
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. Comment |
Track Anthropic 1-hour cache writes, preserve tiered context rates, and expose auditable pricing controls and usage breakdowns.\n\nRefs #376
Make cache and long-context pricing policy visible in the model table while preserving full selectors for editing and routing.
njbrake
force-pushed
the
feat/provider-metered-pricing
branch
from
July 22, 2026 20:27
14ab324 to
855f095
Compare
njbrake
had a problem deploying
to
integration-tests
July 22, 2026 20:27 — with 
GitHub Actions
Failure
GitHub Actions
Failure
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
🧹 Nitpick comments (2)
web/src/pages/ModelsPage.test.tsx (1)
647-759: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winMock
apiFetchrather than globalfetch.These new cases continue through
mockApi(), which replacesglobalThis.fetch. Mock theapiFetchboundary instead so dashboard tests stay aligned with the supported API contract and are less coupled to its transport implementation. As per coding guidelines, “mock only theapiFetchnetwork boundary, not hooks.”🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@web/src/pages/ModelsPage.test.tsx` around lines 647 - 759, Update the tests in ModelsPage.test.tsx to mock the apiFetch boundary instead of relying on mockApi() to replace globalThis.fetch. Adjust the relevant test setup and POST request assertions to use the supported apiFetch mock while preserving the existing pricing behavior and expectations; do not mock hooks or the underlying transport.Source: Coding guidelines
src/gateway/core/config.py (1)
78-87: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick winNice addition, but consider requiring at least one override per tier.
PricingTierConfighappily accepts a tier with onlymin_input_tokensset and no actual price overrides — a silent no-op tier that a config author might not notice is doing nothing. The dashboard'svalidTiers()inModelsPage.tsxalready enforces "at least one override" client-side; mirroring that here would catch the same mistake for anyone authoringconfig.ymldirectly.♻️ Optional validator
+ `@model_validator`(mode="after") + def validate_has_override(self) -> "PricingTierConfig": + fields = ( + self.input_price_per_million, + self.output_price_per_million, + self.cache_read_price_per_million, + self.cache_write_price_per_million, + self.cache_write_1h_price_per_million, + ) + if all(f is None for f in fields): + raise ValueError("pricing tier must override at least one price field") + return self🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@src/gateway/core/config.py` around lines 78 - 87, Update PricingTierConfig to validate that at least one price override field is provided, covering input, output, cache-read, cache-write, and cache-write-1h prices; reject tiers containing only min_input_tokens while preserving the existing individual field constraints.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src/gateway/services/pricing_service.py`:
- Line 247: Update the default_model_pricing() caveat in
src/gateway/services/pricing_service.py at lines 247-247 to state that default
pricing retains tier breakpoints instead of flattening to a base rate. Revise
the limitation in docs/configuration.md at lines 297-312 consistently, removing
or correcting the claim that imported tiers are billed at the base rate.
In `@src/gateway/static/dashboard/assets/ActivityPage-HAV6P_tZ.js`:
- Line 1: Update the source component that renders the expanded activity entry,
specifically the Pe component and its error_message display/copy behavior, so
raw provider errors are never rendered or passed to Te. Replace them with a
sanitized generic error message while retaining raw diagnostics server-side,
then rebuild the generated ActivityPage asset.
In `@web/src/pages/ModelsPage.tsx`:
- Around line 1101-1109: Update PricingPolicyCell to sort row.pricingTiers by
the raw tier.min_input_tokens values before applying formatContext, then build
thresholds from that ordered result. Preserve the existing labels and tier-count
display while ensuring thresholds spanning K/M boundaries appear in numeric
ascending order.
---
Nitpick comments:
In `@src/gateway/core/config.py`:
- Around line 78-87: Update PricingTierConfig to validate that at least one
price override field is provided, covering input, output, cache-read,
cache-write, and cache-write-1h prices; reject tiers containing only
min_input_tokens while preserving the existing individual field constraints.
In `@web/src/pages/ModelsPage.test.tsx`:
- Around line 647-759: Update the tests in ModelsPage.test.tsx to mock the
apiFetch boundary instead of relying on mockApi() to replace globalThis.fetch.
Adjust the relevant test setup and POST request assertions to use the supported
apiFetch mock while preserving the existing pricing behavior and expectations;
do not mock hooks or the underlying transport.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: ae71560e-81cb-4de9-b286-59d24979324c
⛔ Files ignored due to path filters (1)
docs/public/openapi.jsonis excluded by!docs/public/openapi.json
📒 Files selected for processing (43)
alembic/versions/7d9e1f3a5b7c_add_metered_pricing_fields.pydocs/configuration.mdsrc/gateway/api/routes/_pipeline.pysrc/gateway/api/routes/messages.pysrc/gateway/api/routes/models.pysrc/gateway/api/routes/pricing.pysrc/gateway/api/routes/usage.pysrc/gateway/core/config.pysrc/gateway/core/usage.pysrc/gateway/models/entities.pysrc/gateway/services/metered_pricing.pysrc/gateway/services/pricing_init_service.pysrc/gateway/services/pricing_service.pysrc/gateway/static/dashboard/assets/ActivityPage-Dv1t5m0G.jssrc/gateway/static/dashboard/assets/ActivityPage-HAV6P_tZ.jssrc/gateway/static/dashboard/assets/AliasesPage-BWvbF6Ai.jssrc/gateway/static/dashboard/assets/BudgetsPage-CgMXKHu3.jssrc/gateway/static/dashboard/assets/KeysPage-DyFw8uY7.jssrc/gateway/static/dashboard/assets/ModelComboBox-CF4wm-YK.jssrc/gateway/static/dashboard/assets/ModelScopeControl-_-_rnmkv.jssrc/gateway/static/dashboard/assets/ModelsPage-Bp26JK8Z.jssrc/gateway/static/dashboard/assets/ModelsPage-CCg9xDDB.jssrc/gateway/static/dashboard/assets/OverviewPage-BbByzEFd.jssrc/gateway/static/dashboard/assets/ProvidersPage-Cls9MgL7.jssrc/gateway/static/dashboard/assets/SettingsPage-D6ZrjyKX.jssrc/gateway/static/dashboard/assets/ToolsGuardrailsPage-D9fLjk2y.jssrc/gateway/static/dashboard/assets/UsagePage-BiFDzqgv.jssrc/gateway/static/dashboard/assets/UsagePage-Cf1iUtft.jssrc/gateway/static/dashboard/assets/UsersPage-Bq_wHrmP.jssrc/gateway/static/dashboard/assets/index-CSaFgccK.jssrc/gateway/static/dashboard/assets/index-DtlrpKAk.csssrc/gateway/static/dashboard/index.htmlsrc/gateway/streaming.pytests/integration/test_pricing_budget_validation.pytests/integration/test_pricing_config.pytests/unit/test_compute_cost.pytests/unit/test_default_model_pricing.pytests/unit/test_usage_cache_tokens.pyweb/src/api/types.tsweb/src/pages/ActivityPage.tsxweb/src/pages/ModelsPage.test.tsxweb/src/pages/ModelsPage.tsxweb/src/pages/UsagePage.tsx
💤 Files with no reviewable changes (3)
- src/gateway/static/dashboard/assets/UsagePage-Cf1iUtft.js
- src/gateway/static/dashboard/assets/ModelsPage-CCg9xDDB.js
- src/gateway/static/dashboard/assets/ActivityPage-Dv1t5m0G.js
| @@ -0,0 +1 @@ | |||
| import{j as e}from"./tanstack-query-W9y7rsMr.js";import{u as re,r as n}from"./react-q-ooZ0ti.js";import{u as le,a as ne,b as ie,c as oe,P as ce,E as de,F as G,d as Y}from"./index-CSaFgccK.js";import{T as ue,a as me,b as d,L as xe,c as he,d as pe,e as u}from"./Table-DEsIhjZo.js";import{B as j,S as ge}from"./heroui-CewI8xK4.js";const je=new Intl.NumberFormat(void 0,{style:"currency",currency:"USD",maximumFractionDigits:4});function _(s){return s===null?"—":je.format(s)}function o(s){return s===null?"—":s.toLocaleString()}function Z(s){return s===null?"—":s<1e3?`${s} ms`:`${(s/1e3).toFixed(s<1e4?2:1)} s`}function fe(s){const a=new Date(s);return Number.isNaN(a.getTime())?s:a.toLocaleString()}function ve(s){const a=new Date(s).getTime();if(Number.isNaN(a))return s;const r=Math.max(0,Math.round((Date.now()-a)/1e3));if(r<60)return`${r}s ago`;const i=Math.round(r/60);if(i<60)return`${i}m ago`;const m=Math.round(i/60);return m<24?`${m}h ago`:`${Math.round(m/24)}d ago`}const be=3600,w=86400,Se=[{label:"Last hour",seconds:be},{label:"24h",seconds:w},{label:"7d",seconds:7*w},{label:"All",seconds:null}],Ne=[{label:"All",value:""},{label:"Success",value:"success"},{label:"Error",value:"error"}],_e=["/v1/chat/completions","/v1/messages","/v1/responses","/v1/embeddings","/v1/moderations","/v1/audio/transcriptions","/v1/audio/speech","/v1/images/generations","/v1/rerank","/v1/batches","/v1/batches/results"],S=50;function E(s){return new Date(Date.now()-s*1e3).toISOString()}function we({status:s}){const a=s==="error"?"border-red-200 bg-red-50 text-red-700":"border-[var(--otari-line)] bg-[var(--otari-brand-tint)] text-[var(--otari-brand-dark)]";return e.jsx("span",{className:`inline-flex items-center rounded-full border px-2 py-0.5 text-xs font-medium ${a}`,children:s})}async function ke(s,a=navigator.clipboard){if(!a)return!1;try{return await a.writeText(s),!0}catch{return!1}}function Te({text:s}){const[a,r]=n.useState(!1);return e.jsx(j,{size:"sm",variant:"ghost",onPress:async()=>{await ke(s)&&(r(!0),window.setTimeout(()=>r(!1),1500))},children:a?"Copied":"Copy"})}function l({label:s,children:a}){return e.jsxs("div",{className:"flex flex-col gap-0.5",children:[e.jsx("span",{className:"text-[11px] font-medium uppercase tracking-wide text-[var(--otari-muted)]",children:s}),e.jsx("span",{className:"text-sm text-[var(--otari-ink)] break-all",children:a})]})}function Pe({entry:s}){var a;return e.jsxs("div",{className:"flex flex-col gap-4 px-4 py-4",children:[s.error_message?e.jsxs("div",{className:"flex flex-col gap-1.5",children:[e.jsxs("div",{className:"flex items-center justify-between",children:[e.jsx("span",{className:"text-[11px] font-medium uppercase tracking-wide text-[var(--otari-muted)]",children:"Error (provider-reported)"}),e.jsx(Te,{text:s.error_message})]}),e.jsx("pre",{className:"max-h-48 overflow-auto rounded-lg border border-red-200 bg-red-50 p-3 text-xs whitespace-pre-wrap break-all text-red-700",children:s.error_message})]}):null,e.jsxs("div",{className:"grid gap-4 sm:grid-cols-2 lg:grid-cols-4",children:[e.jsx(l,{label:"Provider",children:s.provider??"—"}),e.jsx(l,{label:"Endpoint",children:s.endpoint}),e.jsx(l,{label:"User",children:s.user_id??"—"}),e.jsx(l,{label:"API key",children:s.api_key_id??"—"}),e.jsx(l,{label:"Prompt tokens",children:o(s.prompt_tokens)}),e.jsx(l,{label:"Completion tokens",children:o(s.completion_tokens)}),e.jsx(l,{label:"Total tokens",children:o(s.total_tokens)}),e.jsx(l,{label:"Cost",children:_(s.cost)}),e.jsx(l,{label:"Cache read tokens",children:o(s.cache_read_tokens)}),e.jsx(l,{label:"Cache write tokens",children:o(s.cache_write_tokens)}),e.jsx(l,{label:"1h cache writes",children:o(s.cache_write_1h_tokens??null)}),e.jsx(l,{label:"Total time",children:Z(s.latency_ms)}),e.jsx(l,{label:"Request ID",children:s.id})]}),(a=s.pricing_breakdown)!=null&&a.length?e.jsxs("div",{className:"flex flex-col gap-2",children:[e.jsx("span",{className:"text-[11px] font-medium uppercase tracking-wide text-[var(--otari-muted)]",children:"Billed meters"}),e.jsx("div",{className:"grid gap-2 sm:grid-cols-2 lg:grid-cols-3",children:s.pricing_breakdown.map(r=>e.jsxs(l,{label:r.meter.replaceAll("_"," "),children:[o(r.units)," at ",_(r.rate_per_million)," / 1M, ",_(r.cost)]},r.meter))})]}):null]})}const D=7;function $e(){var q,z,y;const s=le(),[a]=re(),r=a.get("start_date")??void 0,[i,m]=n.useState(r?null:w),[N,k]=n.useState(()=>r??E(w)),[x,A]=n.useState(()=>a.get("status")??""),[c,M]=n.useState(()=>a.get("model")??""),[h,$]=n.useState(""),[p,U]=n.useState(()=>a.get("user_id")??""),[f,T]=n.useState(0),[J,K]=n.useState(null),P=n.useMemo(()=>({start_date:N,status:x||void 0,model:c.trim()||void 0,endpoint:h||void 0,user_id:p||void 0}),[N,x,c,h,p]);n.useEffect(()=>{T(0)},[P]);const g=ne(P,f,S),v=ie(P),Q=n.useMemo(()=>({start_date:N,status:x||void 0,endpoint:h||void 0,user_id:p||void 0}),[N,x,h,p]),F=((z=(q=oe(Q,"day").data)==null?void 0:q.by_model)==null?void 0:z.filter(t=>!t.is_other&&t.key!==null).map(t=>t.key))??[],b=g.data??[],C=((y=v.data)==null?void 0:y.total)??0,L=!!(x||c.trim()||h||p||i!==null),V=t=>{m(t),k(t===null?void 0:E(t))},W=()=>{m(null),k(void 0),A(""),M(""),$(""),U("")},X=()=>{i!==null&&k(E(i)),g.refetch(),v.refetch()},I=b.length>0,O=I?f*S+1:0,R=f*S+b.length,B=v.isSuccess&&!v.isPlaceholderData,ee=B?C===0?"0 of 0":`${O}–${R} of ${C.toLocaleString()}`:I?`${O}–${R}`:"0",se=B?(f+1)*S<C:b.length===S;return e.jsxs("div",{className:"flex flex-col gap-6",children:[e.jsx(ce,{title:"Activity",description:"A per-request log of what the gateway served: tokens, cost, latency, and failures. No request or response content is stored.",action:e.jsx(j,{variant:"outline",onPress:X,isDisabled:g.isFetching,children:"Refresh"})}),e.jsx(de,{error:g.error??v.error}),e.jsxs("div",{className:"flex flex-wrap items-end gap-3",children:[e.jsx("div",{className:"flex flex-wrap gap-2",children:Se.map(t=>e.jsx(j,{size:"sm",variant:i===t.seconds?"primary":"outline",onPress:()=>V(t.seconds),children:t.label},t.label))}),e.jsxs("div",{className:"flex flex-wrap items-end gap-3",children:[e.jsx(G,{id:"filter-status",label:"Status",value:x,onChange:A,children:Ne.map(t=>e.jsx("option",{value:t.value,children:t.label},t.value))}),e.jsxs(G,{id:"filter-endpoint",label:"Endpoint",value:h,onChange:$,children:[e.jsx("option",{value:"",children:"All endpoints"}),_e.map(t=>e.jsx("option",{value:t,children:t},t))]}),e.jsx(Y,{label:"User",value:p,onChange:U,placeholder:"All users",options:(s.data??[]).map(t=>({value:t.user_id,label:t.alias?`${t.alias} (${t.user_id})`:t.user_id}))}),e.jsx(Y,{label:"Model",value:c,onChange:M,allowsCustom:!0,placeholder:"Any model",options:(c&&!F.includes(c)?[c,...F]:F).map(t=>({value:t,label:t}))}),L?e.jsx(j,{size:"sm",variant:"ghost",onPress:W,children:"Clear filters"}):null]})]}),e.jsxs(ue,{children:[e.jsx(me,{children:e.jsxs("tr",{children:[e.jsx(d,{children:"Time"}),e.jsx(d,{children:"User"}),e.jsx(d,{children:"Model"}),e.jsx(d,{className:"text-right",children:"Tokens"}),e.jsx(d,{className:"text-right",children:"Cost"}),e.jsx(d,{className:"text-right",children:"Total time"}),e.jsx(d,{children:"Status"})]})}),e.jsx("tbody",{children:g.isLoading?e.jsx(xe,{colSpan:D}):b.length===0?e.jsx(he,{colSpan:D,children:L?"No requests match these filters.":"No requests recorded yet."}):b.map(t=>{const te=t.status==="error",H=J===t.id;return e.jsxs(n.Fragment,{children:[e.jsxs(pe,{selected:H,className:te?"bg-red-50":"",onClick:()=>K(ae=>ae===t.id?null:t.id),children:[e.jsx(u,{className:"text-[var(--otari-muted)]",children:e.jsx("span",{title:fe(t.timestamp),children:ve(t.timestamp)})}),e.jsx(u,{className:"text-[var(--otari-ink)]",children:t.user_id??"—"}),e.jsx(u,{className:"text-[var(--otari-ink)]",children:t.model}),e.jsx(u,{className:"text-right tabular-nums",children:o(t.total_tokens)}),e.jsx(u,{className:"text-right tabular-nums",children:_(t.cost)}),e.jsx(u,{className:"text-right tabular-nums",children:Z(t.latency_ms)}),e.jsx(u,{children:e.jsx(we,{status:t.status})})]}),H?e.jsx("tr",{className:"border-b border-[var(--otari-line)] bg-[var(--otari-bg)]",children:e.jsx("td",{colSpan:D,children:e.jsx(Pe,{entry:t})})}):null]},t.id)})})]}),e.jsxs("div",{className:"flex items-center justify-between gap-3",children:[e.jsxs("span",{className:"inline-flex items-center gap-2 text-sm text-[var(--otari-muted)]",children:[ee,g.isFetching?e.jsx(ge,{size:"sm"}):null]}),e.jsxs("span",{className:"inline-flex gap-1.5",children:[e.jsx(j,{size:"sm",variant:"outline",isDisabled:f===0,onPress:()=>T(t=>Math.max(0,t-1)),children:"Previous"}),e.jsx(j,{size:"sm",variant:"outline",isDisabled:!se,onPress:()=>T(t=>t+1),children:"Next"})]})]})]})}export{$e as ActivityPage,ke as copyToClipboard}; | |||
There was a problem hiding this comment.
🔒 Security & Privacy | 🟠 Major | 🏗️ Heavy lift
Do not render raw provider errors in the dashboard.
error_message is displayed and copied verbatim. Provider responses can include sensitive request details or credentials; expose only a sanitized generic message and retain raw diagnostics server-side. Fix the source component and rebuild this generated asset.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/gateway/static/dashboard/assets/ActivityPage-HAV6P_tZ.js` at line 1,
Update the source component that renders the expanded activity entry,
specifically the Pe component and its error_message display/copy behavior, so
raw provider errors are never rendered or passed to Te. Replace them with a
sanitized generic error message while retaining raw diagnostics server-side,
then rebuild the generated ActivityPage asset.
Source: Coding guidelines
njbrake
had a problem deploying
to
integration-tests
July 22, 2026 20:38 — with
GitHub Actions
Failure
GitHub Actions
Failure
Member
Author
|
Addressed the current CodeRabbit findings in 156e798:
Also updated the five stale exact-response integration assertions that caused the previous CI failure. Local validation: 870 unit tests passed, dashboard tests passed, lint, type checks, OpenAPI freshness, and dashboard build passed. CodeRabbit is currently rate-limited for a follow-up review. |
There was a problem hiding this comment.
Pull request overview
Adds provider-metered pricing support for cache reads/writes (including Anthropic 1-hour TTL) and whole-request long-context price tiers, persisting billing meters and a per-request pricing breakdown for auditing. The dashboard is updated to surface these new meters and pricing policies.
Changes:
- Introduces tier-aware, meter-based cost calculation and persists
billing_meters+pricing_breakdownonUsageLog. - Extends pricing/config/models/APIs to support
cache_write_1h_price_per_millionandpricing_tiers. - Updates the dashboard UI + tests to display/edit tier + TTL pricing and show 1-hour cache-write usage.
Reviewed changes
Copilot reviewed 46 out of 50 changed files in this pull request and generated 2 comments.
Show a summary per file
| File | Description |
|---|---|
| web/src/pages/UsagePage.tsx | Adds a “1h cache write” stat tile to usage analytics. |
| web/src/pages/ModelsPage.test.tsx | Updates mocks and expands coverage for TTL + tier pricing UI behavior. |
| web/src/pages/ActivityPage.tsx | Avoids exposing provider diagnostics; displays 1h cache writes and pricing breakdown. |
| web/src/pages/ActivityPage.test.tsx | Updates assertions to ensure provider error details are not exposed. |
| web/src/api/types.ts | Extends API types for 1h cache write rates/tokens, billing meters, tier rules, and breakdown lines. |
| tests/unit/test_usage_cache_tokens.py | Adds coverage for TTL detection and Anthropic 1h cache-write usage extraction. |
| tests/unit/test_default_model_pricing.py | Ensures genai-prices tier data is preserved into pricing tiers. |
| tests/unit/test_compute_cost.py | Adds tests for tier “whole-request cliff” semantics and 1h cache-write pricing. |
| tests/unit/test_budget_cost_estimate.py | Adds unit tests for threshold-aware reservation estimates (including cache TTL). |
| tests/integration/test_usage_summary.py | Extends usage totals response coverage to include 1h cache write totals. |
| tests/integration/test_usage_endpoint.py | Extends usage row response shape coverage for new billing/audit fields. |
| tests/integration/test_pricing_config.py | Adds integration coverage for persisting tiers + 1h cache write pricing via API. |
| tests/integration/test_pricing_budget_validation.py | Adds integration validation coverage for tier rules. |
| tests/integration/test_model_aliases.py | Updates alias pricing surface expectations to include new pricing fields. |
| tests/integration/test_alias_api.py | Updates alias API expectations to include new pricing fields. |
| src/gateway/streaming.py | Merges streamed usage including the new 1h cache-write token meter. |
| src/gateway/services/pricing_service.py | Preserves genai-prices tier thresholds and exports them as pricing tiers. |
| src/gateway/services/pricing_init_service.py | Initializes DB pricing from config including 1h cache-write rate and tiers. |
| src/gateway/services/metered_pricing.py | Adds canonical meters, tier rate selection, reservation estimation, and auditable charge lines. |
| src/gateway/services/budget_service.py | Switches budget estimation to the meter-based estimator and adds cache TTL input. |
| src/gateway/models/entities.py | Adds DB fields for tier rules and usage auditability (meters + breakdown). |
| src/gateway/core/usage.py | Adds cache_write_1h_tokens to gateway usage and accessor helpers. |
| src/gateway/core/config.py | Adds config models/validation for tier rules + 1h cache-write pricing. |
| src/gateway/api/routes/usage.py | Extends usage API models and totals aggregation for new fields. |
| src/gateway/api/routes/pricing.py | Extends pricing API schemas/storage for tiers and 1h cache-write rates. |
| src/gateway/api/routes/models.py | Surfaces new pricing fields on model listing/details and alias pricing views. |
| src/gateway/api/routes/messages.py | Captures Anthropic 1h cache-write usage and detects requested cache TTL for estimation. |
| src/gateway/api/routes/_pipeline.py | Switches to metered calculator, persists breakdown, and threads TTL into reservations. |
| docs/public/openapi.json | Regenerates OpenAPI to include new pricing/usage fields. |
| docs/configuration.md | Documents 1h cache-write pricing and long-context tier configuration semantics. |
| alembic/versions/7d9e1f3a5b7c_add_metered_pricing_fields.py | Adds schema migrations for new pricing and usage audit columns. |
| src/gateway/static/dashboard/index.html | Updates built dashboard asset references. |
| src/gateway/static/dashboard/assets/UsersPage-BUuslhzL.js | Updated built dashboard bundle output. |
| src/gateway/static/dashboard/assets/UsagePage-ChSRIieH.js | Updated built dashboard bundle output (new usage UI). |
| src/gateway/static/dashboard/assets/UsagePage-CFdEpDOA.js | Removes old built dashboard chunk. |
| src/gateway/static/dashboard/assets/ToolsGuardrailsPage-B8ikQfKj.js | Updated built dashboard bundle output. |
| src/gateway/static/dashboard/assets/OverviewPage-BOBh0Ajn.js | Updated built dashboard bundle output. |
| src/gateway/static/dashboard/assets/ModelsPage-Cgm8YsX7.js | Removes old built dashboard chunk. |
| src/gateway/static/dashboard/assets/ModelScopeControl-BpXA5VMI.js | Updated built dashboard bundle output. |
| src/gateway/static/dashboard/assets/KeysPage-CvexGYae.js | Updated built dashboard bundle output. |
| src/gateway/static/dashboard/assets/AliasesPage-BSp3jVNj.js | Updated built dashboard bundle output. |
| src/gateway/static/dashboard/assets/ActivityPage-DYwWFsLV.js | Updated built dashboard bundle output (new activity UI). |
| src/gateway/static/dashboard/assets/ActivityPage-CKgTCz2o.js | Removes old built dashboard chunk. |
Files not reviewed (3)
- src/gateway/static/dashboard/assets/ActivityPage-DYwWFsLV.js: Generated file
- src/gateway/static/dashboard/assets/ModelsPage-5SIMv0Ug.js: Generated file
- src/gateway/static/dashboard/assets/UsagePage-ChSRIieH.js: Generated file
Comment on lines
+28
to
+37
| class PricingTier(BaseModel): | ||
| """Whole-request price cliff selected by total billable input tokens.""" | ||
|
|
||
| min_input_tokens: int = Field(gt=0) | ||
| input_price_per_million: float | None = Field(default=None, ge=0) | ||
| output_price_per_million: float | None = Field(default=None, ge=0) | ||
| cache_read_price_per_million: float | None = Field(default=None, ge=0) | ||
| cache_write_price_per_million: float | None = Field(default=None, ge=0) | ||
| cache_write_1h_price_per_million: float | None = Field(default=None, ge=0) | ||
|
|
Member
Author
There was a problem hiding this comment.
Fixed in fc10f6c, which adds the matching per-tier override validator to the pricing API and regression coverage.
Comment on lines
+93
to
+112
| input_tokens = max(estimated_input_tokens, 0.0) | ||
| output_tokens = max(estimated_output_tokens, 0) | ||
| cache_write_tokens = input_tokens if cache_write_ttl is not None else 0.0 | ||
| rates = effective_rates(pricing, input_tokens + cache_write_tokens) | ||
|
|
||
| input_rate = rates["input_price_per_million"] | ||
| output_rate = rates["output_price_per_million"] | ||
| assert input_rate is not None | ||
| assert output_rate is not None | ||
| cost = (input_tokens * input_rate + output_tokens * output_rate) / 1_000_000 | ||
|
|
||
| if cache_write_ttl == "1h": | ||
| cache_write_rate = rates["cache_write_1h_price_per_million"] | ||
| if cache_write_rate is None: | ||
| cache_write_rate = rates["cache_write_price_per_million"] | ||
| else: | ||
| cache_write_rate = rates["cache_write_price_per_million"] | ||
| if cache_write_rate is not None: | ||
| cost += cache_write_tokens * cache_write_rate / 1_000_000 | ||
| return cost |
Member
Author
There was a problem hiding this comment.
Fixed in e1c2d4c. Conservative reservations now price potential cache-write tokens at the selected normal input rate when no cache-write rate is configured, matching final reconciliation. The regression test now asserts the doubled input charge.
khaledosman
approved these changes
Jul 23, 2026
khaledosman
left a comment
khaledosman
left a comment
Member
There was a problem hiding this comment.
Reviewed the full diff against the backend/security/performance and dashboard standards. Billing path is sound: fresh_input stays non-negative in both the additive (Anthropic) and subset (OpenAI) conventions, the 1h-cache-write rate fallback is correct, a configured $0 rate is preserved, and the reserve/reconcile lifecycle is unchanged. Migration is reversible with a server_default, OpenAPI + dashboard bundle regenerated, docs updated, and the unit/integration coverage hits the tricky cases (whole-request cliff, 1h vs 5m rate, $0 rate, conservative over-reservation).
Approving. Three low-severity notes inline, none blocking.
Review created by Claude Code.
| output_price_per_million: float | None = Field(default=None, ge=0) | ||
| cache_read_price_per_million: float | None = Field(default=None, ge=0) | ||
| cache_write_price_per_million: float | None = Field(default=None, ge=0) | ||
| cache_write_1h_price_per_million: float | None = Field(default=None, ge=0) |
Member
There was a problem hiding this comment.
This API tier model lacks the "at least one rate override" check that PricingTierConfig enforces in core/config.py (validate_has_rate_override). Via POST /v1/pricing a tier of just {min_input_tokens} validates and is stored as a silent no-op at billing time. Add the matching model_validator for parity so config and API reject the same input. (Master-key only, so not a security issue.)
Member
Author
There was a problem hiding this comment.
Fixed in fc10f6c. The API now rejects no-op tiers with the same validation as config, and unit plus route coverage cover the rule.
| </div> | ||
| <pre className="max-h-48 overflow-auto rounded-lg border border-red-200 bg-red-50 p-3 text-xs whitespace-pre-wrap break-all text-red-700"> | ||
| {entry.error_message} | ||
| The provider returned an error. Inspect gateway logs for details. |
Member
There was a problem hiding this comment.
The provider-error redaction (dropping CopyButton + the verbatim error_message) is unrelated to pricing; CONTRIBUTING asks to keep diffs focused, so this reads more naturally as its own PR. Also note error_message is still returned by the /v1/usage API, so this hides it in the UI only, not at the boundary. Reasonable hardening either way.
Member
Author
There was a problem hiding this comment.
Kept this focused UI hardening because it resolves the CodeRabbit security finding: the dashboard no longer renders or copies raw provider diagnostics. Agreed that the master-key usage API remains a separate operator/API-boundary policy, so I did not broaden that change here.
| user_id_from_request=str(user_from_metadata) if user_from_metadata else None, | ||
| estimate_prompt_chars=len(str(request.messages)) + len(str(request.system or "")), | ||
| estimate_max_output_tokens=request.max_tokens, | ||
| estimate_cache_write_ttl=_requested_cache_write_ttl( |
Member
There was a problem hiding this comment.
estimate_cache_write_ttl is wired only into the Anthropic /v1/messages path. Anthropic prompt caching driven through /v1/chat/completions won't be reflected in the up-front reservation, so those requests under-reserve (final spend still reconciles correctly). Consistent with prior behavior; flagging only as a known asymmetry.
Member
Author
There was a problem hiding this comment.
Verified before changing: ChatCompletionRequest has no cache_control field and its default Pydantic config drops unknown extras, so the OpenAI-compatible endpoint cannot request Anthropic prompt-cache creation. Native Anthropic cache controls are accepted only on /v1/messages, where the conservative reservation is applied.
njbrake
temporarily deployed
to
integration-tests
July 23, 2026 09:23 — with
GitHub Actions
Inactive
GitHub Actions
Inactive
njbrake
temporarily deployed
to
integration-tests
July 23, 2026 09:24 — with
GitHub Actions
Inactive
GitHub Actions
Inactive
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (1)
src/gateway/static/dashboard/assets/ModelsPage-5SIMv0Ug.js (1)
1-1: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winUse dashboard design tokens instead of raw colors.
The tooltip uses
border-[#c2843a]andtext-[#b45309]. Replace these with--otari-*tokens in the source stylesheet, then rebuild the generated bundle. As per coding guidelines, dashboard colors must use--otari-*design tokens rather than scattered literals.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@src/gateway/static/dashboard/assets/ModelsPage-5SIMv0Ug.js` at line 1, The tooltip component se uses raw warning colors in its className. Replace border-[`#c2843a`] and text-[`#b45309`] with the appropriate --otari-* dashboard design tokens in the source stylesheet, then rebuild the generated ModelsPage bundle so the compiled asset reflects the token-based classes.Source: Coding guidelines
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src/gateway/static/dashboard/assets/ActivityPage-DYwWFsLV.js`:
- Line 1: Update the shared dashboard table-row interaction pattern so rows with
mouse onClick handlers also support keyboard activation via appropriate
focusability and keyboard semantics, preserving existing click behavior. Apply
the fix to request-detail expansion in
src/gateway/static/dashboard/assets/ActivityPage-DYwWFsLV.js:1-1, model-detail
selection in src/gateway/static/dashboard/assets/ModelsPage-5SIMv0Ug.js:1-1, and
usage drill-down navigation in
src/gateway/static/dashboard/assets/UsagePage-ChSRIieH.js:1-1, then regenerate
all affected bundles.
In `@src/gateway/static/dashboard/assets/ModelsPage-5SIMv0Ug.js`:
- Line 1: The model-row selection behavior in Kt currently relies only on
onClick and is inaccessible from the keyboard. Update the row selection
implementation to use a focusable interactive control or add Enter/Space
handling with appropriate focusability and selected-row semantics, while
preserving the existing onSelect behavior that opens model details.
---
Nitpick comments:
In `@src/gateway/static/dashboard/assets/ModelsPage-5SIMv0Ug.js`:
- Line 1: The tooltip component se uses raw warning colors in its className.
Replace border-[`#c2843a`] and text-[`#b45309`] with the appropriate --otari-*
dashboard design tokens in the source stylesheet, then rebuild the generated
ModelsPage bundle so the compiled asset reflects the token-based classes.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 67be0a8b-fb3a-4eef-8cd7-c211d7d0be9f
⛔ Files ignored due to path filters (1)
docs/public/openapi.jsonis excluded by!docs/public/openapi.json
📒 Files selected for processing (47)
alembic/versions/7d9e1f3a5b7c_add_metered_pricing_fields.pydocs/configuration.mdsrc/gateway/api/routes/_pipeline.pysrc/gateway/api/routes/messages.pysrc/gateway/api/routes/models.pysrc/gateway/api/routes/pricing.pysrc/gateway/api/routes/usage.pysrc/gateway/core/config.pysrc/gateway/core/usage.pysrc/gateway/models/entities.pysrc/gateway/services/budget_service.pysrc/gateway/services/metered_pricing.pysrc/gateway/services/pricing_init_service.pysrc/gateway/services/pricing_service.pysrc/gateway/static/dashboard/assets/ActivityPage-DYwWFsLV.jssrc/gateway/static/dashboard/assets/AliasesPage-BSp3jVNj.jssrc/gateway/static/dashboard/assets/BudgetsPage-DlL0354P.jssrc/gateway/static/dashboard/assets/KeysPage-CvexGYae.jssrc/gateway/static/dashboard/assets/ModelScopeControl-BpXA5VMI.jssrc/gateway/static/dashboard/assets/ModelsPage-5SIMv0Ug.jssrc/gateway/static/dashboard/assets/OverviewPage-BOBh0Ajn.jssrc/gateway/static/dashboard/assets/ProvidersPage-DPfGIa86.jssrc/gateway/static/dashboard/assets/SettingsPage-C0JYm9ni.jssrc/gateway/static/dashboard/assets/ToolsGuardrailsPage-B8ikQfKj.jssrc/gateway/static/dashboard/assets/UsagePage-ChSRIieH.jssrc/gateway/static/dashboard/assets/UsersPage-BUuslhzL.jssrc/gateway/static/dashboard/assets/index-DkHDQ_jO.jssrc/gateway/static/dashboard/assets/index-Dl_VAUga.csssrc/gateway/static/dashboard/index.htmlsrc/gateway/streaming.pytests/integration/test_alias_api.pytests/integration/test_model_aliases.pytests/integration/test_pricing_budget_validation.pytests/integration/test_pricing_config.pytests/integration/test_usage_endpoint.pytests/integration/test_usage_summary.pytests/unit/test_budget_cost_estimate.pytests/unit/test_compute_cost.pytests/unit/test_default_model_pricing.pytests/unit/test_pricing_schema.pytests/unit/test_usage_cache_tokens.pyweb/src/api/types.tsweb/src/pages/ActivityPage.test.tsxweb/src/pages/ActivityPage.tsxweb/src/pages/ModelsPage.test.tsxweb/src/pages/ModelsPage.tsxweb/src/pages/UsagePage.tsx
🚧 Files skipped from review as they are similar to previous changes (18)
- web/src/pages/UsagePage.tsx
- tests/unit/test_default_model_pricing.py
- src/gateway/static/dashboard/index.html
- src/gateway/services/pricing_init_service.py
- tests/integration/test_pricing_budget_validation.py
- src/gateway/services/pricing_service.py
- src/gateway/streaming.py
- src/gateway/core/config.py
- src/gateway/api/routes/usage.py
- alembic/versions/7d9e1f3a5b7c_add_metered_pricing_fields.py
- src/gateway/core/usage.py
- src/gateway/api/routes/messages.py
- web/src/pages/ModelsPage.test.tsx
- web/src/api/types.ts
- src/gateway/models/entities.py
- src/gateway/api/routes/pricing.py
- src/gateway/api/routes/_pipeline.py
- web/src/pages/ModelsPage.tsx
| @@ -0,0 +1 @@ | |||
| import{j as e}from"./tanstack-query-W9y7rsMr.js";import{u as le,r as n}from"./react-q-ooZ0ti.js";import{u as re,a as ne,b as ie,c as oe,P as ce,E as de,F as G,d as Y}from"./index-DkHDQ_jO.js";import{T as ue,a as me,b as d,L as xe,c as he,d as pe,e as u}from"./Table-DEsIhjZo.js";import{B as f,S as ge}from"./heroui-CewI8xK4.js";const je=new Intl.NumberFormat(void 0,{style:"currency",currency:"USD",maximumFractionDigits:4});function _(t){return t===null?"—":je.format(t)}function o(t){return t===null?"—":t.toLocaleString()}function Z(t){return t===null?"—":t<1e3?`${t} ms`:`${(t/1e3).toFixed(t<1e4?2:1)} s`}function ve(t){const a=new Date(t);return Number.isNaN(a.getTime())?t:a.toLocaleString()}function be(t){const a=new Date(t).getTime();if(Number.isNaN(a))return t;const r=Math.max(0,Math.round((Date.now()-a)/1e3));if(r<60)return`${r}s ago`;const i=Math.round(r/60);if(i<60)return`${i}m ago`;const m=Math.round(i/60);return m<24?`${m}h ago`:`${Math.round(m/24)}d ago`}const fe=3600,k=86400,Ne=[{label:"Last hour",seconds:fe},{label:"24h",seconds:k},{label:"7d",seconds:7*k},{label:"All",seconds:null}],Se=[{label:"All",value:""},{label:"Success",value:"success"},{label:"Error",value:"error"}],_e=["/v1/chat/completions","/v1/messages","/v1/responses","/v1/embeddings","/v1/moderations","/v1/audio/transcriptions","/v1/audio/speech","/v1/images/generations","/v1/rerank","/v1/batches","/v1/batches/results"],N=50;function D(t){return new Date(Date.now()-t*1e3).toISOString()}function ke({status:t}){const a=t==="error"?"border-red-200 bg-red-50 text-red-700":"border-[var(--otari-line)] bg-[var(--otari-brand-tint)] text-[var(--otari-brand-dark)]";return e.jsx("span",{className:`inline-flex items-center rounded-full border px-2 py-0.5 text-xs font-medium ${a}`,children:t})}function l({label:t,children:a}){return e.jsxs("div",{className:"flex flex-col gap-0.5",children:[e.jsx("span",{className:"text-[11px] font-medium uppercase tracking-wide text-[var(--otari-muted)]",children:t}),e.jsx("span",{className:"text-sm text-[var(--otari-ink)] break-all",children:a})]})}function we({entry:t}){var a;return e.jsxs("div",{className:"flex flex-col gap-4 px-4 py-4",children:[t.error_message?e.jsxs("div",{className:"flex flex-col gap-1.5",children:[e.jsx("div",{className:"flex items-center justify-between",children:e.jsx("span",{className:"text-[11px] font-medium uppercase tracking-wide text-[var(--otari-muted)]",children:"Error"})}),e.jsx("pre",{className:"max-h-48 overflow-auto rounded-lg border border-red-200 bg-red-50 p-3 text-xs whitespace-pre-wrap break-all text-red-700",children:"The provider returned an error. Inspect gateway logs for details."})]}):null,e.jsxs("div",{className:"grid gap-4 sm:grid-cols-2 lg:grid-cols-4",children:[e.jsx(l,{label:"Provider",children:t.provider??"—"}),e.jsx(l,{label:"Endpoint",children:t.endpoint}),e.jsx(l,{label:"User",children:t.user_id??"—"}),e.jsx(l,{label:"API key",children:t.api_key_id??"—"}),e.jsx(l,{label:"Prompt tokens",children:o(t.prompt_tokens)}),e.jsx(l,{label:"Completion tokens",children:o(t.completion_tokens)}),e.jsx(l,{label:"Total tokens",children:o(t.total_tokens)}),e.jsx(l,{label:"Cost",children:_(t.cost)}),e.jsx(l,{label:"Cache read tokens",children:o(t.cache_read_tokens)}),e.jsx(l,{label:"Cache write tokens",children:o(t.cache_write_tokens)}),e.jsx(l,{label:"1h cache writes",children:o(t.cache_write_1h_tokens??null)}),e.jsx(l,{label:"Total time",children:Z(t.latency_ms)}),e.jsx(l,{label:"Request ID",children:t.id})]}),(a=t.pricing_breakdown)!=null&&a.length?e.jsxs("div",{className:"flex flex-col gap-2",children:[e.jsx("span",{className:"text-[11px] font-medium uppercase tracking-wide text-[var(--otari-muted)]",children:"Billed meters"}),e.jsx("div",{className:"grid gap-2 sm:grid-cols-2 lg:grid-cols-3",children:t.pricing_breakdown.map(r=>e.jsxs(l,{label:r.meter.replaceAll("_"," "),children:[o(r.units)," at ",_(r.rate_per_million)," / 1M, ",_(r.cost)]},r.meter))})]}):null]})}const A=7;function Ce(){var B,z,H;const t=re(),[a]=le(),r=a.get("start_date")??void 0,[i,m]=n.useState(r?null:k),[S,w]=n.useState(()=>r??D(k)),[x,C]=n.useState(()=>a.get("status")??""),[c,M]=n.useState(()=>a.get("model")??""),[h,$]=n.useState(""),[p,U]=n.useState(()=>a.get("user_id")??""),[j,T]=n.useState(0),[J,K]=n.useState(null),P=n.useMemo(()=>({start_date:S,status:x||void 0,model:c.trim()||void 0,endpoint:h||void 0,user_id:p||void 0}),[S,x,c,h,p]);n.useEffect(()=>{T(0)},[P]);const g=ne(P,j,N),v=ie(P),Q=n.useMemo(()=>({start_date:S,status:x||void 0,endpoint:h||void 0,user_id:p||void 0}),[S,x,h,p]),F=((z=(B=oe(Q,"day").data)==null?void 0:B.by_model)==null?void 0:z.filter(s=>!s.is_other&&s.key!==null).map(s=>s.key))??[],b=g.data??[],E=((H=v.data)==null?void 0:H.total)??0,I=!!(x||c.trim()||h||p||i!==null),V=s=>{m(s),w(s===null?void 0:D(s))},W=()=>{m(null),w(void 0),C(""),M(""),$(""),U("")},X=()=>{i!==null&&w(D(i)),g.refetch(),v.refetch()},L=b.length>0,O=L?j*N+1:0,R=j*N+b.length,q=v.isSuccess&&!v.isPlaceholderData,ee=q?E===0?"0 of 0":`${O}–${R} of ${E.toLocaleString()}`:L?`${O}–${R}`:"0",se=q?(j+1)*N<E:b.length===N;return e.jsxs("div",{className:"flex flex-col gap-6",children:[e.jsx(ce,{title:"Activity",description:"A per-request log of what the gateway served: tokens, cost, latency, and failures. No request or response content is stored.",action:e.jsx(f,{variant:"outline",onPress:X,isDisabled:g.isFetching,children:"Refresh"})}),e.jsx(de,{error:g.error??v.error}),e.jsxs("div",{className:"flex flex-wrap items-end gap-3",children:[e.jsx("div",{className:"flex flex-wrap gap-2",children:Ne.map(s=>e.jsx(f,{size:"sm",variant:i===s.seconds?"primary":"outline",onPress:()=>V(s.seconds),children:s.label},s.label))}),e.jsxs("div",{className:"flex flex-wrap items-end gap-3",children:[e.jsx(G,{id:"filter-status",label:"Status",value:x,onChange:C,children:Se.map(s=>e.jsx("option",{value:s.value,children:s.label},s.value))}),e.jsxs(G,{id:"filter-endpoint",label:"Endpoint",value:h,onChange:$,children:[e.jsx("option",{value:"",children:"All endpoints"}),_e.map(s=>e.jsx("option",{value:s,children:s},s))]}),e.jsx(Y,{label:"User",value:p,onChange:U,placeholder:"All users",options:(t.data??[]).map(s=>({value:s.user_id,label:s.alias?`${s.alias} (${s.user_id})`:s.user_id}))}),e.jsx(Y,{label:"Model",value:c,onChange:M,allowsCustom:!0,placeholder:"Any model",options:(c&&!F.includes(c)?[c,...F]:F).map(s=>({value:s,label:s}))}),I?e.jsx(f,{size:"sm",variant:"ghost",onPress:W,children:"Clear filters"}):null]})]}),e.jsxs(ue,{children:[e.jsx(me,{children:e.jsxs("tr",{children:[e.jsx(d,{children:"Time"}),e.jsx(d,{children:"User"}),e.jsx(d,{children:"Model"}),e.jsx(d,{className:"text-right",children:"Tokens"}),e.jsx(d,{className:"text-right",children:"Cost"}),e.jsx(d,{className:"text-right",children:"Total time"}),e.jsx(d,{children:"Status"})]})}),e.jsx("tbody",{children:g.isLoading?e.jsx(xe,{colSpan:A}):b.length===0?e.jsx(he,{colSpan:A,children:I?"No requests match these filters.":"No requests recorded yet."}):b.map(s=>{const te=s.status==="error",y=J===s.id;return e.jsxs(n.Fragment,{children:[e.jsxs(pe,{selected:y,className:te?"bg-red-50":"",onClick:()=>K(ae=>ae===s.id?null:s.id),children:[e.jsx(u,{className:"text-[var(--otari-muted)]",children:e.jsx("span",{title:ve(s.timestamp),children:be(s.timestamp)})}),e.jsx(u,{className:"text-[var(--otari-ink)]",children:s.user_id??"—"}),e.jsx(u,{className:"text-[var(--otari-ink)]",children:s.model}),e.jsx(u,{className:"text-right tabular-nums",children:o(s.total_tokens)}),e.jsx(u,{className:"text-right tabular-nums",children:_(s.cost)}),e.jsx(u,{className:"text-right tabular-nums",children:Z(s.latency_ms)}),e.jsx(u,{children:e.jsx(ke,{status:s.status})})]}),y?e.jsx("tr",{className:"border-b border-[var(--otari-line)] bg-[var(--otari-bg)]",children:e.jsx("td",{colSpan:A,children:e.jsx(we,{entry:s})})}):null]},s.id)})})]}),e.jsxs("div",{className:"flex items-center justify-between gap-3",children:[e.jsxs("span",{className:"inline-flex items-center gap-2 text-sm text-[var(--otari-muted)]",children:[ee,g.isFetching?e.jsx(ge,{size:"sm"}):null]}),e.jsxs("span",{className:"inline-flex gap-1.5",children:[e.jsx(f,{size:"sm",variant:"outline",isDisabled:j===0,onPress:()=>T(s=>Math.max(0,s-1)),children:"Previous"}),e.jsx(f,{size:"sm",variant:"outline",isDisabled:!se,onPress:()=>T(s=>s+1),children:"Next"})]})]})]})}export{Ce as ActivityPage}; | |||
There was a problem hiding this comment.
🎯 Functional Correctness | 🟠 Major | ⚡ Quick win
Add keyboard semantics to all interactive dashboard table rows.
These rows use mouse-only onClick handlers, preventing keyboard users from expanding details, opening model details, or drilling into activity. Fix the shared source pattern and regenerate the affected bundles.
src/gateway/static/dashboard/assets/ActivityPage-DYwWFsLV.js#L1-L1: make request-detail expansion keyboard accessible.src/gateway/static/dashboard/assets/ModelsPage-5SIMv0Ug.js#L1-L1: make model-detail selection keyboard accessible.src/gateway/static/dashboard/assets/UsagePage-ChSRIieH.js#L1-L1: make usage drill-down navigation keyboard accessible.
📍 Affects 3 files
src/gateway/static/dashboard/assets/ActivityPage-DYwWFsLV.js#L1-L1(this comment)src/gateway/static/dashboard/assets/ModelsPage-5SIMv0Ug.js#L1-L1src/gateway/static/dashboard/assets/UsagePage-ChSRIieH.js#L1-L1
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/gateway/static/dashboard/assets/ActivityPage-DYwWFsLV.js` at line 1,
Update the shared dashboard table-row interaction pattern so rows with mouse
onClick handlers also support keyboard activation via appropriate focusability
and keyboard semantics, preserving existing click behavior. Apply the fix to
request-detail expansion in
src/gateway/static/dashboard/assets/ActivityPage-DYwWFsLV.js:1-1, model-detail
selection in src/gateway/static/dashboard/assets/ModelsPage-5SIMv0Ug.js:1-1, and
usage drill-down navigation in
src/gateway/static/dashboard/assets/UsagePage-ChSRIieH.js:1-1, then regenerate
all affected bundles.
| @@ -0,0 +1 @@ | |||
| import{j as e}from"./tanstack-query-W9y7rsMr.js";import{g as rt,u as nt,r as c}from"./react-q-ooZ0ti.js";import{v as at,w as st,e as lt,x as ct,P as ot,E as ut,F as E,I as dt,y as ne,z as pt,A as ht,B as xt,D as A,G as $e,H as Ee,C as Ae,h as me}from"./index-DkHDQ_jO.js";import{T as mt,a as ft,d as Ce,b as re,L as vt,e as U,c as bt}from"./Table-DEsIhjZo.js";import{B as L,c as ke,f as K}from"./heroui-CewI8xK4.js";function We(t){const i=t.indexOf(":");return i>0?t.slice(0,i):"—"}function gt(t,i=Date.now()){const r=new Map;for(const n of t){const a=r.get(n.model_key)??[];a.push(n),r.set(n.model_key,a)}const s=[];for(const n of r.values()){const a=[...n].sort((h,p)=>Date.parse(h.effective_at)-Date.parse(p.effective_at)),d=[...a].reverse().find(h=>Date.parse(h.effective_at)<=i);s.push(d??a[0])}return s.sort((n,a)=>n.model_key.localeCompare(a.model_key))}const _t="otari",Me=[{value:"vision",label:"Vision",test:t=>Array.isArray(t.input_modalities)&&t.input_modalities.includes("image")},{value:"tool_call",label:"Tool calling",test:t=>!!t.tool_call},{value:"reasoning",label:"Reasoning",test:t=>!!t.reasoning},{value:"structured_output",label:"Structured output",test:t=>!!t.structured_output},{value:"attachment",label:"Attachments",test:t=>!!t.attachment},{value:"audio",label:"Audio",test:t=>Array.isArray(t.input_modalities)&&t.input_modalities.includes("audio")},{value:"pdf",label:"PDF",test:t=>Array.isArray(t.input_modalities)&&t.input_modalities.includes("pdf")}],jt=[{key:"reasoning",label:"Reasoning"},{key:"tool_call",label:"Tool calling"},{key:"structured_output",label:"Structured output"},{key:"attachment",label:"Attachments"},{key:"temperature",label:"Temperature"}],Te={text:"Text",image:"Image",audio:"Audio",video:"Video",pdf:"PDF"},Pt=[{value:"0",label:"Any context"},{value:"8000",label:"≥ 8K"},{value:"32000",label:"≥ 32K"},{value:"128000",label:"≥ 128K"},{value:"200000",label:"≥ 200K"},{value:"1000000",label:"≥ 1M"}],yt=[{value:"",label:"Any price"},{value:"1",label:"≤ $1 / 1M in"},{value:"3",label:"≤ $3 / 1M in"},{value:"10",label:"≤ $10 / 1M in"},{value:"30",label:"≤ $30 / 1M in"}],Nt=[{value:"",label:"Base prices"},{value:"8000",label:"Compare at 8K"},{value:"128000",label:"Compare at 128K"},{value:"200000",label:"Compare at 200K"},{value:"500000",label:"Compare at 500K"},{value:"1000000",label:"Compare at 1M"}],St=[{value:"all",label:"Any release date"},{value:"365",label:"Past year"},{value:"730",label:"Past 2 years"},{value:"1095",label:"Past 3 years"}],Ct=1440*60*1e3;function Ie(t,i){const r=`${i}:`;return t.startsWith(r)?t.slice(r.length):t}function kt(t,i){const r={inputPrice:t.inputPrice,outputPrice:t.outputPrice,cacheReadPrice:t.cacheReadPrice,cacheWritePrice:t.cacheWritePrice,cacheWrite1hPrice:t.cacheWrite1hPrice};if(i==null)return r;const s=t.pricingTiers.filter(n=>n.min_input_tokens<=i).sort((n,a)=>a.min_input_tokens-n.min_input_tokens)[0];return s?{inputPrice:s.input_price_per_million??r.inputPrice,outputPrice:s.output_price_per_million??r.outputPrice,cacheReadPrice:s.cache_read_price_per_million??r.cacheReadPrice,cacheWritePrice:s.cache_write_price_per_million??r.cacheWritePrice,cacheWrite1hPrice:s.cache_write_1h_price_per_million??r.cacheWrite1hPrice}:r}function ae(t){const i=Number(t);return t.trim()!==""&&Number.isFinite(i)&&i>=0}function R(t){if(t.trim()==="")return!0;const i=Number(t);return Number.isFinite(i)&&i>=0}function V(t){return t.trim()===""?null:Number(t)}function De(t){return t.map((i,r)=>({id:r,minInputTokens:String(i.min_input_tokens),input:i.input_price_per_million==null?"":String(i.input_price_per_million),output:i.output_price_per_million==null?"":String(i.output_price_per_million),cacheRead:i.cache_read_price_per_million==null?"":String(i.cache_read_price_per_million),cacheWrite:i.cache_write_price_per_million==null?"":String(i.cache_write_price_per_million),cacheWrite1h:i.cache_write_1h_price_per_million==null?"":String(i.cache_write_1h_price_per_million)}))}function Oe(t){const i=new Set;return t.every(r=>{const s=Number(r.minInputTokens),n=[r.input,r.output,r.cacheRead,r.cacheWrite,r.cacheWrite1h].some(a=>a.trim()!=="");return!Number.isInteger(s)||s<=0||i.has(s)||!n?!1:(i.add(s),[r.input,r.output,r.cacheRead,r.cacheWrite,r.cacheWrite1h].every(R))})}function Fe(t){return t.map(i=>({min_input_tokens:Number(i.minInputTokens),...i.input.trim()===""?{}:{input_price_per_million:Number(i.input)},...i.output.trim()===""?{}:{output_price_per_million:Number(i.output)},...i.cacheRead.trim()===""?{}:{cache_read_price_per_million:Number(i.cacheRead)},...i.cacheWrite.trim()===""?{}:{cache_write_price_per_million:Number(i.cacheWrite)},...i.cacheWrite1h.trim()===""?{}:{cache_write_1h_price_per_million:Number(i.cacheWrite1h)}}))}function j({value:t,onChange:i,ariaLabel:r}){return e.jsx("input",{type:"number",step:"any",min:"0",inputMode:"decimal","aria-label":r,value:t,onChange:s=>i(s.target.value),className:"w-28 rounded-md border border-[var(--otari-line)] bg-white px-2 py-1 text-right text-sm tabular-nums focus:border-[var(--otari-brand)] focus:outline-none"})}function ze({tiers:t,onChange:i}){const r=(n,a,d)=>{i(t.map(h=>h.id===n?{...h,[a]:d}:h))},s=()=>{const n=t.reduce((a,d)=>Math.max(a,d.id),-1)+1;i([...t,{id:n,minInputTokens:"128000",input:"",output:"",cacheRead:"",cacheWrite:"",cacheWrite1h:""}])};return e.jsxs("div",{className:"flex flex-col gap-2 rounded-lg border border-[var(--otari-line)] p-3",children:[e.jsxs("div",{className:"flex items-center justify-between gap-3",children:[e.jsxs("div",{children:[e.jsx("div",{className:"text-xs font-medium text-[var(--otari-ink)]",children:"Long-context price tiers"}),e.jsx("p",{className:"text-xs text-[var(--otari-muted)]",children:"At a threshold, listed rates replace the base rate for the whole request."})]}),e.jsx(L,{size:"sm",variant:"outline",onPress:s,children:"Add tier"})]}),t.map(n=>e.jsxs("div",{className:"flex flex-wrap items-end gap-2 border-t border-[var(--otari-line)] pt-2",children:[e.jsxs("label",{className:"flex flex-col gap-1 text-xs text-[var(--otari-muted)]",children:["Context ≥ tokens",e.jsx("input",{type:"number",min:"1",step:"1",inputMode:"numeric","aria-label":"Tier context threshold",value:n.minInputTokens,onChange:a=>r(n.id,"minInputTokens",a.target.value),className:"w-28 rounded-md border border-[var(--otari-line)] bg-white px-2 py-1 text-right text-sm tabular-nums focus:border-[var(--otari-brand)] focus:outline-none"})]}),e.jsxs("label",{className:"flex flex-col gap-1 text-xs text-[var(--otari-muted)]",children:["Input",e.jsx(j,{value:n.input,onChange:a=>r(n.id,"input",a),ariaLabel:"Tier input price"})]}),e.jsxs("label",{className:"flex flex-col gap-1 text-xs text-[var(--otari-muted)]",children:["Output",e.jsx(j,{value:n.output,onChange:a=>r(n.id,"output",a),ariaLabel:"Tier output price"})]}),e.jsxs("label",{className:"flex flex-col gap-1 text-xs text-[var(--otari-muted)]",children:["Cache read",e.jsx(j,{value:n.cacheRead,onChange:a=>r(n.id,"cacheRead",a),ariaLabel:"Tier cache read price"})]}),e.jsxs("label",{className:"flex flex-col gap-1 text-xs text-[var(--otari-muted)]",children:["Cache write",e.jsx(j,{value:n.cacheWrite,onChange:a=>r(n.id,"cacheWrite",a),ariaLabel:"Tier cache write price"})]}),e.jsxs("label",{className:"flex flex-col gap-1 text-xs text-[var(--otari-muted)]",children:["1h write",e.jsx(j,{value:n.cacheWrite1h,onChange:a=>r(n.id,"cacheWrite1h",a),ariaLabel:"Tier 1 hour cache write price"})]}),e.jsx(L,{size:"sm",variant:"ghost",onPress:()=>i(t.filter(a=>a.id!==n.id)),children:"Remove"})]},n.id))]})}function Wt({source:t}){return t==="configured"?e.jsx(K,{size:"sm",color:"default",children:"configured"}):t==="default"||t==="alias"?e.jsx(K,{size:"sm",color:"accent",children:t}):e.jsx("span",{className:"text-xs text-[var(--otari-muted)]",children:"not priced"})}function se({label:t,tone:i="info",children:r}){const s=c.useId();return e.jsxs("span",{className:"group relative inline-flex items-center font-normal normal-case",children:[e.jsx("button",{type:"button","aria-label":t,"aria-describedby":s,className:`inline-flex h-4 w-4 items-center justify-center rounded-full border text-[10px] leading-none ${i==="warning"?"border-[#c2843a] text-[#b45309]":"border-[var(--otari-line)] text-[var(--otari-muted)] hover:border-[var(--otari-brand)] hover:text-[var(--otari-brand)]"}`,children:"i"}),e.jsx("span",{id:s,role:"tooltip",className:"pointer-events-none absolute top-full right-0 z-20 mt-1.5 w-72 rounded-lg border border-[var(--otari-line)] bg-[var(--otari-surface)] px-3 py-2 text-left text-xs font-normal whitespace-normal break-words text-[var(--otari-ink)] opacity-0 shadow-lg transition-opacity group-hover:opacity-100 group-focus-within:opacity-100",children:r})]})}function Mt(){const t=xt();return t.data?t.data.default_pricing?e.jsx(se,{label:"How unpriced models are metered",tone:"info",children:"Default pricing is on: models without a configured price are metered using community-maintained rates (the bundled genai-prices dataset). Set a price to override the fallback."}):e.jsxs(se,{label:"How unpriced models are metered",tone:"warning",children:["Default pricing is off: only models with a configured price are metered.",t.data.require_pricing?" Requests for any other model are rejected (HTTP 402) because require_pricing is on.":" Other models are served without cost tracking."]}):null}function T({label:t,value:i}){return e.jsxs("div",{className:"flex items-baseline justify-between gap-3",children:[e.jsx("span",{className:"text-xs text-[var(--otari-muted)]",children:t}),e.jsx("span",{className:"text-right text-sm text-[var(--otari-ink)] tabular-nums",children:i})]})}function ie({title:t,children:i}){return e.jsxs("div",{className:"flex flex-col gap-2",children:[e.jsx("span",{className:"text-xs font-semibold uppercase tracking-wide text-[var(--otari-muted)]",children:t}),i]})}function Tt({row:t}){const i=$e(),r=Ee(),[s,n]=c.useState(!1),[a,d]=c.useState(""),[h,p]=c.useState(""),[v,P]=c.useState(""),[m,S]=c.useState(""),[k,W]=c.useState(""),[D,y]=c.useState([]),Y=()=>{d(t.inputPrice==null?"":String(t.inputPrice)),p(t.outputPrice==null?"":String(t.outputPrice)),P(t.cacheReadPrice==null?"":String(t.cacheReadPrice)),S(t.cacheWritePrice==null?"":String(t.cacheWritePrice)),W(t.cacheWrite1hPrice==null?"":String(t.cacheWrite1hPrice)),y(De(t.pricingTiers)),n(!0)},M=ae(a)&&ae(h)&&R(v)&&R(m)&&R(k)&&Oe(D),G=()=>{M&&i.mutate({model_key:t.key,input_price_per_million:Number(a),output_price_per_million:Number(h),cache_read_price_per_million:V(v),cache_write_price_per_million:V(m),cache_write_1h_price_per_million:V(k),pricing_tiers:Fe(D)},{onSuccess:()=>n(!1)})};return s?e.jsxs("div",{className:"flex flex-col gap-2",children:[e.jsxs("div",{className:"flex items-center justify-between gap-2",children:[e.jsx("span",{className:"text-xs text-[var(--otari-muted)]",children:"Input $ / 1M"}),e.jsx(j,{value:a,onChange:d,ariaLabel:`Input price for ${t.key}`})]}),e.jsxs("div",{className:"flex items-center justify-between gap-2",children:[e.jsx("span",{className:"text-xs text-[var(--otari-muted)]",children:"Output $ / 1M"}),e.jsx(j,{value:h,onChange:p,ariaLabel:`Output price for ${t.key}`})]}),e.jsxs("div",{className:"flex items-center justify-between gap-2",children:[e.jsx("span",{className:"text-xs text-[var(--otari-muted)]",children:"Cache read $ / 1M"}),e.jsx(j,{value:v,onChange:P,ariaLabel:`Cache read price for ${t.key}`})]}),e.jsxs("div",{className:"flex items-center justify-between gap-2",children:[e.jsx("span",{className:"text-xs text-[var(--otari-muted)]",children:"Cache write $ / 1M"}),e.jsx(j,{value:m,onChange:S,ariaLabel:`Cache write price for ${t.key}`})]}),e.jsxs("div",{className:"flex items-center justify-between gap-2",children:[e.jsx("span",{className:"text-xs text-[var(--otari-muted)]",children:"1h cache write $ / 1M"}),e.jsx(j,{value:k,onChange:W,ariaLabel:`1 hour cache write price for ${t.key}`})]}),e.jsx(ze,{tiers:D,onChange:y}),e.jsxs("div",{className:"flex items-center gap-2",children:[e.jsx(L,{size:"sm",variant:"primary",isDisabled:i.isPending||!M,onPress:G,children:"Save"}),e.jsx(L,{size:"sm",variant:"ghost",isDisabled:i.isPending,onPress:()=>n(!1),children:"Cancel"})]}),i.error?e.jsx("span",{className:"text-xs text-red-700",children:me(i.error)}):null]}):e.jsxs("div",{className:"flex flex-col gap-2",children:[e.jsx(T,{label:"Input",value:t.inputPrice==null?"—":`${A(t.inputPrice)} / 1M`}),e.jsx(T,{label:"Output",value:t.outputPrice==null?"—":`${A(t.outputPrice)} / 1M`}),e.jsx(T,{label:"Cache read",value:t.cacheReadPrice==null?"—":`${A(t.cacheReadPrice)} / 1M`}),e.jsx(T,{label:"Cache write",value:t.cacheWritePrice==null?"—":`${A(t.cacheWritePrice)} / 1M`}),e.jsx(T,{label:"1h cache write",value:t.cacheWrite1hPrice==null?"—":`${A(t.cacheWrite1hPrice)} / 1M`}),e.jsx(T,{label:"Context tiers",value:t.pricingTiers.length?`${t.pricingTiers.length} configured`:"—"}),e.jsxs("div",{className:"flex items-center gap-2 pt-1",children:[e.jsx(L,{size:"sm",variant:"outline",onPress:Y,children:t.source==="configured"?"Edit price":"Set price"}),t.source==="configured"?e.jsxs(e.Fragment,{children:[e.jsx(Ae,{confirmLabel:"Reset",isPending:r.isPending,onConfirm:()=>r.mutate(t.key),children:"Reset"}),e.jsx(se,{label:"What reset does",children:"Removes the custom price. The model reverts to the default rate (genai-prices) when default pricing is on, otherwise it is metered at no cost."})]}):null,r.error?e.jsx("span",{className:"text-xs text-red-700",children:me(r.error)}):null]})]})}function It({row:t,metadata:i,metadataAvailable:r,onMakeAlias:s,onClose:n}){const a=(i==null?void 0:i.input_modalities)??[],d=(i==null?void 0:i.output_modalities)??[],h=jt.filter(({key:p})=>i==null?void 0:i[p]);return e.jsx(ke,{children:e.jsxs(ke.Content,{className:"flex flex-col gap-5 p-5",children:[e.jsxs("div",{className:"flex items-start justify-between gap-3",children:[e.jsxs("div",{className:"min-w-0",children:[e.jsxs("div",{className:"flex items-center gap-2",children:[e.jsx("h2",{className:"text-base font-semibold break-all text-[var(--otari-ink)]",children:t.model}),i!=null&&i.deprecated?e.jsx(K,{size:"sm",color:"danger",children:"deprecated"}):null]}),e.jsxs("p",{className:"mt-1 text-xs break-all text-[var(--otari-muted)]",children:["Selector: ",e.jsx("code",{children:t.key})]}),i!=null&&i.family?e.jsx("p",{className:"text-xs text-[var(--otari-muted)]",children:i.family}):null]}),e.jsx("button",{type:"button","aria-label":"Close model details",onClick:n,className:"-mt-1 -mr-1 shrink-0 rounded-md px-1.5 py-0.5 text-lg leading-none text-[var(--otari-muted)] hover:bg-[var(--otari-bg)] hover:text-[var(--otari-ink)]",children:"✕"})]}),i!=null&&i.description?e.jsx("p",{className:"text-sm text-[var(--otari-ink)]",children:i.description}):null,e.jsxs(ie,{title:"Pricing",children:[e.jsxs("div",{className:"flex items-center gap-2",children:[e.jsx(Wt,{source:t.source}),t.isDiscovered?null:e.jsx("span",{className:"text-xs text-[var(--otari-muted)]",children:"not discovered"})]}),e.jsx(Tt,{row:t},t.key),e.jsx(L,{size:"sm",variant:"outline",onPress:()=>s(t.key),children:"Make an alias"})]}),e.jsxs(ie,{title:"Specs",children:[e.jsx(T,{label:"Context window",value:ne(t.contextWindow)}),e.jsx(T,{label:"Max output",value:ne((i==null?void 0:i.max_output_tokens)??null)}),e.jsx(T,{label:"Knowledge cutoff",value:(i==null?void 0:i.knowledge_cutoff)??"—"}),e.jsx(T,{label:"Released",value:ht(i==null?void 0:i.release_date)}),e.jsx(T,{label:"Open weights",value:i?i.open_weights?"Yes":"No":"—"})]}),e.jsx(ie,{title:"Modalities",children:a.length===0&&d.length===0?e.jsx("span",{className:"text-sm text-[var(--otari-muted)]",children:"Unknown."}):e.jsxs("div",{className:"flex flex-col gap-1.5 text-xs text-[var(--otari-muted)]",children:[e.jsxs("div",{className:"flex flex-wrap items-center gap-1",children:[e.jsx("span",{children:"In:"}),a.map(p=>e.jsx(K,{size:"sm",color:"default",children:Te[p]??p},p))]}),e.jsxs("div",{className:"flex flex-wrap items-center gap-1",children:[e.jsx("span",{children:"Out:"}),d.map(p=>e.jsx(K,{size:"sm",color:"default",children:Te[p]??p},p))]})]})}),e.jsx(ie,{title:"Capabilities",children:h.length>0?e.jsx("div",{className:"flex flex-wrap gap-1.5",children:h.map(({key:p,label:v})=>e.jsx(K,{size:"sm",color:"default",children:v},p))}):e.jsx("span",{className:"text-sm text-[var(--otari-muted)]",children:r?"None reported.":"Extended metadata unavailable (models.dev disabled or unreachable)."})})]})})}const Lt=15,$t=[{value:"15",label:"15 per page"},{value:"25",label:"25 per page"},{value:"50",label:"50 per page"}];function Et({value:t,onChange:i,placeholder:r}){return e.jsx("input",{type:"search",value:t,onChange:s=>i(s.target.value),placeholder:r,"aria-label":r,className:"w-full max-w-xs rounded-md border border-[var(--otari-line)] bg-white px-3 py-1.5 text-sm focus:border-[var(--otari-brand)] focus:outline-none"})}function At({page:t,pageCount:i,total:r,pageSize:s,onPage:n,onPageSize:a}){return e.jsxs("div",{className:"flex items-center justify-between px-1 pt-1 text-sm text-[var(--otari-muted)]",children:[e.jsxs("span",{children:[i>1?`Page ${t+1} of ${i} · `:"",pt(r)," model",r===1?"":"s"]}),e.jsxs("span",{className:"inline-flex gap-2",children:[e.jsx(E,{ariaLabel:"Rows per page",value:String(s),onChange:d=>a(Number(d)),options:$t}),i>1?e.jsxs(e.Fragment,{children:[e.jsx(L,{size:"sm",variant:"outline",isDisabled:t===0,onPress:()=>n(t-1),children:"Prev"}),e.jsx(L,{size:"sm",variant:"outline",isDisabled:t>=i-1,onPress:()=>n(t+1),children:"Next"})]}):null]})]})}const Be="otari.dashboard.modelsSort",xe={col:"model",dir:"asc"},Dt=["model","released","input","output"];function Ot(){if(typeof window>"u")return xe;try{const t=window.localStorage.getItem(Be);if(!t)return xe;const i=JSON.parse(t);if(Dt.includes(i.col)&&(i.dir==="asc"||i.dir==="desc"))return{col:i.col,dir:i.dir}}catch{}return xe}function Le({label:t,col:i,sort:r,onSort:s,align:n="left",info:a}){const d=r.col===i;return e.jsx(re,{className:n==="right"?"text-right":void 0,ariaSort:d?r.dir==="asc"?"ascending":"descending":"none",children:e.jsxs("span",{className:`inline-flex items-center gap-1.5 ${n==="right"?"flex-row-reverse":""}`,children:[e.jsxs("button",{type:"button",onClick:()=>s(i),className:`inline-flex items-center gap-1 ${n==="right"?"flex-row-reverse":""} hover:text-[var(--otari-ink)]`,children:[t,e.jsx("span",{className:"text-[10px] text-[var(--otari-muted)]",children:d?r.dir==="asc"?"▲":"▼":"↕"})]}),a]})})}function Ft({row:t,onClose:i}){const r=$e(),s=Ee(),[n,a]=c.useState(t.inputPrice==null?"":String(t.inputPrice)),[d,h]=c.useState(t.outputPrice==null?"":String(t.outputPrice)),[p,v]=c.useState(t.cacheReadPrice==null?"":String(t.cacheReadPrice)),[P,m]=c.useState(t.cacheWritePrice==null?"":String(t.cacheWritePrice)),[S,k]=c.useState(t.cacheWrite1hPrice==null?"":String(t.cacheWrite1hPrice)),[W,D]=c.useState(De(t.pricingTiers)),y=ae(n)&&ae(d)&&R(p)&&R(P)&&R(S)&&Oe(W),Y=()=>{y&&r.mutate({model_key:t.key,input_price_per_million:Number(n),output_price_per_million:Number(d),cache_read_price_per_million:V(p),cache_write_price_per_million:V(P),cache_write_1h_price_per_million:V(S),pricing_tiers:Fe(W)},{onSuccess:i})};return e.jsxs("div",{className:"flex flex-col gap-3 px-4 py-3",children:[e.jsxs("div",{className:"flex flex-wrap items-center gap-3",children:[e.jsx("span",{className:"text-xs font-medium break-all text-[var(--otari-muted)]",children:t.key}),e.jsxs("label",{className:"flex items-center gap-1.5 text-xs text-[var(--otari-muted)]",children:["Input $ / 1M",e.jsx(j,{value:n,onChange:a,ariaLabel:`Input price for ${t.key}`})]}),e.jsxs("label",{className:"flex items-center gap-1.5 text-xs text-[var(--otari-muted)]",children:["Output $ / 1M",e.jsx(j,{value:d,onChange:h,ariaLabel:`Output price for ${t.key}`})]}),e.jsxs("label",{className:"flex items-center gap-1.5 text-xs text-[var(--otari-muted)]",children:["Cache read $ / 1M",e.jsx(j,{value:p,onChange:v,ariaLabel:`Cache read price for ${t.key}`})]}),e.jsxs("label",{className:"flex items-center gap-1.5 text-xs text-[var(--otari-muted)]",children:["Cache write $ / 1M",e.jsx(j,{value:P,onChange:m,ariaLabel:`Cache write price for ${t.key}`})]}),e.jsxs("label",{className:"flex items-center gap-1.5 text-xs text-[var(--otari-muted)]",children:["1h cache write $ / 1M",e.jsx(j,{value:S,onChange:k,ariaLabel:`1 hour cache write price for ${t.key}`})]}),e.jsx(L,{size:"sm",variant:"primary",isDisabled:r.isPending||!y,onPress:Y,children:r.isPending?"Saving…":"Save"}),e.jsx(L,{size:"sm",variant:"ghost",isDisabled:r.isPending,onPress:i,children:"Cancel"}),t.source==="configured"?e.jsxs("span",{className:"inline-flex items-center gap-1",children:[e.jsx(Ae,{confirmLabel:"Reset",isPending:s.isPending,onConfirm:()=>s.mutate(t.key,{onSuccess:i}),children:"Reset"}),e.jsx(se,{label:"What reset does",children:"Removes the custom price. The model reverts to the default rate (genai-prices) when default pricing is on, otherwise it is metered at no cost."})]}):null,r.error||s.error?e.jsx("span",{className:"text-xs text-red-700",children:me(r.error??s.error)}):null]}),e.jsx(ze,{tiers:W,onChange:D})]})}function zt({primary:t,secondary:i,rowKey:r,primaryLabel:s,secondaryLabel:n,onEdit:a}){const d=(h,p)=>e.jsx("button",{type:"button","aria-label":`Edit ${p} price for ${r}`,className:"tabular-nums hover:text-[var(--otari-brand-dark)] hover:underline",onClick:v=>{v.stopPropagation(),a()},children:h==null?"—":A(h)});return e.jsxs("span",{className:"inline-flex items-center justify-end gap-1",children:[d(t,s),e.jsx("span",{className:"text-[var(--otari-muted)]",children:"/"}),d(i,n)]})}function Bt({rates:t,rowKey:i,onEdit:r}){const s=[t.cacheReadPrice==null?null:`R ${A(t.cacheReadPrice)}`,t.cacheWritePrice==null?null:`W ${A(t.cacheWritePrice)}`,t.cacheWrite1hPrice==null?null:`1h ${A(t.cacheWrite1hPrice)}`].filter(n=>n!==null);return e.jsx("button",{type:"button","aria-label":`Edit caching price for ${i}`,className:"max-w-44 text-right text-xs leading-5 text-[var(--otari-muted)] hover:text-[var(--otari-brand-dark)] hover:underline",onClick:n=>{n.stopPropagation(),r()},children:s.length>0?s.join(" · "):"Input-rate fallback"})}function Rt({row:t,onEdit:i}){const r=[...t.pricingTiers].sort((n,a)=>n.min_input_tokens-a.min_input_tokens).map(n=>ne(n.min_input_tokens)),s=r.length===0?"Base only":`${r.length} tier${r.length===1?"":"s"} · ≥ ${r.join(", ")}`;return e.jsx("button",{type:"button","aria-label":`Edit pricing policy for ${t.key}`,className:"max-w-40 text-right text-xs leading-5 text-[var(--otari-muted)] hover:text-[var(--otari-brand-dark)] hover:underline",onClick:n=>{n.stopPropagation(),i()},children:s})}function Kt({rows:t,isLoading:i,empty:r,sort:s,onSort:n,selectedKey:a,onSelect:d,pricingKey:h,onSetPricingKey:p,comparisonContextTokens:v}){const P=v==null?"Base":`at ${ne(v)}`;return e.jsxs(mt,{children:[e.jsx(ft,{children:e.jsxs(Ce,{children:[e.jsx(Le,{label:"Model",col:"model",sort:s,onSort:n}),e.jsx(re,{children:"Provider"}),e.jsx(Le,{label:`${P} in / out $ / 1M`,col:"input",sort:s,onSort:n,align:"right",info:e.jsx(Mt,{})}),e.jsxs(re,{className:"text-right",children:["Caching ",v==null?"policy":P]}),e.jsx(re,{className:"text-right",children:"Pricing policy"})]})}),e.jsx("tbody",{children:i?e.jsx(vt,{colSpan:5}):t.length>0?t.map(m=>{const S=()=>p(h===m.key?null:m.key),k=kt(m,v);return e.jsxs(c.Fragment,{children:[e.jsxs(Ce,{onClick:()=>d(m.key),selected:m.key===a,children:[e.jsxs(U,{className:"font-medium break-all",children:[m.model,e.jsx("span",{className:"sr-only",children:m.key})]}),e.jsx(U,{className:"text-[var(--otari-muted)]",children:m.provider}),e.jsx(U,{className:"text-right",children:e.jsx(zt,{primary:k.inputPrice,secondary:k.outputPrice,rowKey:m.key,primaryLabel:"input",secondaryLabel:"output",onEdit:S})}),e.jsx(U,{className:"text-right",children:e.jsx(Bt,{rates:k,rowKey:m.key,onEdit:S})}),e.jsx(U,{className:"text-right",children:e.jsx(Rt,{row:m,onEdit:S})})]}),h===m.key?e.jsx("tr",{className:"border-b border-[var(--otari-line)] bg-[var(--otari-bg)] last:border-b-0",children:e.jsx("td",{colSpan:5,children:e.jsx(Ft,{row:m,onClose:()=>p(null)})})}):null]},m.key)}):e.jsx(bt,{colSpan:5,children:r})})]})}function Vt({providers:t}){return t.length===0?null:e.jsxs(dt,{tone:"warning",children:["Could not list ",t.map(i=>i.provider).join(", "),". Check that provider's credentials in config.yml; its models are missing from the list below."]})}function Jt(){var je,Pe,ye;const t=rt(),[i]=nt(),r=at(),s=st(),n=lt(),a=ct(),[d,h]=c.useState(""),[p,v]=c.useState(0),[P,m]=c.useState(Lt),[S,k]=c.useState(null),[W,D]=c.useState(null),[y,Y]=c.useState(Ot);c.useEffect(()=>{try{window.localStorage.setItem(Be,JSON.stringify(y))}catch{}},[y]);const[M,G]=c.useState(i.get("provider")||"all"),[F,Re]=c.useState("all"),[H,Ke]=c.useState("all"),[q,Ve]=c.useState("all"),[le,Ye]=c.useState("0"),[J,He]=c.useState(""),[Z,qe]=c.useState("all"),[ce,we]=c.useState(""),O=((je=a.data)==null?void 0:je.models)??{},Ue=((Pe=a.data)==null?void 0:Pe.available)??!1,fe=c.useMemo(()=>{var l;return new Set((((l=n.data)==null?void 0:l.providers)??[]).flatMap(f=>f.models.map(o=>o.key)))},[n.data]),Ge=l=>{h(l),v(0)},z=l=>f=>{l(f),v(0)},Je=l=>{Y(f=>f.col===l?{col:l,dir:f.dir==="asc"?"desc":"asc"}:{col:l,dir:l==="released"?"desc":"asc"}),v(0)},X=c.useMemo(()=>{var C,g,_,$,I,B,Ne;const l=new Map(gt(s.data??[]).map(u=>[u.model_key,u])),f=[],o=new Set,b=(u,he,Se,x)=>{if(o.has(u))return;o.add(u);const N=l.get(u);f.push({key:u,model:Ie(he,Se),provider:Se,isDiscovered:fe.has(u),contextWindow:(x==null?void 0:x.contextWindow)??null,inputPrice:N?N.input_price_per_million:(x==null?void 0:x.inputPrice)??null,outputPrice:N?N.output_price_per_million:(x==null?void 0:x.outputPrice)??null,cacheReadPrice:N?N.cache_read_price_per_million:(x==null?void 0:x.cacheReadPrice)??null,cacheWritePrice:N?N.cache_write_price_per_million:(x==null?void 0:x.cacheWritePrice)??null,cacheWrite1hPrice:N?N.cache_write_1h_price_per_million??null:(x==null?void 0:x.cacheWrite1hPrice)??null,pricingTiers:N?N.pricing_tiers??[]:(x==null?void 0:x.pricingTiers)??[],source:N?"configured":(x==null?void 0:x.source)??"none"})};for(const u of((C=r.data)==null?void 0:C.data)??[]){if(u.owned_by===_t)continue;const he=u.pricing_source==="default"?"default":u.pricing?"configured":"none";b(u.id,u.id,u.owned_by||We(u.id),{key:u.id,model:u.id,provider:u.owned_by,contextWindow:u.context_window,inputPrice:((g=u.pricing)==null?void 0:g.input_price_per_million)??null,outputPrice:((_=u.pricing)==null?void 0:_.output_price_per_million)??null,cacheReadPrice:(($=u.pricing)==null?void 0:$.cache_read_price_per_million)??null,cacheWritePrice:((I=u.pricing)==null?void 0:I.cache_write_price_per_million)??null,cacheWrite1hPrice:((B=u.pricing)==null?void 0:B.cache_write_1h_price_per_million)??null,pricingTiers:((Ne=u.pricing)==null?void 0:Ne.pricing_tiers)??[],source:he})}for(const u of l.keys())b(u,u,We(u));return f},[r.data,s.data,fe]),Ze=c.useMemo(()=>new Map(X.map(l=>[l.key,l])),[X]),Q=c.useMemo(()=>{var o,b,C;const l=X.map(g=>{var _,$;return{...g,contextWindow:g.contextWindow??((_=O[g.key])==null?void 0:_.context_window)??null,releaseDate:(($=O[g.key])==null?void 0:$.release_date)??null}}),f=new Set(l.map(g=>g.key));for(const g of((o=n.data)==null?void 0:o.providers)??[])for(const _ of g.models)f.has(_.key)||(f.add(_.key),l.push({key:_.key,model:Ie(_.key,g.provider),provider:g.provider,isDiscovered:!0,contextWindow:((b=O[_.key])==null?void 0:b.context_window)??null,releaseDate:((C=O[_.key])==null?void 0:C.release_date)??null,inputPrice:null,outputPrice:null,cacheReadPrice:null,cacheWritePrice:null,cacheWrite1hPrice:null,pricingTiers:[],source:"none"}));return l},[X,n.data,O]),Xe=(((ye=n.data)==null?void 0:ye.providers)??[]).filter(l=>!l.ok),ee=c.useMemo(()=>{const l=Array.from(new Set(Q.map(f=>f.provider))).sort((f,o)=>f.localeCompare(o));return[{value:"all",label:"All providers"},...l.map(f=>({value:f,label:f}))]},[Q]);c.useEffect(()=>{M==="all"||ee.length<=1||ee.some(l=>l.value===M)||G("all")},[ee,M]);const w=d.trim().toLowerCase(),oe=Number(le)||0,ue=J===""?Number.POSITIVE_INFINITY:Number(J),Qe=ce===""?null:Number(ce),de=Z==="all"?null:Date.now()-Number(Z)*Ct,pe=c.useMemo(()=>{const l=o=>{if(w&&!o.key.toLowerCase().includes(w)&&!o.provider.toLowerCase().includes(w)||M!=="all"&&o.provider!==M||F==="configured"&&o.source!=="configured"||F==="default"&&o.source!=="default"||F==="priced"&&o.inputPrice==null||F==="unpriced"&&o.inputPrice!=null||H==="discovered"&&!o.isDiscovered||H==="custom"&&o.isDiscovered)return!1;if(q!=="all"){const b=Me.find(g=>g.value===q),C=O[o.key];if(!b||!C||!b.test(C))return!1}if(oe>0&&(o.contextWindow==null||o.contextWindow<oe)||ue!==Number.POSITIVE_INFINITY&&(o.inputPrice==null||o.inputPrice>ue))return!1;if(de!=null){const b=o.releaseDate?Date.parse(o.releaseDate):Number.NaN;if(Number.isNaN(b)||b<de)return!1}return!0},f=(o,b)=>{const C=y.dir==="asc"?1:-1;if(y.col==="model")return o.model.localeCompare(b.model)*C;if(y.col==="released"){const I=o.releaseDate??null,B=b.releaseDate??null;return!I&&!B?o.model.localeCompare(b.model):I?B?(I<B?-1:I>B?1:0)*C||o.model.localeCompare(b.model):-1:1}const g=I=>y.col==="input"?I.inputPrice:I.outputPrice,_=g(o),$=g(b);return _==null&&$==null?o.model.localeCompare(b.model):_==null?1:$==null?-1:(_-$)*C||o.model.localeCompare(b.model)};return Q.filter(l).sort(f)},[Q,w,M,F,H,q,oe,ue,de,O,y]),ve=pe.length,be=Math.max(1,Math.ceil(ve/P)),ge=Math.min(p,be-1),_e=ge*P,et=pe.slice(_e,_e+P),tt=r.isLoading||s.isLoading||n.isLoading,it=w!==""||M!=="all"||F!=="all"||H!=="all"||q!=="all"||le!=="0"||J!==""||Z!=="all"?"No models match your filters.":"No models yet. Add a provider on the Providers page, or price a model below.",te=W?pe.find(l=>l.key===W)??Ze.get(W)??null:null;return e.jsxs("div",{className:"flex flex-col gap-6",children:[e.jsx(ot,{title:"Models",description:"Every model your providers can serve. Set a price on any model so budgets and usage tracking work."}),e.jsx(ut,{error:r.error??s.error??n.error??a.error}),e.jsxs("div",{className:`grid gap-4 lg:items-start ${te?"lg:grid-cols-[minmax(0,1fr)_360px]":"grid-cols-1"}`,children:[e.jsxs("div",{className:"flex min-w-0 flex-col gap-3",children:[e.jsxs("div",{className:"flex flex-wrap items-center gap-2",children:[e.jsx(Et,{value:d,onChange:Ge,placeholder:"Search models…"}),e.jsx(E,{ariaLabel:"Filter by provider",value:M,onChange:z(G),options:ee}),e.jsx(E,{ariaLabel:"Filter by pricing",value:F,onChange:z(Re),options:[{value:"all",label:"Any pricing"},{value:"configured",label:"Custom price"},{value:"default",label:"Default price"},{value:"priced",label:"Priced"},{value:"unpriced",label:"Unpriced"}]}),e.jsx(E,{ariaLabel:"Filter by source",value:H,onChange:z(Ke),options:[{value:"all",label:"Any source"},{value:"discovered",label:"Discovered"},{value:"custom",label:"Custom (not discovered)"}]}),e.jsx(E,{ariaLabel:"Filter by capability",value:q,onChange:z(Ve),options:[{value:"all",label:"Any capability"},...Me.map(l=>({value:l.value,label:l.label}))]}),e.jsx(E,{ariaLabel:"Minimum context window",value:le,onChange:z(Ye),options:Pt}),e.jsx(E,{ariaLabel:"Maximum input price",value:J,onChange:z(He),options:yt}),e.jsx(E,{ariaLabel:"Compare prices at context",value:ce,onChange:we,options:Nt}),e.jsx(E,{ariaLabel:"Filter by release date",value:Z,onChange:z(qe),options:St})]}),e.jsx(Vt,{providers:Xe}),e.jsx(Kt,{rows:et,isLoading:tt,empty:it,sort:y,onSort:Je,selectedKey:W,onSelect:D,pricingKey:S,onSetPricingKey:k,comparisonContextTokens:Qe}),e.jsx("div",{className:"flex flex-wrap items-center justify-between gap-2",children:e.jsx(At,{page:ge,pageCount:be,total:ve,pageSize:P,onPage:v,onPageSize:l=>{m(l),v(0)}})})]}),te?e.jsx("aside",{className:"lg:sticky lg:top-4",children:e.jsx(It,{row:te,metadata:O[te.key],metadataAvailable:Ue,onMakeAlias:l=>t(`/aliases?target=${encodeURIComponent(l)}`),onClose:()=>D(null)})}):null]})]})}export{Jt as ModelsPage}; | |||
There was a problem hiding this comment.
🎯 Functional Correctness | 🟠 Major | ⚡ Quick win
Make model-row selection keyboard accessible.
The model table row relies on onClick alone, so keyboard users cannot open model details. Use a focusable control or add Enter/Space handling and appropriate row state semantics.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/gateway/static/dashboard/assets/ModelsPage-5SIMv0Ug.js` at line 1, The
model-row selection behavior in Kt currently relies only on onClick and is
inaccessible from the keyboard. Update the row selection implementation to use a
focusable interactive control or add Enter/Space handling with appropriate
focusability and selected-row semantics, while preserving the existing onSelect
behavior that opens model details.
The pre-call estimate treated a requested Anthropic cache write as an additional input meter, charging every estimated prompt token once at the input rate and again at the cache-write rate, and selecting the context tier from double the token count. Actual billing charges a cache-write token only at the write rate, so the reservation over-reserved (roughly input + write per token) and could reject requests that would comfortably fit their budget. Each prompt token is billed as exactly one of fresh input, cache read, or cache write, so the true upper bound for the input side is the token count times the dearest applicable rate. Reserve the input at max(input rate, cache-write rate) when a write is requested, and select the tier from the estimated input, which approximates the real billable total. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
njbrake
temporarily deployed
to
integration-tests
July 23, 2026 10:53 — with
GitHub Actions
Inactive
GitHub Actions
Inactive
There was a problem hiding this comment.
🧹 Nitpick comments (2)
tests/unit/test_budget_cost_estimate.py (2)
94-103: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick winCover the one-hour fallback separately from unpriced writes.
Setting both cache-write rates to
Noneonly verifies input-rate fallback. It does not test the required case where the 1-hour rate is unset but the base cache-write rate remains configured. Keep the base rate configured and set onlycache_write_1h_price_per_million=None, then expect the base cache-write rate.Suggested test adjustment
- _pricing(cache_write_price_per_million=None, cache_write_1h_price_per_million=None), + _pricing(cache_write_1h_price_per_million=None), ... - assert estimate == pytest.approx(1_000 * 3.0 / 1_000_000) + assert estimate == pytest.approx(1_000 * 3.75 / 1_000_000)🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/unit/test_budget_cost_estimate.py` around lines 94 - 103, Update test_estimate_cost_falls_back_to_input_when_cache_write_is_unpriced to keep the base cache-write price configured while setting only cache_write_1h_price_per_million to None, and assert the estimate uses the configured base cache-write rate for the 1-hour TTL.
68-79: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick winMake the tiered cache-write assertion distinguish the selected meter.
The expected
7.5can still pass ifcache_write_ttlis ignored, because the estimator uses the maximum of the input and cache-write rates. Give the tier’s cache-write rate a distinct, higher value so this test actually catches regressions in tiered cache-write selection. Also, the comment should say “estimated input tokens,” not “real billable total,” since this pre-call estimator selects tiers from the estimate.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/unit/test_budget_cost_estimate.py` around lines 68 - 79, The test test_estimate_cost_reserves_cache_write_using_the_context_tier should configure the pricing fixture with a distinct, higher tiered cache-write rate so the expected cost specifically verifies cache-write meter selection rather than the maximum-rate fallback. Update the assertion to use that cache-write rate and revise the comment to say the tier is selected from the estimated input tokens.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Nitpick comments:
In `@tests/unit/test_budget_cost_estimate.py`:
- Around line 94-103: Update
test_estimate_cost_falls_back_to_input_when_cache_write_is_unpriced to keep the
base cache-write price configured while setting only
cache_write_1h_price_per_million to None, and assert the estimate uses the
configured base cache-write rate for the 1-hour TTL.
- Around line 68-79: The test
test_estimate_cost_reserves_cache_write_using_the_context_tier should configure
the pricing fixture with a distinct, higher tiered cache-write rate so the
expected cost specifically verifies cache-write meter selection rather than the
maximum-rate fallback. Update the assertion to use that cache-write rate and
revise the comment to say the tier is selected from the estimated input tokens.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 5186d367-52c7-4ddf-be8b-0fb2249f8347
📒 Files selected for processing (3)
src/gateway/services/budget_service.pysrc/gateway/services/metered_pricing.pytests/unit/test_budget_cost_estimate.py
🚧 Files skipped from review as they are similar to previous changes (2)
- src/gateway/services/budget_service.py
- src/gateway/services/metered_pricing.py
14 tasks
14 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Adds provider-metered usage pricing for cache reads, regular cache writes, Anthropic one-hour cache writes, and request-level long-context price tiers. Usage records now retain billing meters and a pricing breakdown for auditability.
The standalone dashboard now compares base and effective rates at a selected context size, presents compact cache and tier policies, supports inline tier editing, and avoids repeating the provider instance in model labels.
PR Type
Relevant issues
Closes #357
Closes #372
Refs #376
Refs mozilla-ai/any-llm#1195
Checklist
tests/unit,tests/integration).make lint,make typecheck,make test).uv run python scripts/generate_openapi.py).Validation:
make lint,make typecheck,uv run pytest tests/unit(857 passed, 1 skipped),npm --prefix web run typecheck,npm --prefix web test(195 passed), dashboard production build, and OpenAPI freshness check all pass. Integration tests could not run locally because Docker is unavailable.AI Usage
AI Model/Tool used: Codex
Any additional AI details you would like to share: Implementation and validation were performed interactively with maintainer direction.
Summary
Technical notes