@zaip-ai/cli 2.6.0 → 3.0.0

package/templates/.agents/dashboard/README.md

@@ -34,6 +34,16 @@ previews e bancos de teste. O launcher não depende delas nem gera novos bancos
   o mesmo arquivo em uma transação, incluindo os dados das primeiras telas.
 - Leitores continuam usando a última transação confirmada durante a atualização. Após o commit,
   o servidor renova suas consultas e caches e informa `current`. Uma falha reverte a transação.
+  A verificação de integridade e schema ocorre no processo de atualização; o servidor exige
+  saída zero e recibo válido para o caminho exato antes de renovar o leitor. Recibo ausente,
+  schema incompatível ou erro do processo não autorizam a renovação. Essa checagem pesada
+  não se repete no event loop HTTP durante a navegação.
+- Os gráficos PRD/SPEC reutilizam as métricas já fornecidas pela projeção rápida. Carregar
+  detalhes continua disponível quando essas métricas estão ausentes; clicar nos gráficos
+  não deve disparar uma leitura redundante de todo o histórico diário.
+- A conexão de presença do navegador recebe heartbeat SSE a cada 10 segundos. Fechar a última
+  conexão mantém o encerramento automático; o heartbeat evita que uma conexão ainda aberta
+  seja confundida com uma conexão ociosa expirada. O banco permanece preservado.
 - O inventário mantém checkpoints por arquivo no próprio SQLite: caminho, tamanho, identidade,
   datas de modificação e alteração, hash e última atualização processada. Arquivos intactos
   reutilizam seu hash; arquivos novos ou alterados têm o conteúdo verificado novamente.
@@ -86,6 +96,12 @@ concorrentes. Não copie um arquivo SQLite ativo sem considerar seu WAL.
 
 ## Limites
 
+O período selecionado pertence à interface, não às respostas assíncronas. As datas de cobertura
+de **Todo o Período** são derivadas dos dados disponíveis sem disparar uma nova seleção; a chegada
+de uma resposta antiga não deve substituir uma escolha recente. Selecionar novamente o mesmo
+intervalo não abre um bloqueio de carregamento sem consulta nova. A troca real de intervalo
+invalida a resposta anterior e mantém os limites de data comuns às abas e às exportações.
+
 - O `node:sqlite` do Node 22.16 ainda emite `ExperimentalWarning`.
 - A primeira criação ou uma mudança incompatível de esquema exige preparação antes de servir.
   Atualizações volumosas podem demorar; o banco válido existente permite abrir a interface durante

package/templates/.agents/dashboard/apps/frontend/dist/assets/{CostCalculationGuidePage-CGxeQak4.js → CostCalculationGuidePage-CEDZjl4V.js}

@@ -1 +1 @@
-import{n as e}from"./rolldown-runtime-Bh1tDfsg.js";import{C as t,x as n}from"./cost-monitor-B6aZfo5f.js";var r=e(t(),1),i=n(),a={input:1e5,cache:4e4,output:2e4,reasoning:8e3},o=e=>Math.round(e*1e6)/1e6;function s(e,t,n,r,i){let a=o(Math.max(e.input-e.cache,0)/1e6*t.input*n),s=o(e.cache/1e6*(t.cache??0)*n),c=o(e.output/1e6*t.output*r),l=o(a+s+c);return{usage:e,rates:t,inputMultiplier:n,outputMultiplier:r,inputUsd:a,cacheUsd:s,outputUsd:c,reasoningUsd:0,totalUsd:l,totalBrl:i===null?null:o(l*i)}}function c(e,t){if(t===`standard`)return{input:e.input_per_million_usd,cache:e.cached_input_per_million_usd,output:e.output_per_million_usd};let n=e.priority_input_per_million_usd,r=e.priority_output_per_million_usd;return n==null||r==null?null:{input:n,cache:e.priority_cached_input_per_million_usd??null,output:r}}function l(e,t,n){let r=c(e,t);return r===null?null:s(a,r,1,1,n)}function u(e,t){return e.long_context_threshold_tokens===null?null:s({input:e.long_context_threshold_tokens+5e4,cache:a.cache,output:a.output,reasoning:a.reasoning},c(e,`standard`),e.long_context_input_multiplier,e.long_context_output_multiplier,t)}var d={"pt-BR":{eyebrow:`CONFIGURAÇÕES · TRANSPARÊNCIA`,title:`Como o custo é calculado`,intro:`Do menor grão — uma chamada LLM — até o total: veja quais tokens entram na conta e como modelo, esforço, contexto e velocidade mudam o resultado.`,back:`Voltar ao dashboard`,model:`Modelo usado nos exemplos`,currency:`Moeda`,unavailable:`Indisponível`,millionRate:`por 1 milhão de tokens`,rateTitle:`1. Comece pelas tarifas do modelo`,rateIntro:`O modelo escolhe a tabela de preços. Priority/Fast usa outra tabela quando o catálogo publica essas tarifas.`,standard:`Standard`,priority:`Priority / Fast`,input:`Input`,cache:`Cache`,output:`Output`,reasoning:`Reasoning`,ratio:`Priority ÷ Standard`,included:`Incluído no output (+0)`,noPublishedRate:`Sem tarifa publicada (custo adicional 0)`,conceptsTitle:`2. Entenda cada parte da chamada`,concepts:[[`Chamada LLM`,`É o menor grão do cálculo: uma requisição ao provedor, com um modelo e um tier de serviço.`],[`Input tokens`,`Tudo que entra no modelo: instruções, histórico, contexto, ferramentas e a mensagem atual.`],[`Cache tokens`,`São parte do input, não tokens extras. Primeiro subtraímos o cache do input e depois aplicamos sua tarifa própria.`],[`Output tokens`,`É o total produzido pelo modelo. O custo usa o output total informado pela telemetria.`],[`Reasoning tokens`,`São parte do output. O cálculo atual cobra o output uma vez e mantém reasoningUsd = 0, evitando cobrança dupla.`],[`Modelo e esforço`,`O modelo muda as tarifas. O esforço não tem multiplicador próprio, mas pode gerar mais reasoning, output ou ciclos e aumentar o custo indiretamente.`],[`Multiplicador de contexto`,`Acima do limite do modelo, o multiplicador de input vale também para cache; o multiplicador de output vale para todo o output.`],[`Multiplicador de velocidade`,`É a razão derivada entre o total Priority/Fast e o Standard. Serve para comparar; nunca é multiplicado de novo no total.`]],exampleTitle:`3. Exemplo simples, parcela por parcela`,exampleIntro:`Uma chamada com 100 mil tokens de input, dos quais 40 mil vieram do cache, e 20 mil tokens de output, dos quais 8 mil são reasoning.`,billableInput:`Input faturável`,operation:`Operação`,cost:`Custo`,total:`Total da chamada`,speedMultiplier:`Multiplicador de velocidade derivado`,longTitle:`4. Exemplo com contexto longo`,longIntro:`Para realmente ultrapassar o limite, o exemplo usa o threshold do modelo + 50 mil tokens de input. A conta abaixo continua no tier Standard.`,threshold:`Limite do modelo`,longInput:`Input do exemplo longo`,priorityLong:`Priority/Fast + contexto longo`,priorityLongNote:`Indisponível: sem uma combinação publicada no catálogo, o dashboard não inventa uma tarifa.`,finalNote:`Regra prática: escolha o modelo e o tier, separe input de cache, cobre o output total uma única vez e só então aplique câmbio. Mais esforço só aumenta o custo quando aumenta o uso registrado.`},"en-US":{eyebrow:`SETTINGS · TRANSPARENCY`,title:`How cost is calculated`,intro:`From the smallest grain — one LLM call — to the total: see which tokens enter the calculation and how model, effort, context, and speed affect it.`,back:`Back to dashboard`,model:`Model used in the examples`,currency:`Currency`,unavailable:`Unavailable`,millionRate:`per 1 million tokens`,rateTitle:`1. Start with the model rates`,rateIntro:`The model selects the price table. Priority/Fast uses a different table when the catalog publishes those rates.`,standard:`Standard`,priority:`Priority / Fast`,input:`Input`,cache:`Cache`,output:`Output`,reasoning:`Reasoning`,ratio:`Priority ÷ Standard`,included:`Included in output (+0)`,noPublishedRate:`No published rate (additional cost 0)`,conceptsTitle:`2. Understand each part of the call`,concepts:[[`LLM call`,`The smallest calculation grain: one provider request with one model and one service tier.`],[`Input tokens`,`Everything sent to the model: instructions, history, context, tools, and the current message.`],[`Cache tokens`,`They are part of input, not extra tokens. Cache is subtracted from input first, then its own rate is applied.`],[`Output tokens`,`Everything produced by the model. Cost uses the total output reported by telemetry.`],[`Reasoning tokens`,`They are part of output. Current math charges output once and keeps reasoningUsd = 0, preventing double billing.`],[`Model and effort`,`The model changes rates. Effort has no direct multiplier, but it can produce more reasoning, output, or cycles and indirectly raise cost.`],[`Context multiplier`,`Above the model threshold, the input multiplier also applies to cache; the output multiplier applies to all output.`],[`Speed multiplier`,`It is the derived ratio between Priority/Fast and Standard totals. It is only a comparison and is never multiplied into the total again.`]],exampleTitle:`3. A simple example, component by component`,exampleIntro:`One call with 100 thousand input tokens, including 40 thousand cache tokens, and 20 thousand output tokens, including 8 thousand reasoning tokens.`,billableInput:`Billable input`,operation:`Operation`,cost:`Cost`,total:`Call total`,speedMultiplier:`Derived speed multiplier`,longTitle:`4. Long-context example`,longIntro:`To actually exceed the limit, this example uses the model threshold + 50 thousand input tokens. The calculation remains on Standard tier.`,threshold:`Model threshold`,longInput:`Long-example input`,priorityLong:`Priority/Fast + long context`,priorityLongNote:`Unavailable: without a published combination in the catalog, the dashboard does not invent a rate.`,finalNote:`Rule of thumb: choose model and tier, separate input from cache, charge total output once, then apply FX. Higher effort only raises cost when it raises recorded usage.`}};function f(e,t,n,r){if(t===`BRL`&&n===null)return r;let i=t===`BRL`?e*n:e;return new Intl.NumberFormat(t===`BRL`?`pt-BR`:`en-US`,{style:`currency`,currency:t,minimumFractionDigits:2,maximumFractionDigits:6}).format(i)}function p(e,t,n,r,i){return e==null?i:f(e,t,n,r)}function m(e,t){return new Intl.NumberFormat(t).format(e)}function h({example:e,currency:t,fxRate:n,locale:r,labels:a}){let o=a.unavailable,s=Math.max(e.usage.input-e.usage.cache,0),c=e.inputMultiplier===1?``:` × ${e.inputMultiplier}`,l=e.outputMultiplier===1?``:` × ${e.outputMultiplier}`,u=e.rates.cache??0,d=[{label:a.billableInput,operation:`${m(e.usage.input,r)} − ${m(e.usage.cache,r)} = ${m(s,r)} tokens`,cost:null},{label:a.input,operation:`${m(s,r)} ÷ 1M × ${p(e.rates.input,t,n,o,a.noPublishedRate)}${c}`,cost:e.inputUsd},{label:a.cache,operation:`${m(e.usage.cache,r)} ÷ 1M × ${p(u,t,n,o,a.noPublishedRate)}${c}`,cost:e.cacheUsd},{label:a.output,operation:`${m(e.usage.output,r)} ÷ 1M × ${p(e.rates.output,t,n,o,a.noPublishedRate)}${l}`,cost:e.outputUsd},{label:a.reasoning,operation:`${m(e.usage.reasoning,r)} tokens ⊂ ${a.output} → +0`,cost:e.reasoningUsd}];return(0,i.jsxs)(`div`,{className:`cost-guide-calculation`,children:[(0,i.jsxs)(`div`,{className:`cost-guide-calculation-head`,"aria-hidden":`true`,children:[(0,i.jsx)(`span`,{}),(0,i.jsx)(`span`,{children:a.operation}),(0,i.jsx)(`span`,{children:a.cost})]}),d.map(e=>(0,i.jsxs)(`div`,{className:`cost-guide-calculation-line`,children:[(0,i.jsx)(`strong`,{children:e.label}),(0,i.jsx)(`code`,{children:e.operation}),(0,i.jsx)(`span`,{children:e.cost===null?`—`:f(e.cost,t,n,o)})]},e.label)),(0,i.jsxs)(`div`,{className:`cost-guide-calculation-total`,children:[(0,i.jsx)(`strong`,{children:a.total}),(0,i.jsx)(`span`,{children:f(e.totalUsd,t,n,o)})]})]})}function g({catalog:e,fxRate:t,locale:n,onClose:a}){let o=d[n],[s,c]=(0,r.useState)(e.models[0]?.id??``),[f,g]=(0,r.useState)(`USD`),_=(0,r.useMemo)(()=>e.models.find(e=>e.id===s)??e.models[0]??null,[e.models,s]);if(_===null)return(0,i.jsxs)(`main`,{className:`cost-guide`,children:[(0,i.jsxs)(`button`,{onClick:a,type:`button`,children:[`← `,o.back]}),(0,i.jsx)(`h1`,{children:o.title}),(0,i.jsx)(`p`,{children:o.unavailable})]});let v=t?.rate??null,y=l(_,`standard`,v),b=l(_,`priority`,v),x=u(_,v),S=b!==null&&y.totalUsd>0?b.totalUsd/y.totalUsd:null,C=[{label:o.input,standard:_.input_per_million_usd,priority:_.priority_input_per_million_usd},{label:o.cache,standard:_.cached_input_per_million_usd,priority:_.priority_cached_input_per_million_usd},{label:o.output,standard:_.output_per_million_usd,priority:_.priority_output_per_million_usd}];return(0,i.jsxs)(`main`,{className:`cost-guide`,children:[(0,i.jsxs)(`header`,{children:[(0,i.jsxs)(`div`,{children:[(0,i.jsx)(`small`,{children:o.eyebrow}),(0,i.jsx)(`h1`,{children:o.title}),(0,i.jsx)(`p`,{children:o.intro})]}),(0,i.jsxs)(`button`,{onClick:a,type:`button`,children:[`← `,o.back]})]}),(0,i.jsxs)(`section`,{className:`cost-guide-controls`,"aria-label":o.model,children:[(0,i.jsxs)(`label`,{children:[o.model,(0,i.jsx)(`select`,{value:_.id,onChange:e=>c(e.target.value),children:e.models.map(e=>(0,i.jsx)(`option`,{value:e.id,children:e.id},e.id))})]}),(0,i.jsxs)(`label`,{children:[o.currency,(0,i.jsxs)(`select`,{value:f,onChange:e=>g(e.target.value),children:[(0,i.jsx)(`option`,{value:`USD`,children:`USD`}),(0,i.jsxs)(`option`,{value:`BRL`,children:[`BRL `,v===null?`(${o.unavailable})`:`(1 USD = ${v.toFixed(4)} BRL)`]})]})]})]}),(0,i.jsxs)(`section`,{className:`cost-guide-section`,children:[(0,i.jsxs)(`div`,{className:`cost-guide-section-heading`,children:[(0,i.jsx)(`h2`,{children:o.rateTitle}),(0,i.jsx)(`p`,{children:o.rateIntro})]}),(0,i.jsx)(`div`,{className:`cost-guide-table-wrap`,children:(0,i.jsxs)(`table`,{children:[(0,i.jsxs)(`caption`,{children:[_.id,` · `,o.millionRate]}),(0,i.jsx)(`thead`,{children:(0,i.jsxs)(`tr`,{children:[(0,i.jsx)(`th`,{scope:`col`}),(0,i.jsx)(`th`,{scope:`col`,children:o.standard}),(0,i.jsx)(`th`,{scope:`col`,children:o.priority}),(0,i.jsx)(`th`,{scope:`col`,children:o.ratio})]})}),(0,i.jsxs)(`tbody`,{children:[C.map(e=>(0,i.jsxs)(`tr`,{children:[(0,i.jsx)(`th`,{scope:`row`,children:e.label}),(0,i.jsx)(`td`,{children:p(e.standard,f,v,o.unavailable,o.noPublishedRate)}),(0,i.jsx)(`td`,{children:p(e.priority,f,v,o.unavailable,o.noPublishedRate)}),(0,i.jsx)(`td`,{children:e.standard!==null&&e.standard>0&&e.priority!==null&&e.priority!==void 0?`${(e.priority/e.standard).toFixed(2)}×`:o.unavailable})]},e.label)),(0,i.jsxs)(`tr`,{children:[(0,i.jsx)(`th`,{scope:`row`,children:o.reasoning}),(0,i.jsx)(`td`,{colSpan:3,children:o.included})]})]})]})})]}),(0,i.jsxs)(`section`,{className:`cost-guide-section`,children:[(0,i.jsx)(`div`,{className:`cost-guide-section-heading`,children:(0,i.jsx)(`h2`,{children:o.conceptsTitle})}),(0,i.jsx)(`div`,{className:`cost-guide-blocks`,children:o.concepts.map(([e,t])=>(0,i.jsxs)(`article`,{children:[(0,i.jsx)(`strong`,{children:e}),(0,i.jsx)(`p`,{children:t})]},e))})]}),(0,i.jsxs)(`section`,{className:`cost-guide-section`,children:[(0,i.jsxs)(`div`,{className:`cost-guide-section-heading`,children:[(0,i.jsx)(`h2`,{children:o.exampleTitle}),(0,i.jsx)(`p`,{children:o.exampleIntro})]}),(0,i.jsxs)(`div`,{className:`cost-guide-example-grid`,children:[(0,i.jsxs)(`article`,{className:`cost-guide-example-card`,children:[(0,i.jsx)(`h3`,{children:o.standard}),(0,i.jsx)(h,{example:y,currency:f,fxRate:v,locale:n,labels:o})]}),(0,i.jsxs)(`article`,{className:`cost-guide-example-card`,children:[(0,i.jsx)(`h3`,{children:o.priority}),b===null?(0,i.jsx)(`p`,{className:`cost-guide-unavailable`,children:o.unavailable}):(0,i.jsx)(h,{example:b,currency:f,fxRate:v,locale:n,labels:o})]})]}),(0,i.jsxs)(`div`,{className:`cost-guide-speed`,children:[(0,i.jsx)(`span`,{children:o.speedMultiplier}),(0,i.jsx)(`strong`,{children:S===null?o.unavailable:`${S.toFixed(2)}×`}),(0,i.jsxs)(`code`,{children:[o.priority,` `,o.total,` ÷ `,o.standard,` `,o.total]})]})]}),(0,i.jsxs)(`section`,{className:`cost-guide-section`,children:[(0,i.jsxs)(`div`,{className:`cost-guide-section-heading`,children:[(0,i.jsx)(`h2`,{children:o.longTitle}),(0,i.jsx)(`p`,{children:o.longIntro})]}),(0,i.jsxs)(`div`,{className:`cost-guide-context-summary`,children:[(0,i.jsxs)(`span`,{children:[o.threshold,`:`,` `,(0,i.jsx)(`strong`,{children:_.long_context_threshold_tokens===null?o.unavailable:m(_.long_context_threshold_tokens,n)})]}),(0,i.jsxs)(`span`,{children:[o.longInput,`:`,` `,(0,i.jsx)(`strong`,{children:x===null?o.unavailable:m(x.usage.input,n)})]}),(0,i.jsxs)(`span`,{children:[`Input/cache: `,(0,i.jsxs)(`strong`,{children:[_.long_context_input_multiplier,`×`]})]}),(0,i.jsxs)(`span`,{children:[`Output: `,(0,i.jsxs)(`strong`,{children:[_.long_context_output_multiplier,`×`]})]})]}),x===null?(0,i.jsx)(`p`,{className:`cost-guide-unavailable`,children:o.unavailable}):(0,i.jsx)(h,{example:x,currency:f,fxRate:v,locale:n,labels:o}),(0,i.jsxs)(`div`,{className:`cost-guide-unavailable-combination`,children:[(0,i.jsxs)(`strong`,{children:[o.priorityLong,`: `,o.unavailable]}),(0,i.jsx)(`p`,{children:o.priorityLongNote})]})]}),(0,i.jsx)(`aside`,{className:`cost-guide-final-note`,children:o.finalNote})]})}export{g as CostCalculationGuidePage};
+import{n as e}from"./rolldown-runtime-Bh1tDfsg.js";import{E as t,w as n}from"./cost-monitor-06__WZco.js";var r=e(t(),1),i=n(),a={input:1e5,cache:4e4,output:2e4,reasoning:8e3},o=e=>Math.round(e*1e6)/1e6;function s(e,t,n,r,i){let a=o(Math.max(e.input-e.cache,0)/1e6*t.input*n),s=o(e.cache/1e6*(t.cache??0)*n),c=o(e.output/1e6*t.output*r),l=o(a+s+c);return{usage:e,rates:t,inputMultiplier:n,outputMultiplier:r,inputUsd:a,cacheUsd:s,outputUsd:c,reasoningUsd:0,totalUsd:l,totalBrl:i===null?null:o(l*i)}}function c(e,t){if(t===`standard`)return{input:e.input_per_million_usd,cache:e.cached_input_per_million_usd,output:e.output_per_million_usd};let n=e.priority_input_per_million_usd,r=e.priority_output_per_million_usd;return n==null||r==null?null:{input:n,cache:e.priority_cached_input_per_million_usd??null,output:r}}function l(e,t,n){let r=c(e,t);return r===null?null:s(a,r,1,1,n)}function u(e,t){return e.long_context_threshold_tokens===null?null:s({input:e.long_context_threshold_tokens+5e4,cache:a.cache,output:a.output,reasoning:a.reasoning},c(e,`standard`),e.long_context_input_multiplier,e.long_context_output_multiplier,t)}var d={"pt-BR":{eyebrow:`CONFIGURAÇÕES · TRANSPARÊNCIA`,title:`Como o custo é calculado`,intro:`Do menor grão — uma chamada LLM — até o total: veja quais tokens entram na conta e como modelo, esforço, contexto e velocidade mudam o resultado.`,back:`Voltar ao dashboard`,model:`Modelo usado nos exemplos`,currency:`Moeda`,unavailable:`Indisponível`,millionRate:`por 1 milhão de tokens`,rateTitle:`1. Comece pelas tarifas do modelo`,rateIntro:`O modelo escolhe a tabela de preços. Priority/Fast usa outra tabela quando o catálogo publica essas tarifas.`,standard:`Standard`,priority:`Priority / Fast`,input:`Input`,cache:`Cache`,output:`Output`,reasoning:`Reasoning`,ratio:`Priority ÷ Standard`,included:`Incluído no output (+0)`,noPublishedRate:`Sem tarifa publicada (custo adicional 0)`,conceptsTitle:`2. Entenda cada parte da chamada`,concepts:[[`Chamada LLM`,`É o menor grão do cálculo: uma requisição ao provedor, com um modelo e um tier de serviço.`],[`Input tokens`,`Tudo que entra no modelo: instruções, histórico, contexto, ferramentas e a mensagem atual.`],[`Cache tokens`,`São parte do input, não tokens extras. Primeiro subtraímos o cache do input e depois aplicamos sua tarifa própria.`],[`Output tokens`,`É o total produzido pelo modelo. O custo usa o output total informado pela telemetria.`],[`Reasoning tokens`,`São parte do output. O cálculo atual cobra o output uma vez e mantém reasoningUsd = 0, evitando cobrança dupla.`],[`Modelo e esforço`,`O modelo muda as tarifas. O esforço não tem multiplicador próprio, mas pode gerar mais reasoning, output ou ciclos e aumentar o custo indiretamente.`],[`Multiplicador de contexto`,`Acima do limite do modelo, o multiplicador de input vale também para cache; o multiplicador de output vale para todo o output.`],[`Multiplicador de velocidade`,`É a razão derivada entre o total Priority/Fast e o Standard. Serve para comparar; nunca é multiplicado de novo no total.`]],exampleTitle:`3. Exemplo simples, parcela por parcela`,exampleIntro:`Uma chamada com 100 mil tokens de input, dos quais 40 mil vieram do cache, e 20 mil tokens de output, dos quais 8 mil são reasoning.`,billableInput:`Input faturável`,operation:`Operação`,cost:`Custo`,total:`Total da chamada`,speedMultiplier:`Multiplicador de velocidade derivado`,longTitle:`4. Exemplo com contexto longo`,longIntro:`Para realmente ultrapassar o limite, o exemplo usa o threshold do modelo + 50 mil tokens de input. A conta abaixo continua no tier Standard.`,threshold:`Limite do modelo`,longInput:`Input do exemplo longo`,priorityLong:`Priority/Fast + contexto longo`,priorityLongNote:`Indisponível: sem uma combinação publicada no catálogo, o dashboard não inventa uma tarifa.`,finalNote:`Regra prática: escolha o modelo e o tier, separe input de cache, cobre o output total uma única vez e só então aplique câmbio. Mais esforço só aumenta o custo quando aumenta o uso registrado.`},"en-US":{eyebrow:`SETTINGS · TRANSPARENCY`,title:`How cost is calculated`,intro:`From the smallest grain — one LLM call — to the total: see which tokens enter the calculation and how model, effort, context, and speed affect it.`,back:`Back to dashboard`,model:`Model used in the examples`,currency:`Currency`,unavailable:`Unavailable`,millionRate:`per 1 million tokens`,rateTitle:`1. Start with the model rates`,rateIntro:`The model selects the price table. Priority/Fast uses a different table when the catalog publishes those rates.`,standard:`Standard`,priority:`Priority / Fast`,input:`Input`,cache:`Cache`,output:`Output`,reasoning:`Reasoning`,ratio:`Priority ÷ Standard`,included:`Included in output (+0)`,noPublishedRate:`No published rate (additional cost 0)`,conceptsTitle:`2. Understand each part of the call`,concepts:[[`LLM call`,`The smallest calculation grain: one provider request with one model and one service tier.`],[`Input tokens`,`Everything sent to the model: instructions, history, context, tools, and the current message.`],[`Cache tokens`,`They are part of input, not extra tokens. Cache is subtracted from input first, then its own rate is applied.`],[`Output tokens`,`Everything produced by the model. Cost uses the total output reported by telemetry.`],[`Reasoning tokens`,`They are part of output. Current math charges output once and keeps reasoningUsd = 0, preventing double billing.`],[`Model and effort`,`The model changes rates. Effort has no direct multiplier, but it can produce more reasoning, output, or cycles and indirectly raise cost.`],[`Context multiplier`,`Above the model threshold, the input multiplier also applies to cache; the output multiplier applies to all output.`],[`Speed multiplier`,`It is the derived ratio between Priority/Fast and Standard totals. It is only a comparison and is never multiplied into the total again.`]],exampleTitle:`3. A simple example, component by component`,exampleIntro:`One call with 100 thousand input tokens, including 40 thousand cache tokens, and 20 thousand output tokens, including 8 thousand reasoning tokens.`,billableInput:`Billable input`,operation:`Operation`,cost:`Cost`,total:`Call total`,speedMultiplier:`Derived speed multiplier`,longTitle:`4. Long-context example`,longIntro:`To actually exceed the limit, this example uses the model threshold + 50 thousand input tokens. The calculation remains on Standard tier.`,threshold:`Model threshold`,longInput:`Long-example input`,priorityLong:`Priority/Fast + long context`,priorityLongNote:`Unavailable: without a published combination in the catalog, the dashboard does not invent a rate.`,finalNote:`Rule of thumb: choose model and tier, separate input from cache, charge total output once, then apply FX. Higher effort only raises cost when it raises recorded usage.`}};function f(e,t,n,r){if(t===`BRL`&&n===null)return r;let i=t===`BRL`?e*n:e;return new Intl.NumberFormat(t===`BRL`?`pt-BR`:`en-US`,{style:`currency`,currency:t,minimumFractionDigits:2,maximumFractionDigits:6}).format(i)}function p(e,t,n,r,i){return e==null?i:f(e,t,n,r)}function m(e,t){return new Intl.NumberFormat(t).format(e)}function h({example:e,currency:t,fxRate:n,locale:r,labels:a}){let o=a.unavailable,s=Math.max(e.usage.input-e.usage.cache,0),c=e.inputMultiplier===1?``:` × ${e.inputMultiplier}`,l=e.outputMultiplier===1?``:` × ${e.outputMultiplier}`,u=e.rates.cache??0,d=[{label:a.billableInput,operation:`${m(e.usage.input,r)} − ${m(e.usage.cache,r)} = ${m(s,r)} tokens`,cost:null},{label:a.input,operation:`${m(s,r)} ÷ 1M × ${p(e.rates.input,t,n,o,a.noPublishedRate)}${c}`,cost:e.inputUsd},{label:a.cache,operation:`${m(e.usage.cache,r)} ÷ 1M × ${p(u,t,n,o,a.noPublishedRate)}${c}`,cost:e.cacheUsd},{label:a.output,operation:`${m(e.usage.output,r)} ÷ 1M × ${p(e.rates.output,t,n,o,a.noPublishedRate)}${l}`,cost:e.outputUsd},{label:a.reasoning,operation:`${m(e.usage.reasoning,r)} tokens ⊂ ${a.output} → +0`,cost:e.reasoningUsd}];return(0,i.jsxs)(`div`,{className:`cost-guide-calculation`,children:[(0,i.jsxs)(`div`,{className:`cost-guide-calculation-head`,"aria-hidden":`true`,children:[(0,i.jsx)(`span`,{}),(0,i.jsx)(`span`,{children:a.operation}),(0,i.jsx)(`span`,{children:a.cost})]}),d.map(e=>(0,i.jsxs)(`div`,{className:`cost-guide-calculation-line`,children:[(0,i.jsx)(`strong`,{children:e.label}),(0,i.jsx)(`code`,{children:e.operation}),(0,i.jsx)(`span`,{children:e.cost===null?`—`:f(e.cost,t,n,o)})]},e.label)),(0,i.jsxs)(`div`,{className:`cost-guide-calculation-total`,children:[(0,i.jsx)(`strong`,{children:a.total}),(0,i.jsx)(`span`,{children:f(e.totalUsd,t,n,o)})]})]})}function g({catalog:e,fxRate:t,locale:n,onClose:a}){let o=d[n],[s,c]=(0,r.useState)(e.models[0]?.id??``),[f,g]=(0,r.useState)(`USD`),_=(0,r.useMemo)(()=>e.models.find(e=>e.id===s)??e.models[0]??null,[e.models,s]);if(_===null)return(0,i.jsxs)(`main`,{className:`cost-guide`,children:[(0,i.jsxs)(`button`,{onClick:a,type:`button`,children:[`← `,o.back]}),(0,i.jsx)(`h1`,{children:o.title}),(0,i.jsx)(`p`,{children:o.unavailable})]});let v=t?.rate??null,y=l(_,`standard`,v),b=l(_,`priority`,v),x=u(_,v),S=b!==null&&y.totalUsd>0?b.totalUsd/y.totalUsd:null,C=[{label:o.input,standard:_.input_per_million_usd,priority:_.priority_input_per_million_usd},{label:o.cache,standard:_.cached_input_per_million_usd,priority:_.priority_cached_input_per_million_usd},{label:o.output,standard:_.output_per_million_usd,priority:_.priority_output_per_million_usd}];return(0,i.jsxs)(`main`,{className:`cost-guide`,children:[(0,i.jsxs)(`header`,{children:[(0,i.jsxs)(`div`,{children:[(0,i.jsx)(`small`,{children:o.eyebrow}),(0,i.jsx)(`h1`,{children:o.title}),(0,i.jsx)(`p`,{children:o.intro})]}),(0,i.jsxs)(`button`,{onClick:a,type:`button`,children:[`← `,o.back]})]}),(0,i.jsxs)(`section`,{className:`cost-guide-controls`,"aria-label":o.model,children:[(0,i.jsxs)(`label`,{children:[o.model,(0,i.jsx)(`select`,{value:_.id,onChange:e=>c(e.target.value),children:e.models.map(e=>(0,i.jsx)(`option`,{value:e.id,children:e.id},e.id))})]}),(0,i.jsxs)(`label`,{children:[o.currency,(0,i.jsxs)(`select`,{value:f,onChange:e=>g(e.target.value),children:[(0,i.jsx)(`option`,{value:`USD`,children:`USD`}),(0,i.jsxs)(`option`,{value:`BRL`,children:[`BRL `,v===null?`(${o.unavailable})`:`(1 USD = ${v.toFixed(4)} BRL)`]})]})]})]}),(0,i.jsxs)(`section`,{className:`cost-guide-section`,children:[(0,i.jsxs)(`div`,{className:`cost-guide-section-heading`,children:[(0,i.jsx)(`h2`,{children:o.rateTitle}),(0,i.jsx)(`p`,{children:o.rateIntro})]}),(0,i.jsx)(`div`,{className:`cost-guide-table-wrap`,children:(0,i.jsxs)(`table`,{children:[(0,i.jsxs)(`caption`,{children:[_.id,` · `,o.millionRate]}),(0,i.jsx)(`thead`,{children:(0,i.jsxs)(`tr`,{children:[(0,i.jsx)(`th`,{scope:`col`}),(0,i.jsx)(`th`,{scope:`col`,children:o.standard}),(0,i.jsx)(`th`,{scope:`col`,children:o.priority}),(0,i.jsx)(`th`,{scope:`col`,children:o.ratio})]})}),(0,i.jsxs)(`tbody`,{children:[C.map(e=>(0,i.jsxs)(`tr`,{children:[(0,i.jsx)(`th`,{scope:`row`,children:e.label}),(0,i.jsx)(`td`,{children:p(e.standard,f,v,o.unavailable,o.noPublishedRate)}),(0,i.jsx)(`td`,{children:p(e.priority,f,v,o.unavailable,o.noPublishedRate)}),(0,i.jsx)(`td`,{children:e.standard!==null&&e.standard>0&&e.priority!==null&&e.priority!==void 0?`${(e.priority/e.standard).toFixed(2)}×`:o.unavailable})]},e.label)),(0,i.jsxs)(`tr`,{children:[(0,i.jsx)(`th`,{scope:`row`,children:o.reasoning}),(0,i.jsx)(`td`,{colSpan:3,children:o.included})]})]})]})})]}),(0,i.jsxs)(`section`,{className:`cost-guide-section`,children:[(0,i.jsx)(`div`,{className:`cost-guide-section-heading`,children:(0,i.jsx)(`h2`,{children:o.conceptsTitle})}),(0,i.jsx)(`div`,{className:`cost-guide-blocks`,children:o.concepts.map(([e,t])=>(0,i.jsxs)(`article`,{children:[(0,i.jsx)(`strong`,{children:e}),(0,i.jsx)(`p`,{children:t})]},e))})]}),(0,i.jsxs)(`section`,{className:`cost-guide-section`,children:[(0,i.jsxs)(`div`,{className:`cost-guide-section-heading`,children:[(0,i.jsx)(`h2`,{children:o.exampleTitle}),(0,i.jsx)(`p`,{children:o.exampleIntro})]}),(0,i.jsxs)(`div`,{className:`cost-guide-example-grid`,children:[(0,i.jsxs)(`article`,{className:`cost-guide-example-card`,children:[(0,i.jsx)(`h3`,{children:o.standard}),(0,i.jsx)(h,{example:y,currency:f,fxRate:v,locale:n,labels:o})]}),(0,i.jsxs)(`article`,{className:`cost-guide-example-card`,children:[(0,i.jsx)(`h3`,{children:o.priority}),b===null?(0,i.jsx)(`p`,{className:`cost-guide-unavailable`,children:o.unavailable}):(0,i.jsx)(h,{example:b,currency:f,fxRate:v,locale:n,labels:o})]})]}),(0,i.jsxs)(`div`,{className:`cost-guide-speed`,children:[(0,i.jsx)(`span`,{children:o.speedMultiplier}),(0,i.jsx)(`strong`,{children:S===null?o.unavailable:`${S.toFixed(2)}×`}),(0,i.jsxs)(`code`,{children:[o.priority,` `,o.total,` ÷ `,o.standard,` `,o.total]})]})]}),(0,i.jsxs)(`section`,{className:`cost-guide-section`,children:[(0,i.jsxs)(`div`,{className:`cost-guide-section-heading`,children:[(0,i.jsx)(`h2`,{children:o.longTitle}),(0,i.jsx)(`p`,{children:o.longIntro})]}),(0,i.jsxs)(`div`,{className:`cost-guide-context-summary`,children:[(0,i.jsxs)(`span`,{children:[o.threshold,`:`,` `,(0,i.jsx)(`strong`,{children:_.long_context_threshold_tokens===null?o.unavailable:m(_.long_context_threshold_tokens,n)})]}),(0,i.jsxs)(`span`,{children:[o.longInput,`:`,` `,(0,i.jsx)(`strong`,{children:x===null?o.unavailable:m(x.usage.input,n)})]}),(0,i.jsxs)(`span`,{children:[`Input/cache: `,(0,i.jsxs)(`strong`,{children:[_.long_context_input_multiplier,`×`]})]}),(0,i.jsxs)(`span`,{children:[`Output: `,(0,i.jsxs)(`strong`,{children:[_.long_context_output_multiplier,`×`]})]})]}),x===null?(0,i.jsx)(`p`,{className:`cost-guide-unavailable`,children:o.unavailable}):(0,i.jsx)(h,{example:x,currency:f,fxRate:v,locale:n,labels:o}),(0,i.jsxs)(`div`,{className:`cost-guide-unavailable-combination`,children:[(0,i.jsxs)(`strong`,{children:[o.priorityLong,`: `,o.unavailable]}),(0,i.jsx)(`p`,{children:o.priorityLongNote})]})]}),(0,i.jsx)(`aside`,{className:`cost-guide-final-note`,children:o.finalNote})]})}export{g as CostCalculationGuidePage};