feat: 加入预估价格，优化移动UI

Co-authored-by: Copilot <copilot@github.com>

index.html

@@ -7,7 +7,7 @@
       name="description"
       content="一个本地优先的 OpenAI-compatible LLM 聊天面板。"
     />
-    <meta name="theme-color" content="#111111" />
+    <meta name="theme-color" content="#ffffff" />
     <meta name="mobile-web-app-capable" content="yes" />
     <meta name="apple-mobile-web-app-capable" content="yes" />
     <meta name="apple-mobile-web-app-title" content="LLM Chat" />

public/manifest.webmanifest

@@ -9,7 +9,7 @@
   "display": "standalone",
   "orientation": "any",
   "background_color": "#ffffff",
-  "theme_color": "#111111",
+  "theme_color": "#ffffff",
   "icons": [
     {
       "src": "/icons/icon-192.png",

src/App.tsx

@@ -17,7 +17,6 @@ import { PromptsView } from "@/components/app/prompts-view"
 import { AppSidebar } from "@/components/app/sidebar"
 import { SettingsView } from "@/components/app/settings-view"
 import { useTheme } from "@/components/theme-provider"
-import { Badge } from "@/components/ui/badge"
 import { Button } from "@/components/ui/button"
 import {
   createDefaultData,
@@ -35,6 +34,7 @@ import {
   fetchModels,
   trimContextMessages,
 } from "@/lib/openai"
+import { calculateUsageCost } from "@/lib/pricing"
 import {
   clearStoredAppData,
   loadAppData,
@@ -91,28 +91,18 @@ const VIEW_ITEMS: {
   { key: "about", label: "关于", icon: Info },
 ]
 
-const STATUS_TEXT: Record<RequestState | "done", string> = {
+const DATA_SAVE_DELAY_MS = 800
+const DEEP_SEARCH_BATCH_SIZE = 32
+const DEEP_SEARCH_SNIPPET_LIMIT = 3
+
+const STATUS_TEXT: Record<RequestState, string> = {
   idle: "就绪",
   requesting: "请求中",
   generating: "生成中",
   failed: "失败",
   stopped: "已停止",
-  done: "完成",
 }
 
-const STATUS_TONE: Record<RequestState | "done", Notice["tone"]> = {
-  idle: "default",
-  requesting: "warning",
-  generating: "warning",
-  failed: "danger",
-  stopped: "default",
-  done: "success",
-}
-
-const DATA_SAVE_DELAY_MS = 800
-const DEEP_SEARCH_BATCH_SIZE = 32
-const DEEP_SEARCH_SNIPPET_LIMIT = 3
-
 function isBusy(status: RequestState) {
   return status === "requesting" || status === "generating"
 }
@@ -167,6 +157,42 @@ function yieldToBrowser() {
   return new Promise<void>((resolve) => window.setTimeout(resolve, 0))
 }
 
+function getStatusLight(
+  configIsValid: boolean,
+  status: RequestState,
+  errors: string[]
+) {
+  if (!configIsValid) {
+    return {
+      label: `配置待完善${errors.length ? `：${errors.join("，")}` : ""}`,
+      className: "bg-amber-500 shadow-amber-500/50",
+      pulse: false,
+    }
+  }
+
+  if (status === "failed") {
+    return {
+      label: "配置有效，当前会话失败",
+      className: "bg-destructive shadow-destructive/50",
+      pulse: false,
+    }
+  }
+
+  if (status === "requesting" || status === "generating") {
+    return {
+      label: `配置有效，${STATUS_TEXT[status]}`,
+      className: "bg-amber-500 shadow-amber-500/50",
+      pulse: true,
+    }
+  }
+
+  return {
+    label: `配置有效，${STATUS_TEXT[status]}`,
+    className: "bg-emerald-500 shadow-emerald-500/50",
+    pulse: false,
+  }
+}
+
 function updateSessionInData(
   source: AppData,
   sessionId: string,
@@ -864,6 +890,7 @@ function AppContent() {
       const elapsedMs = performance.now() - startedAt
       const content = result.content || accumulated
       const reasoningContent = result.reasoningContent || accumulatedReasoning
+      const cost = calculateUsageCost(result.usage, config.model?.pricing)
       const toolCalls = result.toolCalls?.map((toolCall) => ({
         ...toolCall,
         status: "pending" as const,
@@ -887,6 +914,7 @@ function AppContent() {
               toolCalls,
               elapsedMs,
               usage: result.usage,
+              cost,
               updatedAt: nowIso(),
             })
           )
@@ -1652,6 +1680,11 @@ function AppContent() {
   }
 
   const isCurrentBusy = isBusy(currentSession.status)
+  const statusLight = getStatusLight(
+    configIsValid,
+    currentSession.status,
+    runtimeConfig.errors
+  )
 
   return (
     <div
@@ -1710,16 +1743,22 @@ function AppContent() {
               })}
             </nav>
           </div>
-          <div className="flex items-center gap-2">
-            <Badge tone={configIsValid ? "success" : "warning"}>
-              {configIsValid ? "配置有效" : "待配置"}
-            </Badge>
-            <Badge tone={STATUS_TONE[currentSession.status]}>
-              {STATUS_TEXT[currentSession.status]}
-            </Badge>
-          </div>
         </header>
 
+        <div
+          className="fixed right-3 top-3 z-50 flex size-5 items-center justify-center rounded-full border bg-background/80 shadow-sm backdrop-blur"
+          title={statusLight.label}
+          aria-label={statusLight.label}
+        >
+          <span
+            className={cn(
+              "block size-2.5 rounded-full shadow-[0_0_10px]",
+              statusLight.className,
+              statusLight.pulse && "animate-pulse"
+            )}
+          />
+        </div>
+
         {notice ? (
           <div className="notice-wrap pointer-events-none absolute left-0 right-0 top-14 z-20 px-4 bg-background">
             <div
@@ -1758,6 +1797,7 @@ function AppContent() {
             editingMessageId={editingMessageId}
             editingContent={editingContent}
             showPromptEditor={showPromptEditor}
+            headerOpen={data.settings.chatHeaderOpen}
             quickConfigOpen={data.settings.chatQuickConfigOpen}
             isBusy={isCurrentBusy}
             configIsValid={configIsValid}
@@ -1778,6 +1818,9 @@ function AppContent() {
             onServiceChange={applySessionService}
             onModelChange={applySessionModel}
             onPromptChange={applySessionPrompt}
+            onHeaderOpenChange={(open) =>
+              updateSettings({ chatHeaderOpen: open })
+            }
             onQuickConfigOpenChange={(open) =>
               updateSettings({ chatQuickConfigOpen: open })
             }
@@ -1789,6 +1832,16 @@ function AppContent() {
                 updatedAt: nowIso(),
               }))
             }
+            onReasoningEffortChange={(reasoningEffort) => {
+              const modelId = runtimeConfig.model?.id
+
+              if (modelId) {
+                updateModelParameter(modelId, "reasoningEffort", reasoningEffort)
+                return
+              }
+
+              updateSettings({ reasoningEffort })
+            }}
             onOpenSettings={() => setActiveView("settings")}
           />
         ) : null}

src/components/app/chat-view.tsx

@@ -26,6 +26,7 @@ import {
   ASK_QUESTIONS_TOOL_NAME,
   parseAskQuestionsArguments,
 } from "@/lib/tools"
+import { formatMessageCost } from "@/lib/pricing"
 import { cn } from "@/lib/utils"
 import type {
   AskQuestionsAnswer,
@@ -36,6 +37,7 @@ import type {
   ModelConfig,
   PromptTemplate,
   RequestState,
+  ReasoningEffort,
   RuntimeConfig,
   ServiceConfig,
 } from "@/types"
@@ -49,6 +51,14 @@ const STATUS_TEXT: Record<RequestState | "done", string> = {
   done: "完成",
 }
 
+const REASONING_EFFORT_OPTIONS: { value: ReasoningEffort; label: string }[] = [
+  { value: "default", label: "默认" },
+  { value: "minimal", label: "minimal" },
+  { value: "low", label: "low" },
+  { value: "medium", label: "medium" },
+  { value: "high", label: "high" },
+]
+
 function formatMessageTime(value: string) {
   return new Intl.DateTimeFormat("zh-CN", {
     hour: "2-digit",
@@ -73,19 +83,22 @@ function formatUsage(message: ChatMessage) {
     return ""
   }
 
+  const cachedPromptTokens =
+    message.usage.promptCacheHitTokens ?? message.usage.cachedPromptTokens
   const parts = [
     message.usage.promptTokens !== undefined
-      ? `输入 ${message.usage.promptTokens}`
+      ? `In: ${message.usage.promptTokens}`
       : "",
+    cachedPromptTokens !== undefined ? `(Cached: ${cachedPromptTokens})` : "",
     message.usage.completionTokens !== undefined
-      ? `输出 ${message.usage.completionTokens}`
+      ? `Out: ${message.usage.completionTokens}`
       : "",
     message.usage.totalTokens !== undefined
-      ? `总计 ${message.usage.totalTokens}`
+      ? `All: ${message.usage.totalTokens}`
       : "",
   ].filter(Boolean)
 
-  return parts.length ? parts.join(" / ") : ""
+  return parts.length ? parts.join(" ") : ""
 }
 
 function getMessageCopyContent(message: ChatMessage) {
@@ -111,6 +124,7 @@ type ChatViewProps = {
   editingMessageId: string | null
   editingContent: string
   showPromptEditor: boolean
+  headerOpen: boolean
   quickConfigOpen: boolean
   isBusy: boolean
   configIsValid: boolean
@@ -133,9 +147,11 @@ type ChatViewProps = {
   onServiceChange: (serviceId: string) => void
   onModelChange: (modelId: string) => void
   onPromptChange: (promptId: string) => void
+  onHeaderOpenChange: (open: boolean) => void
   onQuickConfigOpenChange: (open: boolean) => void
   onPromptEditorToggle: () => void
   onSystemPromptChange: (value: string) => void
+  onReasoningEffortChange: (value: ReasoningEffort) => void
   onOpenSettings: () => void
 }
 
@@ -149,6 +165,7 @@ export function ChatView({
   editingMessageId,
   editingContent,
   showPromptEditor,
+  headerOpen,
   quickConfigOpen,
   isBusy,
   configIsValid,
@@ -167,13 +184,21 @@ export function ChatView({
   onServiceChange,
   onModelChange,
   onPromptChange,
+  onHeaderOpenChange,
   onQuickConfigOpenChange,
   onPromptEditorToggle,
   onSystemPromptChange,
+  onReasoningEffortChange,
   onOpenSettings,
 }: ChatViewProps) {
   const canSend = chatInput.trim().length > 0 && !isBusy
   const messagesScrollRef = React.useRef<HTMLDivElement>(null)
+  const reasoningMenuRef = React.useRef<HTMLDivElement>(null)
+  const [reasoningMenuOpen, setReasoningMenuOpen] = React.useState(false)
+  const selectedReasoningLabel =
+    REASONING_EFFORT_OPTIONS.find(
+      (option) => option.value === runtimeConfig.parameters.reasoningEffort
+    )?.label ?? runtimeConfig.parameters.reasoningEffort
 
   React.useEffect(() => {
     const element = messagesScrollRef.current
@@ -186,35 +211,105 @@ export function ChatView({
     element.scrollTop = element.scrollHeight
   }, [currentSession.id])
 
+  React.useEffect(() => {
+    if (!reasoningMenuOpen) {
+      return undefined
+    }
+
+    const handlePointerDown = (event: PointerEvent) => {
+      if (
+        reasoningMenuRef.current &&
+        !reasoningMenuRef.current.contains(event.target as Node)
+      ) {
+        setReasoningMenuOpen(false)
+      }
+    }
+
+    const handleKeyDown = (event: KeyboardEvent) => {
+      if (event.key === "Escape") {
+        setReasoningMenuOpen(false)
+      }
+    }
+
+    document.addEventListener("pointerdown", handlePointerDown)
+    document.addEventListener("keydown", handleKeyDown)
+
+    return () => {
+      document.removeEventListener("pointerdown", handlePointerDown)
+      document.removeEventListener("keydown", handleKeyDown)
+    }
+  }, [reasoningMenuOpen])
+
   return (
     <section className="content-view chat-view">
-      <div className="chat-header border-b px-4 py-3">
+      {headerOpen ? (
-        <div className="min-w-0 flex-1">
+        <div className="chat-header border-b px-4 py-3">
-          <div className="flex flex-wrap items-center gap-2">
+          <div className="min-w-0 flex-1">
-            <h1 className="truncate text-base font-semibold">
+            <div className="flex flex-wrap items-center gap-2">
-              {currentSession.title}
+              <h1 className="truncate text-base font-semibold">
-            </h1>
+                {currentSession.title}
-            {currentSession.temporary ? <Badge tone="muted">临时</Badge> : null}
+              </h1>
+              {currentSession.temporary ? (
+                <Badge tone="muted">临时</Badge>
+              ) : null}
+            </div>
+            <div className="mt-1 flex flex-wrap items-center gap-2 text-xs text-muted-foreground">
+              <span>{runtimeConfig.service?.name || "未选择服务"}</span>
+              <span>{runtimeConfig.modelName || "未选择模型"}</span>
+              <span>{currentSession.messages.length} 条消息</span>
+            </div>
           </div>
-          <div className="mt-1 flex flex-wrap items-center gap-2 text-xs text-muted-foreground">
+          <div className="chat-header-actions flex flex-wrap items-center gap-1">
-            <span>{runtimeConfig.service?.name || "未选择服务"}</span>
+            <Button
-            <span>{runtimeConfig.modelName || "未选择模型"}</span>
+              type="button"
-            <span>{currentSession.messages.length} 条消息</span>
+              variant="ghost"
+              size="icon-sm"
+              title={quickConfigOpen ? "收起配置" : "展开配置"}
+              aria-label={quickConfigOpen ? "收起配置" : "展开配置"}
+              aria-expanded={quickConfigOpen}
+              onClick={() => onQuickConfigOpenChange(!quickConfigOpen)}
+            >
+              <Settings />
+            </Button>
+            <Button
+              type="button"
+              variant="ghost"
+              size="icon-sm"
+              title="收起聊天信息"
+              aria-label="收起聊天信息"
+              aria-expanded={headerOpen}
+              onClick={() => onHeaderOpenChange(false)}
+            >
+              <ChevronUp />
+            </Button>
           </div>
         </div>
-        <div className="chat-header-actions flex flex-wrap items-center gap-2">
+      ) : (
+        <div className="flex justify-end gap-1 border-b px-4 py-1.5">
           <Button
             type="button"
-            variant="outline"
+            variant="ghost"
-            size="sm"
+            size="icon-sm"
+            title={quickConfigOpen ? "收起配置" : "展开配置"}
+            aria-label={quickConfigOpen ? "收起配置" : "展开配置"}
             aria-expanded={quickConfigOpen}
             onClick={() => onQuickConfigOpenChange(!quickConfigOpen)}
           >
-            {quickConfigOpen ? <ChevronUp /> : <ChevronDown />}
+            <Settings />
-            {quickConfigOpen ? "收起配置" : "展开配置"}
+          </Button>
+          <Button
+            type="button"
+            variant="ghost"
+            size="icon-sm"
+            title="展开聊天信息"
+            aria-label="展开聊天信息"
+            aria-expanded={headerOpen}
+            onClick={() => onHeaderOpenChange(true)}
+          >
+            <ChevronDown />
           </Button>
         </div>
-      </div>
+      )}
 
       {quickConfigOpen ? (
         <div className="quick-config border-b px-4 py-3">
@@ -368,23 +463,50 @@ export function ChatView({
             disabled={isBusy}
           />
           <div className="mt-2 flex flex-wrap items-center justify-between gap-2">
-            <div className="flex flex-wrap gap-2 text-xs text-muted-foreground">
+            <div className="flex flex-wrap gap-1">
-              <span>
+              <div ref={reasoningMenuRef} className="relative">
-                流式：{runtimeConfig.parameters.stream ? "开启" : "关闭"}
+                <Button
-              </span>
+                  type="button"
-              <span>温度：{runtimeConfig.parameters.temperature}</span>
+                  variant="ghost"
-              <span>
+                  size="icon-sm"
-                上下文：
+                  title={`推理强度 reasoning_effort：${selectedReasoningLabel}`}
-                {runtimeConfig.parameters.maxContextLength > 0
+                  aria-label="推理强度 reasoning_effort"
-                  ? `${runtimeConfig.parameters.maxContextLength} 字符`
+                  aria-haspopup="menu"
-                  : "无限"}
+                  aria-expanded={reasoningMenuOpen}
-              </span>
+                  disabled={isBusy}
-              <span>
+                  onClick={() => setReasoningMenuOpen((open) => !open)}
-                推理：
+                >
-                {runtimeConfig.parameters.reasoningEffort === "default"
+                  <SlidersHorizontal />
-                  ? "默认"
+                </Button>
-                  : runtimeConfig.parameters.reasoningEffort}
+                {reasoningMenuOpen ? (
-              </span>
+                  <div className="absolute bottom-full left-0 z-30 mb-2 w-40 rounded-lg border bg-popover p-1 text-sm text-popover-foreground shadow-lg">
+                    {REASONING_EFFORT_OPTIONS.map((option) => {
+                      const selected =
+                        option.value === runtimeConfig.parameters.reasoningEffort
+
+                      return (
+                        <button
+                          key={option.value}
+                          type="button"
+                          role="menuitemradio"
+                          aria-checked={selected}
+                          className={cn(
+                            "flex h-8 w-full items-center justify-between rounded-md px-2 text-left text-sm hover:bg-muted focus-visible:bg-muted focus-visible:outline-none",
+                            selected && "bg-muted font-medium"
+                          )}
+                          onClick={() => {
+                            onReasoningEffortChange(option.value)
+                            setReasoningMenuOpen(false)
+                          }}
+                        >
+                          <span>{option.label}</span>
+                          {selected ? <span className="text-xs">✓</span> : null}
+                        </button>
+                      )
+                    })}
+                  </div>
+                ) : null}
+              </div>
             </div>
             <div className="flex gap-2">
               {isBusy ? (
@@ -439,6 +561,7 @@ function MessageItem({
   onSaveEditAndSend,
 }: MessageItemProps) {
   const usage = formatUsage(message)
+  const cost = formatMessageCost(message.cost)
   const duration = formatDuration(message.elapsedMs)
   const isUserMessage = message.role === "user"
   const roleLabel =
@@ -566,10 +689,11 @@ function MessageItem({
             {message.error}
           </div>
         ) : null}
-        {duration || usage ? (
+        {duration || usage || cost ? (
           <div className="mt-2 flex flex-wrap gap-2 text-xs opacity-70">
-            {duration ? <span>耗时 {duration}</span> : null}
+            {duration ? <span>{duration}</span> : null}
             {usage ? <span>Token {usage}</span> : null}
+            {cost ? <span>{cost}</span> : null}
           </div>
         ) : null}
       </div>

src/components/app/models-view.tsx

@@ -52,6 +52,25 @@ export function ModelsView({
   onUpdateModel,
   onUpdateParameter,
 }: ModelsViewProps) {
+  function parseNonNegativeNumber(value: string, fallback = 0) {
+    const numericValue = Number(value)
+
+    return Number.isFinite(numericValue) && numericValue >= 0
+      ? numericValue
+      : fallback
+  }
+
+  function updatePricing(
+    model: ModelConfig,
+    patch: Partial<ModelConfig["pricing"]>
+  ) {
+    onUpdateModel(model.id, (item) => ({
+      ...item,
+      pricing: { ...item.pricing, ...patch },
+      updatedAt: nowIso(),
+    }))
+  }
+
   return (
     <section className="content-view overflow-y-auto p-4">
       <div className="mx-auto grid max-w-5xl gap-4">
@@ -252,6 +271,64 @@ export function ModelsView({
                   }
                 />
               </div>
+              <div className="mt-3 grid gap-3 md:grid-cols-2 lg:grid-cols-4">
+                <LabelledField label="货币">
+                  <Input
+                    value={model.pricing.currency}
+                    onChange={(event) =>
+                      updatePricing(model, { currency: event.target.value })
+                    }
+                    placeholder="USD"
+                  />
+                </LabelledField>
+                <LabelledField label="输入价 / 100万 token">
+                  <Input
+                    type="number"
+                    min="0"
+                    step="0.000001"
+                    value={model.pricing.inputPerMillionTokens}
+                    onChange={(event) =>
+                      updatePricing(model, {
+                        inputPerMillionTokens: parseNonNegativeNumber(
+                          event.target.value
+                        ),
+                      })
+                    }
+                  />
+                </LabelledField>
+                <LabelledField label="输出价 / 100万 token">
+                  <Input
+                    type="number"
+                    min="0"
+                    step="0.000001"
+                    value={model.pricing.outputPerMillionTokens}
+                    onChange={(event) =>
+                      updatePricing(model, {
+                        outputPerMillionTokens: parseNonNegativeNumber(
+                          event.target.value
+                        ),
+                      })
+                    }
+                  />
+                </LabelledField>
+                <LabelledField label="缓存输入价 / 100万 token">
+                  <Input
+                    type="number"
+                    min="0"
+                    step="0.000001"
+                    value={model.pricing.cachedInputPerMillionTokens ?? ""}
+                    onChange={(event) =>
+                      updatePricing(model, {
+                        cachedInputPerMillionTokens:
+                          event.target.value === ""
+                            ? undefined
+                            : parseNonNegativeNumber(event.target.value),
+                      })
+                    }
+                    placeholder="空则按输入价"
+                  />
+                </LabelledField>
+              </div>
             </Panel>
           ))}
         </div>

src/index.css

@@ -378,10 +378,11 @@
 
   .chat-header-actions {
     width: 100%;
+    justify-content: flex-end;
   }
 
   .chat-header-actions > button {
-    flex: 1 1 120px;
+    flex: 0 0 auto;
   }
 
   .message-bubble {

src/lib/default-data.ts

@@ -5,6 +5,7 @@ import type {
   ChatSession,
   ModelConfig,
   ModelParameters,
+  ModelPricing,
   PromptTemplate,
   ServiceConfig,
 } from "@/types"
@@ -22,6 +23,12 @@ export const DEFAULT_MODEL_PARAMETERS: ModelParameters = {
   reasoningEffort: "default",
 }
 
+export const DEFAULT_MODEL_PRICING: ModelPricing = {
+  currency: "USD",
+  inputPerMillionTokens: 0,
+  outputPerMillionTokens: 0,
+}
+
 export function createId(prefix: string) {
   const randomId =
     typeof crypto !== "undefined" && "randomUUID" in crypto
@@ -88,6 +95,7 @@ export function createModelConfig(
     serviceId,
     isDefault,
     parameters: { ...DEFAULT_MODEL_PARAMETERS },
+    pricing: { ...DEFAULT_MODEL_PRICING },
     updatedAt: nowIso(),
   }
 }
@@ -118,6 +126,7 @@ export function createDefaultSettings(
     activeServiceId: service.id,
     activeModelId: model.id,
     systemPrompt: DEFAULT_SYSTEM_PROMPT,
+    chatHeaderOpen: true,
     chatQuickConfigOpen: true,
     theme: "system",
     compactMode: false,

src/lib/openai.ts

@@ -91,12 +91,37 @@ function normalizeUsage(value: unknown): TokenUsage | undefined {
   const promptTokens = usage.prompt_tokens
   const completionTokens = usage.completion_tokens
   const totalTokens = usage.total_tokens
+  const promptTokensDetails = usage.prompt_tokens_details
+  const completionTokensDetails = usage.completion_tokens_details
+  const cachedPromptTokens =
+    typeof promptTokensDetails === "object" && promptTokensDetails !== null
+      ? (promptTokensDetails as Record<string, unknown>).cached_tokens
+      : undefined
+  const reasoningTokens =
+    typeof completionTokensDetails === "object" &&
+    completionTokensDetails !== null
+      ? (completionTokensDetails as Record<string, unknown>).reasoning_tokens
+      : undefined
+  const promptCacheHitTokens = usage.prompt_cache_hit_tokens
+  const promptCacheMissTokens = usage.prompt_cache_miss_tokens
 
   return {
     promptTokens: typeof promptTokens === "number" ? promptTokens : undefined,
     completionTokens:
       typeof completionTokens === "number" ? completionTokens : undefined,
     totalTokens: typeof totalTokens === "number" ? totalTokens : undefined,
+    cachedPromptTokens:
+      typeof cachedPromptTokens === "number" ? cachedPromptTokens : undefined,
+    promptCacheHitTokens:
+      typeof promptCacheHitTokens === "number"
+        ? promptCacheHitTokens
+        : undefined,
+    promptCacheMissTokens:
+      typeof promptCacheMissTokens === "number"
+        ? promptCacheMissTokens
+        : undefined,
+    reasoningTokens:
+      typeof reasoningTokens === "number" ? reasoningTokens : undefined,
   }
 }
 
@@ -303,6 +328,10 @@ export async function createChatCompletion({
     body.reasoning_effort = parameters.reasoningEffort
   }
 
+  if (parameters.stream) {
+    body.stream_options = { include_usage: true }
+  }
+
   if (tools?.length) {
     body.tools = tools
     body.tool_choice = toolChoice ?? "auto"

src/lib/pricing.ts

@@ -0,0 +1,91 @@
+import type { MessageCost, ModelPricing, TokenUsage } from "@/types"
+
+const TOKENS_PER_MILLION = 1_000_000
+
+function nonNegativeNumber(value: number | undefined) {
+  return typeof value === "number" && Number.isFinite(value) && value >= 0
+    ? value
+    : undefined
+}
+
+function priceValue(value: number | undefined) {
+  return nonNegativeNumber(value) ?? 0
+}
+
+function formatDecimal(value: number) {
+  if (value > 0 && value < 0.000001) {
+    return "<0.000001"
+  }
+
+  const maximumFractionDigits = value >= 1 ? 4 : 6
+
+  return new Intl.NumberFormat("zh-CN", {
+    maximumFractionDigits,
+  }).format(value)
+}
+
+export function calculateUsageCost(
+  usage: TokenUsage | undefined,
+  pricing: ModelPricing | undefined
+): MessageCost | undefined {
+  if (!usage || !pricing) {
+    return undefined
+  }
+
+  const inputPrice = priceValue(pricing.inputPerMillionTokens)
+  const outputPrice = priceValue(pricing.outputPerMillionTokens)
+  const cachedInputPrice =
+    nonNegativeNumber(pricing.cachedInputPerMillionTokens) ?? inputPrice
+
+  if (inputPrice === 0 && outputPrice === 0 && cachedInputPrice === 0) {
+    return undefined
+  }
+
+  const promptTokens = nonNegativeNumber(usage.promptTokens)
+  const cachedPromptTokens =
+    nonNegativeNumber(usage.promptCacheHitTokens) ??
+    nonNegativeNumber(usage.cachedPromptTokens)
+  const promptCacheMissTokens = nonNegativeNumber(usage.promptCacheMissTokens)
+  const billablePromptTokens =
+    promptCacheMissTokens ??
+    (promptTokens !== undefined
+      ? Math.max(0, promptTokens - (cachedPromptTokens ?? 0))
+      : undefined)
+  const completionTokens = nonNegativeNumber(usage.completionTokens)
+  const promptCost =
+    billablePromptTokens !== undefined
+      ? (billablePromptTokens * inputPrice) / TOKENS_PER_MILLION
+      : undefined
+  const cachedPromptCost =
+    cachedPromptTokens !== undefined
+      ? (cachedPromptTokens * cachedInputPrice) / TOKENS_PER_MILLION
+      : undefined
+  const completionCost =
+    completionTokens !== undefined
+      ? (completionTokens * outputPrice) / TOKENS_PER_MILLION
+      : undefined
+  const totalCost =
+    (promptCost ?? 0) + (cachedPromptCost ?? 0) + (completionCost ?? 0)
+
+  if (totalCost === 0) {
+    return undefined
+  }
+
+  return {
+    currency: pricing.currency.trim() || "USD",
+    promptCost,
+    cachedPromptCost,
+    completionCost,
+    totalCost,
+  }
+}
+
+export function formatMessageCost(cost: MessageCost | undefined) {
+  if (!cost || !Number.isFinite(cost.totalCost) || cost.totalCost <= 0) {
+    return ""
+  }
+
+  const currency = cost.currency.trim() || "USD"
+
+  return `${currency} ${formatDecimal(cost.totalCost)}`
+}

src/lib/storage.ts

@@ -1,5 +1,6 @@
 import {
   DEFAULT_MODEL_PARAMETERS,
+  DEFAULT_MODEL_PRICING,
   createDefaultData,
   nowIso,
 } from "@/lib/default-data"
@@ -10,6 +11,7 @@ import type {
   ChatMessage,
   ChatSession,
   ModelConfig,
+  ModelPricing,
   PromptTemplate,
   ToolChoiceMode,
   ToolSettings,
@@ -30,6 +32,42 @@ function isToolChoiceMode(value: unknown): value is ToolChoiceMode {
   return value === "auto" || value === "none" || value === "required"
 }
 
+function normalizeNonNegativeNumber(value: unknown, fallback: number) {
+  return typeof value === "number" && Number.isFinite(value) && value >= 0
+    ? value
+    : fallback
+}
+
+function normalizeOptionalNonNegativeNumber(value: unknown) {
+  return typeof value === "number" && Number.isFinite(value) && value >= 0
+    ? value
+    : undefined
+}
+
+function normalizeModelPricing(value: unknown): ModelPricing {
+  if (!isRecord(value)) {
+    return { ...DEFAULT_MODEL_PRICING }
+  }
+
+  return {
+    currency:
+      typeof value.currency === "string" && value.currency.trim()
+        ? value.currency.trim()
+        : DEFAULT_MODEL_PRICING.currency,
+    inputPerMillionTokens: normalizeNonNegativeNumber(
+      value.inputPerMillionTokens,
+      DEFAULT_MODEL_PRICING.inputPerMillionTokens
+    ),
+    outputPerMillionTokens: normalizeNonNegativeNumber(
+      value.outputPerMillionTokens,
+      DEFAULT_MODEL_PRICING.outputPerMillionTokens
+    ),
+    cachedInputPerMillionTokens: normalizeOptionalNonNegativeNumber(
+      value.cachedInputPerMillionTokens
+    ),
+  }
+}
+
 function normalizeToolSettings(
   value: unknown,
   fallback: ToolSettings
@@ -128,6 +166,10 @@ function normalizeData(value: unknown): AppData {
     })),
     settings: {
       ...settings,
+      chatHeaderOpen:
+        typeof settings.chatHeaderOpen === "boolean"
+          ? settings.chatHeaderOpen
+          : fallback.settings.chatHeaderOpen,
       chatQuickConfigOpen:
         typeof settings.chatQuickConfigOpen === "boolean"
           ? settings.chatQuickConfigOpen
@@ -146,6 +188,7 @@ function normalizeData(value: unknown): AppData {
         ...DEFAULT_MODEL_PARAMETERS,
         ...(isRecord(model.parameters) ? model.parameters : {}),
       },
+      pricing: normalizeModelPricing(model.pricing),
     })),
     prompts: safePrompts,
   }

src/types.ts

@@ -27,6 +27,25 @@ export type TokenUsage = {
   promptTokens?: number
   completionTokens?: number
   totalTokens?: number
+  cachedPromptTokens?: number
+  promptCacheHitTokens?: number
+  promptCacheMissTokens?: number
+  reasoningTokens?: number
+}
+
+export type ModelPricing = {
+  currency: string
+  inputPerMillionTokens: number
+  outputPerMillionTokens: number
+  cachedInputPerMillionTokens?: number
+}
+
+export type MessageCost = {
+  currency: string
+  promptCost?: number
+  cachedPromptCost?: number
+  completionCost?: number
+  totalCost: number
 }
 
 export type ModelParameters = {
@@ -111,6 +130,7 @@ export type ChatMessage = {
   model?: string
   elapsedMs?: number
   usage?: TokenUsage
+  cost?: MessageCost
   error?: string
   reasoningContent?: string
   toolCallId?: string
@@ -156,6 +176,7 @@ export type ModelConfig = {
   serviceId?: string
   isDefault?: boolean
   parameters: ModelParameters
+  pricing: ModelPricing
   updatedAt: string
 }
 
@@ -188,6 +209,7 @@ export type AppSettings = ModelParameters & {
   activeServiceId: string
   activeModelId: string
   systemPrompt: string
+  chatHeaderOpen: boolean
   chatQuickConfigOpen: boolean
   theme: ThemeMode
   compactMode: boolean