jinzd-ai-cli 0.4.54 → 0.4.55

package/dist/{chunk-FOFQAEU6.js → chunk-DJ342VFS.js}

@@ -6,7 +6,7 @@ import { platform } from "os";
 import chalk from "chalk";
 
 // src/core/constants.ts
-var VERSION = "0.4.54";
+var VERSION = "0.4.55";
 var APP_NAME = "ai-cli";
 var CONFIG_DIR_NAME = ".aicli";
 var CONFIG_FILE_NAME = "config.json";

package/dist/{chunk-6FYFVPVE.js → chunk-JL5NK6AR.js}

@@ -7,7 +7,7 @@ import {
   ProviderNotFoundError,
   RateLimitError,
   schemaToJsonSchema
-} from "./chunk-TAR67QTH.js";
+} from "./chunk-YQEIQJ6K.js";
 import {
   APP_NAME,
   CONFIG_DIR_NAME,
@@ -20,7 +20,7 @@ import {
   MCP_TOOL_PREFIX,
   PLUGINS_DIR_NAME,
   VERSION
-} from "./chunk-NP5KZVP6.js";
+} from "./chunk-W7QVBFIJ.js";
 
 // src/config/config-manager.ts
 import { readFileSync, writeFileSync, existsSync, mkdirSync } from "fs";
@@ -312,6 +312,7 @@ var BaseProvider = class {
 };
 
 // src/providers/claude.ts
+var CACHE_MIN_SYSTEM_CHARS = 2e3;
 var ClaudeProvider = class extends BaseProvider {
   client;
   info = {
@@ -382,6 +383,52 @@ var ClaudeProvider = class extends BaseProvider {
     }
     return blocks.length > 0 ? blocks : "";
   }
+  /**
+   * Build a cacheable system prompt payload.
+   * When the prompt is long enough to be worth caching, return an array with a
+   * single text block carrying `cache_control: { type: 'ephemeral' }`. This caches
+   * system + memory + context files across every request in an agentic loop.
+   * Short prompts pass through as a plain string (no caching overhead).
+   */
+  buildSystemParam(systemPrompt) {
+    if (!systemPrompt) return void 0;
+    if (systemPrompt.length < CACHE_MIN_SYSTEM_CHARS) return systemPrompt;
+    return [
+      {
+        type: "text",
+        text: systemPrompt,
+        cache_control: { type: "ephemeral" }
+      }
+    ];
+  }
+  /**
+   * Mark the last tool definition with `cache_control: ephemeral` so the entire
+   * tool block (all 24+ tools) is cached together. Anthropic caches everything
+   * up to and including a cache breakpoint, so one marker covers all tools.
+   * Returns a new array — does not mutate the input.
+   */
+  addToolsCacheControl(tools) {
+    if (tools.length === 0) return tools;
+    const last = tools[tools.length - 1];
+    return [
+      ...tools.slice(0, -1),
+      { ...last, cache_control: { type: "ephemeral" } }
+    ];
+  }
+  /** Extract usage (including cache fields) from an Anthropic response. */
+  extractUsage(u) {
+    const usage = {
+      inputTokens: u.input_tokens,
+      outputTokens: u.output_tokens
+    };
+    if (u.cache_creation_input_tokens != null && u.cache_creation_input_tokens > 0) {
+      usage.cacheCreationTokens = u.cache_creation_input_tokens;
+    }
+    if (u.cache_read_input_tokens != null && u.cache_read_input_tokens > 0) {
+      usage.cacheReadTokens = u.cache_read_input_tokens;
+    }
+    return usage;
+  }
   /**
    * 构建 Extended Thinking 参数。
    * - thinking 启用时 temperature 必须为 1 或不设置（Anthropic API 要求）
@@ -432,7 +479,7 @@ var ClaudeProvider = class extends BaseProvider {
       const response = await this.client.messages.create({
         model: request.model,
         messages,
-        system: request.systemPrompt,
+        system: this.buildSystemParam(request.systemPrompt),
         max_tokens: request.maxTokens ?? 8192,
         temperature,
         thinking
@@ -441,10 +488,7 @@ var ClaudeProvider = class extends BaseProvider {
       return {
         content,
         model: response.model,
-        usage: {
-          inputTokens: response.usage.input_tokens,
-          outputTokens: response.usage.output_tokens
-        }
+        usage: this.extractUsage(response.usage)
       };
     } catch (err) {
       throw this.wrapError(err);
@@ -460,7 +504,7 @@ var ClaudeProvider = class extends BaseProvider {
       const stream = this.client.messages.stream({
         model: request.model,
         messages,
-        system: request.systemPrompt,
+        system: this.buildSystemParam(request.systemPrompt),
         max_tokens: request.maxTokens ?? 8192,
         temperature,
         thinking
@@ -493,20 +537,22 @@ var ClaudeProvider = class extends BaseProvider {
   }
   async chatWithTools(request, tools) {
     try {
-      const anthropicTools = tools.map((t) => ({
-        name: t.name,
-        description: t.description,
-        input_schema: {
-          type: "object",
-          properties: Object.fromEntries(
-            Object.entries(t.parameters).map(([key, schema]) => [
-              key,
-              schemaToJsonSchema(schema)
-            ])
-          ),
-          required: Object.entries(t.parameters).filter(([, s]) => s.required).map(([k]) => k)
-        }
-      }));
+      const anthropicTools = this.addToolsCacheControl(
+        tools.map((t) => ({
+          name: t.name,
+          description: t.description,
+          input_schema: {
+            type: "object",
+            properties: Object.fromEntries(
+              Object.entries(t.parameters).map(([key, schema]) => [
+                key,
+                schemaToJsonSchema(schema)
+              ])
+            ),
+            required: Object.entries(t.parameters).filter(([, s]) => s.required).map(([k]) => k)
+          }
+        }))
+      );
       const baseMessages = request.messages.filter((m) => m.role !== "system").map((m) => ({ role: m.role, content: this.contentToClaudeParts(m.content) }));
       const extraMessages = request._extraMessages ?? [];
       const allMessages = [...baseMessages, ...extraMessages];
@@ -515,15 +561,12 @@ var ClaudeProvider = class extends BaseProvider {
         model: request.model,
         messages: allMessages,
         tools: anthropicTools,
-        system: request.systemPrompt,
+        system: this.buildSystemParam(request.systemPrompt),
         max_tokens: request.maxTokens ?? 8192,
         temperature,
         thinking
       });
-      const usage = {
-        inputTokens: response.usage.input_tokens,
-        outputTokens: response.usage.output_tokens
-      };
+      const usage = this.extractUsage(response.usage);
       const toolUseBlocks = response.content.filter(
         (b) => b.type === "tool_use"
       );
@@ -547,20 +590,22 @@ var ClaudeProvider = class extends BaseProvider {
    * 同时收集原始 content blocks 供 buildToolResultMessages 使用。
    */
   async *chatWithToolsStream(request, tools) {
-    const anthropicTools = tools.map((t) => ({
-      name: t.name,
-      description: t.description,
-      input_schema: {
-        type: "object",
-        properties: Object.fromEntries(
-          Object.entries(t.parameters).map(([key, schema]) => [
-            key,
-            schemaToJsonSchema(schema)
-          ])
-        ),
-        required: Object.entries(t.parameters).filter(([, s]) => s.required).map(([k]) => k)
-      }
-    }));
+    const anthropicTools = this.addToolsCacheControl(
+      tools.map((t) => ({
+        name: t.name,
+        description: t.description,
+        input_schema: {
+          type: "object",
+          properties: Object.fromEntries(
+            Object.entries(t.parameters).map(([key, schema]) => [
+              key,
+              schemaToJsonSchema(schema)
+            ])
+          ),
+          required: Object.entries(t.parameters).filter(([, s]) => s.required).map(([k]) => k)
+        }
+      }))
+    );
     const baseMessages = request.messages.filter((m) => m.role !== "system").map((m) => ({ role: m.role, content: this.contentToClaudeParts(m.content) }));
     const extraMessages = request._extraMessages ?? [];
     const allMessages = [...baseMessages, ...extraMessages];
@@ -572,7 +617,7 @@ var ClaudeProvider = class extends BaseProvider {
         model: request.model,
         messages: allMessages,
         tools: anthropicTools,
-        system: request.systemPrompt,
+        system: this.buildSystemParam(request.systemPrompt),
         max_tokens: request.maxTokens ?? 8192,
         temperature,
         thinking
@@ -580,7 +625,13 @@ var ClaudeProvider = class extends BaseProvider {
       let currentBlockType = null;
       let currentToolIndex = 0;
       let currentBlockData = {};
+      let startUsage = null;
       for await (const event of stream) {
+        if (event.type === "message_start") {
+          const msgUsage = event.message?.usage;
+          if (msgUsage) startUsage = msgUsage;
+          continue;
+        }
         if (event.type === "content_block_start") {
           const block = event.content_block;
           currentBlockType = block.type;
@@ -641,15 +692,17 @@ var ClaudeProvider = class extends BaseProvider {
           currentBlockType = null;
           currentBlockData = {};
         } else if (event.type === "message_delta") {
-          const usage = event.usage;
-          if (usage) {
+          const deltaUsage = event.usage;
+          if (deltaUsage) {
             doneEmitted = true;
             yield {
               type: "done",
-              usage: {
-                inputTokens: usage.input_tokens ?? 0,
-                outputTokens: usage.output_tokens ?? 0
-              },
+              usage: this.extractUsage({
+                input_tokens: startUsage?.input_tokens ?? deltaUsage.input_tokens ?? 0,
+                output_tokens: deltaUsage.output_tokens ?? 0,
+                cache_creation_input_tokens: startUsage?.cache_creation_input_tokens,
+                cache_read_input_tokens: startUsage?.cache_read_input_tokens
+              }),
               rawContent: rawContentBlocks
             };
           }
@@ -1003,6 +1056,16 @@ Node.js does not automatically use system proxies. Try one of the following:
 
 // src/providers/openai-compatible.ts
 import OpenAI from "openai";
+function toUsage(u) {
+  if (!u) return void 0;
+  const cached = u.prompt_tokens_details?.cached_tokens ?? 0;
+  const usage = {
+    inputTokens: Math.max(0, u.prompt_tokens - cached),
+    outputTokens: u.completion_tokens
+  };
+  if (cached > 0) usage.cacheReadTokens = cached;
+  return usage;
+}
 var OpenAICompatibleProvider = class extends BaseProvider {
   client;
   defaultTimeout = 6e4;
@@ -1056,10 +1119,7 @@ var OpenAICompatibleProvider = class extends BaseProvider {
       return {
         content: firstChoice.message.content ?? "",
         model: response.model,
-        usage: response.usage ? {
-          inputTokens: response.usage.prompt_tokens,
-          outputTokens: response.usage.completion_tokens
-        } : void 0
+        usage: toUsage(response.usage)
       };
     } catch (err) {
       throw this.wrapError(err);
@@ -1088,10 +1148,7 @@ var OpenAICompatibleProvider = class extends BaseProvider {
           yield {
             delta: "",
             done: true,
-            usage: {
-              inputTokens: chunk.usage.prompt_tokens,
-              outputTokens: chunk.usage.completion_tokens
-            }
+            usage: toUsage(chunk.usage)
           };
           continue;
         }
@@ -1159,10 +1216,7 @@ var OpenAICompatibleProvider = class extends BaseProvider {
         return { content: "", usage: void 0 };
       }
       const message = firstChoice.message;
-      const usage = response.usage ? {
-        inputTokens: response.usage.prompt_tokens,
-        outputTokens: response.usage.completion_tokens
-      } : void 0;
+      const usage = toUsage(response.usage);
       const reasoningContent = message.reasoning_content;
       if (message.tool_calls && message.tool_calls.length > 0) {
         const toolCalls = message.tool_calls.map((tc) => {
@@ -1275,10 +1329,7 @@ var OpenAICompatibleProvider = class extends BaseProvider {
           }
           yield {
             type: "done",
-            usage: {
-              inputTokens: chunk.usage.prompt_tokens,
-              outputTokens: chunk.usage.completion_tokens
-            }
+            usage: toUsage(chunk.usage)
           };
           continue;
         }
@@ -2331,7 +2382,12 @@ var Session = class _Session {
   updated;
   messages = [];
   title;
-  tokenUsage = { inputTokens: 0, outputTokens: 0 };
+  tokenUsage = {
+    inputTokens: 0,
+    outputTokens: 0,
+    cacheCreationTokens: 0,
+    cacheReadTokens: 0
+  };
   checkpoints = [];
   constructor(id, provider, model) {
     this.id = id;
@@ -2359,11 +2415,18 @@ var Session = class _Session {
   addTokenUsage(usage) {
     this.tokenUsage.inputTokens += usage.inputTokens;
     this.tokenUsage.outputTokens += usage.outputTokens;
+    this.tokenUsage.cacheCreationTokens += usage.cacheCreationTokens ?? 0;
+    this.tokenUsage.cacheReadTokens += usage.cacheReadTokens ?? 0;
   }
   clear() {
     this.messages = [];
     this.title = void 0;
-    this.tokenUsage = { inputTokens: 0, outputTokens: 0 };
+    this.tokenUsage = {
+      inputTokens: 0,
+      outputTokens: 0,
+      cacheCreationTokens: 0,
+      cacheReadTokens: 0
+    };
     this.updated = /* @__PURE__ */ new Date();
   }
   /**
@@ -2491,7 +2554,9 @@ var Session = class _Session {
     if (tu && typeof tu === "object") {
       session.tokenUsage = {
         inputTokens: typeof tu.inputTokens === "number" ? tu.inputTokens : 0,
-        outputTokens: typeof tu.outputTokens === "number" ? tu.outputTokens : 0
+        outputTokens: typeof tu.outputTokens === "number" ? tu.outputTokens : 0,
+        cacheCreationTokens: typeof tu.cacheCreationTokens === "number" ? tu.cacheCreationTokens : 0,
+        cacheReadTokens: typeof tu.cacheReadTokens === "number" ? tu.cacheReadTokens : 0
       };
     }
     if (Array.isArray(d.checkpoints)) {
@@ -3495,6 +3560,87 @@ async function setupProxy(configProxy) {
   }
 }
 
+// src/core/pricing.ts
+var PRICING_TABLE = {
+  // ── Anthropic Claude ──────────────────────────────────────────
+  "claude-opus-4-6": { input: 15, output: 75, cacheWrite: 18.75, cacheRead: 1.5 },
+  "claude-opus-4-5": { input: 15, output: 75, cacheWrite: 18.75, cacheRead: 1.5 },
+  "claude-sonnet-4-6": { input: 3, output: 15, cacheWrite: 3.75, cacheRead: 0.3 },
+  "claude-sonnet-4-5-20250929": { input: 3, output: 15, cacheWrite: 3.75, cacheRead: 0.3 },
+  "claude-haiku-4-5-20251001": { input: 1, output: 5, cacheWrite: 1.25, cacheRead: 0.1 },
+  "claude-haiku-4-5": { input: 1, output: 5, cacheWrite: 1.25, cacheRead: 0.1 },
+  // Legacy Claude 3.x families (prefix fallback handles minor date suffixes)
+  "claude-3-5-sonnet": { input: 3, output: 15, cacheWrite: 3.75, cacheRead: 0.3 },
+  "claude-3-5-haiku": { input: 0.8, output: 4, cacheWrite: 1, cacheRead: 0.08 },
+  "claude-3-opus": { input: 15, output: 75, cacheWrite: 18.75, cacheRead: 1.5 },
+  // ── OpenAI ────────────────────────────────────────────────────
+  "gpt-4o": { input: 2.5, output: 10, cacheRead: 1.25 },
+  "gpt-4o-mini": { input: 0.15, output: 0.6, cacheRead: 0.075 },
+  "gpt-4-turbo": { input: 10, output: 30 },
+  "gpt-4": { input: 30, output: 60 },
+  "gpt-4.1": { input: 2, output: 8, cacheRead: 0.5 },
+  "gpt-4.1-mini": { input: 0.4, output: 1.6, cacheRead: 0.1 },
+  "gpt-4.1-nano": { input: 0.1, output: 0.4, cacheRead: 0.025 },
+  "o1": { input: 15, output: 60, cacheRead: 7.5 },
+  "o1-mini": { input: 3, output: 12, cacheRead: 1.5 },
+  "o3": { input: 10, output: 40, cacheRead: 2.5 },
+  "o3-mini": { input: 1.1, output: 4.4, cacheRead: 0.55 },
+  // ── Google Gemini ─────────────────────────────────────────────
+  "gemini-2.5-pro": { input: 1.25, output: 10 },
+  "gemini-2.5-flash": { input: 0.3, output: 2.5 },
+  "gemini-2.0-flash": { input: 0.1, output: 0.4 },
+  "gemini-1.5-pro": { input: 1.25, output: 5 },
+  "gemini-1.5-flash": { input: 0.075, output: 0.3 },
+  // ── DeepSeek ──────────────────────────────────────────────────
+  "deepseek-chat": { input: 0.27, output: 1.1, cacheRead: 0.07 },
+  "deepseek-reasoner": { input: 0.55, output: 2.19, cacheRead: 0.14 },
+  "deepseek-v3": { input: 0.27, output: 1.1, cacheRead: 0.07 },
+  // ── Moonshot Kimi ─────────────────────────────────────────────
+  "moonshot-v1-8k": { input: 0.17, output: 0.17 },
+  "moonshot-v1-32k": { input: 0.33, output: 0.33 },
+  "moonshot-v1-128k": { input: 0.83, output: 0.83 },
+  "kimi-k2": { input: 0.6, output: 2.5 },
+  "kimi-latest": { input: 0.6, output: 2.5 },
+  // ── Zhipu GLM ─────────────────────────────────────────────────
+  "glm-4-plus": { input: 0.7, output: 0.7 },
+  "glm-4": { input: 0.14, output: 0.14 },
+  "glm-4-flash": { input: 0, output: 0 },
+  "glm-4.5": { input: 0.29, output: 1.14 },
+  "glm-4.6": { input: 0.6, output: 2.2 }
+  // ── OpenRouter (pass-through — actual cost depends on underlying model) ──
+  // Left empty; callers should resolve via underlying model ID.
+  // ── Ollama (local, zero cost) ─────────────────────────────────
+  // Handled via provider check below.
+};
+var FREE_PROVIDERS = /* @__PURE__ */ new Set(["ollama"]);
+function getPricing(provider, model) {
+  if (FREE_PROVIDERS.has(provider.toLowerCase())) {
+    return { input: 0, output: 0 };
+  }
+  const key = model.toLowerCase();
+  if (PRICING_TABLE[key]) return PRICING_TABLE[key];
+  const keys = Object.keys(PRICING_TABLE).sort((a, b) => b.length - a.length);
+  for (const k of keys) {
+    if (key.startsWith(k)) return PRICING_TABLE[k];
+  }
+  return null;
+}
+function computeCost(provider, model, usage) {
+  const p = getPricing(provider, model);
+  if (!p) return null;
+  const input = usage.inputTokens * p.input;
+  const output = usage.outputTokens * p.output;
+  const cacheWrite = (usage.cacheCreationTokens ?? 0) * (p.cacheWrite ?? p.input);
+  const cacheRead = (usage.cacheReadTokens ?? 0) * (p.cacheRead ?? p.input);
+  return (input + output + cacheWrite + cacheRead) / 1e6;
+}
+function formatCost(amount) {
+  if (amount === 0) return "$0.0000";
+  if (amount < 0.01) return `$${amount.toFixed(4)}`;
+  if (amount < 1) return `$${amount.toFixed(3)}`;
+  return `$${amount.toFixed(2)}`;
+}
+
 // src/repl/dev-state.ts
 import { existsSync as existsSync5, readFileSync as readFileSync4, writeFileSync as writeFileSync3, unlinkSync as unlinkSync2, mkdirSync as mkdirSync4 } from "fs";
 import { join as join5 } from "path";
@@ -3601,6 +3747,9 @@ export {
   getGitRoot,
   getGitContext,
   formatGitContextForPrompt,
+  getPricing,
+  computeCost,
+  formatCost,
   parseSimpleYaml,
   SNAPSHOT_PROMPT,
   sessionHasMeaningfulContent,

package/dist/{chunk-NP5KZVP6.js → chunk-W7QVBFIJ.js}

@@ -8,7 +8,7 @@ import { platform } from "os";
 import chalk from "chalk";
 
 // src/core/constants.ts
-var VERSION = "0.4.54";
+var VERSION = "0.4.55";
 var APP_NAME = "ai-cli";
 var CONFIG_DIR_NAME = ".aicli";
 var CONFIG_FILE_NAME = "config.json";

package/dist/{chunk-TAR67QTH.js → chunk-YQEIQJ6K.js}

@@ -10,7 +10,7 @@ import {
   SUBAGENT_DEFAULT_MAX_ROUNDS,
   SUBAGENT_MAX_ROUNDS_LIMIT,
   runTestsTool
-} from "./chunk-NP5KZVP6.js";
+} from "./chunk-W7QVBFIJ.js";
 
 // src/tools/builtin/bash.ts
 import { execSync } from "child_process";

package/dist/{hub-6V54V4O3.js → hub-AUWP4SWJ.js}

@@ -385,7 +385,7 @@ ${content}`);
   }
 }
 async function runTaskMode(config, providers, configManager, topic) {
-  const { TaskOrchestrator } = await import("./task-orchestrator-TSY7CJE6.js");
+  const { TaskOrchestrator } = await import("./task-orchestrator-MWO6A4KQ.js");
   const orchestrator = new TaskOrchestrator(config, providers, configManager);
   let interrupted = false;
   const onSigint = () => {

package/dist/index.js

@@ -11,20 +11,23 @@ import {
   buildPhantomCorrectionMessage,
   buildWriteRoundReminder,
   clearDevState,
+  computeCost,
   detectsHallucinatedFileOp,
   extractWrittenFilePaths,
   findPhantomClaims,
+  formatCost,
   formatGitContextForPrompt,
   getContentText,
   getGitContext,
   getGitRoot,
+  getPricing,
   hadPreviousWriteToolCalls,
   loadDevState,
   parseSimpleYaml,
   saveDevState,
   sessionHasMeaningfulContent,
   setupProxy
-} from "./chunk-6FYFVPVE.js";
+} from "./chunk-JL5NK6AR.js";
 import {
   ToolExecutor,
   ToolRegistry,
@@ -38,7 +41,7 @@ import {
   spawnAgentContext,
   theme,
   undoStack
-} from "./chunk-TAR67QTH.js";
+} from "./chunk-YQEIQJ6K.js";
 import {
   fileCheckpoints
 } from "./chunk-4BKXL7SM.js";
@@ -63,7 +66,7 @@ import {
   SKILLS_DIR_NAME,
   VERSION,
   buildUserIdentityPrompt
-} from "./chunk-NP5KZVP6.js";
+} from "./chunk-W7QVBFIJ.js";
 
 // src/index.ts
 import { program } from "commander";
@@ -487,8 +490,12 @@ Error${typeName}: ${lines.join("\n")}
   renderUsage(usage, sessionTotal) {
     const total = usage.inputTokens + usage.outputTokens;
     let line = theme.dim("\u{1F4CA} ") + theme.dim(`in ${usage.inputTokens.toLocaleString()}`) + theme.dim(" + ") + theme.dim(`out ${usage.outputTokens.toLocaleString()}`) + theme.dim(` = ${total.toLocaleString()} tokens`);
+    const cacheRead = usage.cacheReadTokens ?? 0;
+    if (cacheRead > 0) {
+      line += theme.dim(`  \u2502  cache: ${cacheRead.toLocaleString()}`);
+    }
     if (sessionTotal) {
-      const sessionSum = sessionTotal.inputTokens + sessionTotal.outputTokens;
+      const sessionSum = sessionTotal.inputTokens + sessionTotal.outputTokens + (sessionTotal.cacheCreationTokens ?? 0) + (sessionTotal.cacheReadTokens ?? 0);
       line += theme.dim(`  \u2502  session total: ${sessionSum.toLocaleString()}`);
     }
     process.stdout.write(line + "\n\n");
@@ -1217,11 +1224,18 @@ function createDefaultCommands() {
         if (sys) {
           console.log(`  System   : ${sys.slice(0, 60)}...`);
         }
-        const totalTokens = tokenUsage.inputTokens + tokenUsage.outputTokens;
+        const cacheRead = tokenUsage.cacheReadTokens ?? 0;
+        const cacheCreate = tokenUsage.cacheCreationTokens ?? 0;
+        const totalTokens = tokenUsage.inputTokens + tokenUsage.outputTokens + cacheRead + cacheCreate;
         if (totalTokens > 0) {
+          const cacheSuffix = cacheRead > 0 || cacheCreate > 0 ? ` [cache: +${cacheCreate.toLocaleString()} / -${cacheRead.toLocaleString()}]` : "";
           console.log(
-            `  Tokens   : in ${tokenUsage.inputTokens.toLocaleString()} + out ${tokenUsage.outputTokens.toLocaleString()} = ${totalTokens.toLocaleString()} (session total)`
+            `  Tokens   : in ${tokenUsage.inputTokens.toLocaleString()} + out ${tokenUsage.outputTokens.toLocaleString()} = ${totalTokens.toLocaleString()}${cacheSuffix}`
           );
+          const cost = computeCost(ctx.getCurrentProvider(), ctx.getCurrentModel(), tokenUsage);
+          if (cost != null) {
+            console.log(`  Cost     : ${formatCost(cost)} (session total)`);
+          }
         }
         const ctxWindowSize = ctx.getContextWindowSize();
         if (ctxWindowSize > 0) {
@@ -1823,7 +1837,7 @@ ${hint}` : "")
     },
     {
       name: "cost",
-      description: "Show session token usage summary",
+      description: "Show session token usage, prompt-cache hits, and USD cost",
       usage: "/cost [reset]",
       execute(args, ctx) {
         const sub = args[0]?.toLowerCase();
@@ -1832,24 +1846,65 @@ ${hint}` : "")
           ctx.renderer.printSuccess("Session token counters reset.");
           return;
         }
-        const usage = ctx.getSessionTokenUsage();
-        const totalTokens = usage.inputTokens + usage.outputTokens;
+        const session = ctx.sessions.current;
+        const usage = session?.tokenUsage ?? {
+          inputTokens: 0,
+          outputTokens: 0,
+          cacheCreationTokens: 0,
+          cacheReadTokens: 0
+        };
+        const cacheCreate = usage.cacheCreationTokens ?? 0;
+        const cacheRead = usage.cacheReadTokens ?? 0;
+        const totalTokens = usage.inputTokens + usage.outputTokens + cacheCreate + cacheRead;
         if (totalTokens === 0) {
           ctx.renderer.printInfo("No token usage recorded this session.");
           return;
         }
+        const provider = ctx.getCurrentProvider();
+        const model = ctx.getCurrentModel();
+        const cost = computeCost(provider, model, usage);
+        const pricing = getPricing(provider, model);
+        let savings = null;
+        if (cost != null && pricing && cacheRead > 0) {
+          const costWithoutCache = computeCost(provider, model, {
+            inputTokens: usage.inputTokens + cacheRead,
+            outputTokens: usage.outputTokens,
+            cacheCreationTokens: cacheCreate,
+            cacheReadTokens: 0
+          });
+          if (costWithoutCache != null) savings = costWithoutCache - cost;
+        }
         console.log();
-        console.log(theme.heading("  \u{1F4CA} Session Token Usage"));
-        console.log(theme.dim("  " + "\u2500".repeat(40)));
-        console.log(theme.dim("  Input tokens  : ") + chalk2.white(usage.inputTokens.toLocaleString()));
-        console.log(theme.dim("  Output tokens : ") + chalk2.white(usage.outputTokens.toLocaleString()));
-        console.log(theme.dim("  Total tokens  : ") + chalk2.bold.white(totalTokens.toLocaleString()));
-        console.log(theme.dim("  " + "\u2500".repeat(40)));
-        const session = ctx.sessions.current;
+        console.log(theme.heading("  \u{1F4B0} Session Cost & Token Usage"));
+        console.log(theme.dim("  " + "\u2500".repeat(48)));
+        console.log(theme.dim("  Input (uncached) : ") + chalk2.white(usage.inputTokens.toLocaleString().padStart(12)));
+        console.log(theme.dim("  Output           : ") + chalk2.white(usage.outputTokens.toLocaleString().padStart(12)));
+        if (cacheCreate > 0) {
+          console.log(theme.dim("  Cache write      : ") + chalk2.yellow(cacheCreate.toLocaleString().padStart(12)));
+        }
+        if (cacheRead > 0) {
+          const pct = Math.round(cacheRead / (cacheRead + usage.inputTokens) * 100);
+          console.log(
+            theme.dim("  Cache read       : ") + chalk2.green(cacheRead.toLocaleString().padStart(12)) + theme.dim(`  (${pct}% hit rate)`)
+          );
+        }
+        console.log(theme.dim("  Total tokens     : ") + chalk2.bold.white(totalTokens.toLocaleString().padStart(12)));
+        console.log(theme.dim("  " + "\u2500".repeat(48)));
+        if (cost != null) {
+          console.log(theme.dim("  Cost             : ") + chalk2.bold.cyan(formatCost(cost).padStart(12)));
+          if (savings != null && savings > 0) {
+            console.log(
+              theme.dim("  Cache savings    : ") + chalk2.green(`-${formatCost(savings)}`.padStart(12)) + theme.dim(`  (vs no cache)`)
+            );
+          }
+        } else {
+          console.log(theme.dim("  Cost             : ") + theme.dim("      \u2014 (pricing unknown)"));
+        }
+        console.log(theme.dim("  " + "\u2500".repeat(48)));
         if (session) {
-          console.log(theme.dim("  Provider      : ") + theme.dim(ctx.getCurrentProvider()));
-          console.log(theme.dim("  Model         : ") + theme.dim(ctx.getCurrentModel()));
-          console.log(theme.dim("  Messages      : ") + theme.dim(String(session.messages.length)));
+          console.log(theme.dim("  Provider         : ") + theme.dim(provider));
+          console.log(theme.dim("  Model            : ") + theme.dim(model));
+          console.log(theme.dim("  Messages         : ") + theme.dim(String(session.messages.length)));
         }
         console.log();
       }
@@ -2106,7 +2161,7 @@ ${hint}` : "")
       usage: "/test [command|filter]",
       async execute(args, ctx) {
         try {
-          const { executeTests } = await import("./run-tests-P53FNUJY.js");
+          const { executeTests } = await import("./run-tests-X4PCLXA2.js");
           const argStr = args.join(" ").trim();
           let testArgs = {};
           if (argStr) {
@@ -3344,7 +3399,14 @@ var Repl = class {
   /** 当前加载的层级上下文（全局/项目/子目录） */
   contextLayers = [];
   /** 本次会话累计 token 用量 */
-  sessionTokenUsage = { inputTokens: 0, outputTokens: 0 };
+  sessionTokenUsage = { inputTokens: 0, outputTokens: 0, cacheCreationTokens: 0, cacheReadTokens: 0 };
+  /** Fold a single-request TokenUsage (with optional cache fields) into sessionTokenUsage. */
+  addSessionUsage(u) {
+    this.sessionTokenUsage.inputTokens += u.inputTokens;
+    this.sessionTokenUsage.outputTokens += u.outputTokens;
+    this.sessionTokenUsage.cacheCreationTokens += u.cacheCreationTokens ?? 0;
+    this.sessionTokenUsage.cacheReadTokens += u.cacheReadTokens ?? 0;
+  }
   /** 启动时检测到的 Git 分支（无 git 仓库时为 null） */
   gitBranch = null;
   /** MCP 多服务器管理器（无 MCP 配置时为 null） */
@@ -4607,8 +4669,7 @@ Session '${this.resumeSessionId}' not found.
         session.addMessage({ role: "assistant", content, timestamp: /* @__PURE__ */ new Date() });
         this.events.emit("message.after", { content });
         if (usage) {
-          this.sessionTokenUsage.inputTokens += usage.inputTokens;
-          this.sessionTokenUsage.outputTokens += usage.outputTokens;
+          this.addSessionUsage(usage);
           session.addTokenUsage(usage);
           if (showTokens && !tokensShown) {
             this.renderer.renderUsage(usage, this.sessionTokenUsage);
@@ -4637,8 +4698,7 @@ Session '${this.resumeSessionId}' not found.
         session.addMessage({ role: "assistant", content: response.content, timestamp: /* @__PURE__ */ new Date() });
         this.events.emit("message.after", { content: response.content });
         if (response.usage) {
-          this.sessionTokenUsage.inputTokens += response.usage.inputTokens;
-          this.sessionTokenUsage.outputTokens += response.usage.outputTokens;
+          this.addSessionUsage(response.usage);
           session.addTokenUsage(response.usage);
           if (this.shouldShowTokens()) {
             this.renderer.renderUsage(response.usage, this.sessionTokenUsage);
@@ -4797,7 +4857,7 @@ You have a maximum of ${maxToolRounds} tool call rounds for this task. Plan effi
     const modelParams = this.getModelParams();
     const useStreaming = this.config.get("ui").streaming;
     const spinner = this.renderer.showSpinner("Thinking...");
-    const roundUsage = { inputTokens: 0, outputTokens: 0 };
+    const roundUsage = { inputTokens: 0, outputTokens: 0, cacheCreationTokens: 0, cacheReadTokens: 0 };
     const supportsStreamingTools = useStreaming && typeof provider.chatWithToolsStream === "function";
     let consecutiveFreeRounds = 0;
     let lastToolCallSignature = "";
@@ -4904,6 +4964,8 @@ You have a maximum of ${maxToolRounds} tool call rounds for this task. Plan effi
         if (result.usage) {
           roundUsage.inputTokens += result.usage.inputTokens;
           roundUsage.outputTokens += result.usage.outputTokens;
+          roundUsage.cacheCreationTokens += result.usage.cacheCreationTokens ?? 0;
+          roundUsage.cacheReadTokens += result.usage.cacheReadTokens ?? 0;
         }
         if ("content" in result) {
           const hasWriteTools = toolDefs.some((t) => t.name === "write_file" || t.name === "edit_file");
@@ -4954,8 +5016,7 @@ You have a maximum of ${maxToolRounds} tool call rounds for this task. Plan effi
           });
           this.events.emit("message.after", { content: finalContent });
           if (roundUsage.inputTokens > 0 || roundUsage.outputTokens > 0) {
-            this.sessionTokenUsage.inputTokens += roundUsage.inputTokens;
-            this.sessionTokenUsage.outputTokens += roundUsage.outputTokens;
+            this.addSessionUsage(roundUsage);
             session.addTokenUsage(roundUsage);
             if (this.shouldShowTokens()) {
               this.renderer.renderUsage(roundUsage, this.sessionTokenUsage);
@@ -4993,6 +5054,8 @@ You have a maximum of ${maxToolRounds} tool call rounds for this task. Plan effi
               if (genUsage) {
                 roundUsage.inputTokens += genUsage.inputTokens;
                 roundUsage.outputTokens += genUsage.outputTokens;
+                roundUsage.cacheCreationTokens += genUsage.cacheCreationTokens ?? 0;
+                roundUsage.cacheReadTokens += genUsage.cacheReadTokens ?? 0;
               }
               session.addMessage({ role: "assistant", content: genContent, timestamp: /* @__PURE__ */ new Date() });
               this.events.emit("message.after", { content: genContent });
@@ -5007,8 +5070,7 @@ You have a maximum of ${maxToolRounds} tool call rounds for this task. Plan effi
               const newMsgs2 = provider.buildToolResultMessages(result.toolCalls, syntheticResults, reasoningContent2);
               extraMessages.push(...newMsgs2);
               if (roundUsage.inputTokens > 0 || roundUsage.outputTokens > 0) {
-                this.sessionTokenUsage.inputTokens += roundUsage.inputTokens;
-                this.sessionTokenUsage.outputTokens += roundUsage.outputTokens;
+                this.addSessionUsage(roundUsage);
                 session.addTokenUsage(roundUsage);
                 if (teeShowTokens && !teeTokShown) {
                   this.renderer.renderUsage(roundUsage, this.sessionTokenUsage);
@@ -5204,6 +5266,8 @@ You have a maximum of ${maxToolRounds} tool call rounds for this task. Plan effi
           if (summaryResult.usage) {
             roundUsage.inputTokens += summaryResult.usage.inputTokens;
             roundUsage.outputTokens += summaryResult.usage.outputTokens;
+            roundUsage.cacheCreationTokens += summaryResult.usage.cacheCreationTokens ?? 0;
+            roundUsage.cacheReadTokens += summaryResult.usage.cacheReadTokens ?? 0;
           }
         } else {
           this.renderer.renderError(
@@ -5218,8 +5282,7 @@ Tip: You can continue the conversation by asking the AI to proceed.`
         );
       }
       if (roundUsage.inputTokens > 0 || roundUsage.outputTokens > 0) {
-        this.sessionTokenUsage.inputTokens += roundUsage.inputTokens;
-        this.sessionTokenUsage.outputTokens += roundUsage.outputTokens;
+        this.addSessionUsage(roundUsage);
         session.addTokenUsage(roundUsage);
         if (this.shouldShowTokens()) {
           this.renderer.renderUsage(roundUsage, this.sessionTokenUsage);
@@ -5316,7 +5379,12 @@ Tip: You can continue the conversation by asking the AI to proceed.`
       },
       getSessionTokenUsage: () => ({ ...this.sessionTokenUsage }),
       resetSessionTokenUsage: () => {
-        this.sessionTokenUsage = { inputTokens: 0, outputTokens: 0 };
+        this.sessionTokenUsage = {
+          inputTokens: 0,
+          outputTokens: 0,
+          cacheCreationTokens: 0,
+          cacheReadTokens: 0
+        };
       },
       getGitBranch: () => this.gitBranch,
       getLastResponse: () => lastResponseStore.content,
@@ -5493,7 +5561,7 @@ program.command("web").description("Start Web UI server with browser-based chat
     console.error("Error: Invalid port number. Must be between 1 and 65535.");
     process.exit(1);
   }
-  const { startWebServer } = await import("./server-BQHIMEBH.js");
+  const { startWebServer } = await import("./server-YPAZWGUE.js");
   await startWebServer({ port, host: options.host });
 });
 program.command("user [action] [username]").description("Manage Web UI users (list | create <name> | delete <name> | reset-password <name> | migrate <name>)").action(async (action, username) => {
@@ -5726,7 +5794,7 @@ program.command("hub [topic]").description("Start multi-agent hub (discuss / bra
     }),
     config.get("customProviders")
   );
-  const { startHub } = await import("./hub-6V54V4O3.js");
+  const { startHub } = await import("./hub-AUWP4SWJ.js");
   await startHub(
     {
       topic: topic ?? "",

package/dist/{run-tests-6G65OGSL.js → run-tests-I6UDHVIS.js}

@@ -1,7 +1,7 @@
 import {
   executeTests,
   runTestsTool
-} from "./chunk-FOFQAEU6.js";
+} from "./chunk-DJ342VFS.js";
 export {
   executeTests,
   runTestsTool

package/dist/{run-tests-P53FNUJY.js → run-tests-X4PCLXA2.js}

@@ -2,7 +2,7 @@
 import {
   executeTests,
   runTestsTool
-} from "./chunk-NP5KZVP6.js";
+} from "./chunk-W7QVBFIJ.js";
 export {
   executeTests,
   runTestsTool

package/dist/{server-BQHIMEBH.js → server-YPAZWGUE.js}

@@ -7,7 +7,9 @@ import {
   SessionManager,
   SkillManager,
   TOOL_CALL_REMINDER,
+  computeCost,
   detectsHallucinatedFileOp,
+  formatCost,
   formatGitContextForPrompt,
   getContentText,
   getGitContext,
@@ -15,7 +17,7 @@ import {
   hadPreviousWriteToolCalls,
   loadDevState,
   setupProxy
-} from "./chunk-6FYFVPVE.js";
+} from "./chunk-JL5NK6AR.js";
 import {
   AuthManager
 } from "./chunk-BYNY5JPB.js";
@@ -34,7 +36,7 @@ import {
   spawnAgentContext,
   truncateOutput,
   undoStack
-} from "./chunk-TAR67QTH.js";
+} from "./chunk-YQEIQJ6K.js";
 import "./chunk-4BKXL7SM.js";
 import {
   AGENTIC_BEHAVIOR_GUIDELINE,
@@ -54,7 +56,7 @@ import {
   SKILLS_DIR_NAME,
   VERSION,
   buildUserIdentityPrompt
-} from "./chunk-NP5KZVP6.js";
+} from "./chunk-W7QVBFIJ.js";
 
 // src/web/server.ts
 import express from "express";
@@ -483,7 +485,17 @@ var SessionHandler = class _SessionHandler {
   currentModel;
   planMode = false;
   runtimeThinking = null;
-  sessionTokenUsage = { inputTokens: 0, outputTokens: 0 };
+  sessionTokenUsage = { inputTokens: 0, outputTokens: 0, cacheCreationTokens: 0, cacheReadTokens: 0 };
+  /** Accumulate a TokenUsage (with optional cache fields) into sessionTokenUsage. */
+  addWebSessionUsage(u) {
+    this.sessionTokenUsage.inputTokens += u.inputTokens;
+    this.sessionTokenUsage.outputTokens += u.outputTokens;
+    this.sessionTokenUsage.cacheCreationTokens += u.cacheCreationTokens ?? 0;
+    this.sessionTokenUsage.cacheReadTokens += u.cacheReadTokens ?? 0;
+  }
+  resetWebSessionUsage() {
+    this.sessionTokenUsage = { inputTokens: 0, outputTokens: 0, cacheCreationTokens: 0, cacheReadTokens: 0 };
+  }
   abortController = null;
   userInterjection = null;
   processing = false;
@@ -547,6 +559,7 @@ var SessionHandler = class _SessionHandler {
       displayName: p.info.displayName,
       models: p.info.models.map((m) => ({ id: m.id, name: m.displayName ?? m.id }))
     }));
+    const costUsd = computeCost(this.currentProvider, this.currentModel, this.sessionTokenUsage);
     this.send({
       type: "status",
       provider: this.currentProvider,
@@ -557,6 +570,7 @@ var SessionHandler = class _SessionHandler {
       planMode: this.planMode,
       thinkingMode: this.runtimeThinking ?? false,
       tokenUsage: { ...this.sessionTokenUsage },
+      costUsd,
       providers: providerList
     });
   }
@@ -724,8 +738,7 @@ var SessionHandler = class _SessionHandler {
         if (chunk.done) {
           this.send({ type: "response_done", content: fullContent, usage: chunk.usage });
           if (chunk.usage) {
-            this.sessionTokenUsage.inputTokens += chunk.usage.inputTokens;
-            this.sessionTokenUsage.outputTokens += chunk.usage.outputTokens;
+            this.addWebSessionUsage(chunk.usage);
             session.addTokenUsage(chunk.usage);
           }
           break;
@@ -759,7 +772,7 @@ You have a maximum of ${maxToolRounds} tool call rounds for this task. Plan effi
 - When remaining rounds are low, focus on completing the current task and summarizing.`;
     const systemPrompt = baseSystemPrompt + roundBudgetHint;
     const modelParams = this.getModelParams();
-    const roundUsage = { inputTokens: 0, outputTokens: 0 };
+    const roundUsage = { inputTokens: 0, outputTokens: 0, cacheCreationTokens: 0, cacheReadTokens: 0 };
     const supportsStreamingTools = typeof provider.chatWithToolsStream === "function";
     let consecutiveFreeRounds = 0;
     const warnNoteAt = Math.max(10, Math.floor(maxToolRounds * 0.2));
@@ -829,6 +842,8 @@ You have a maximum of ${maxToolRounds} tool call rounds for this task. Plan effi
         if (result.usage) {
           roundUsage.inputTokens += result.usage.inputTokens;
           roundUsage.outputTokens += result.usage.outputTokens;
+          roundUsage.cacheCreationTokens += result.usage.cacheCreationTokens ?? 0;
+          roundUsage.cacheReadTokens += result.usage.cacheReadTokens ?? 0;
         }
         if (result.content && !result.toolCalls) {
           const hasWriteTools = toolDefs.some((t) => t.name === "write_file" || t.name === "edit_file");
@@ -843,8 +858,7 @@ You have a maximum of ${maxToolRounds} tool call rounds for this task. Plan effi
           }
           this.send({ type: "response_done", content: result.content, usage: roundUsage });
           session.addMessage({ role: "assistant", content: result.content, timestamp: /* @__PURE__ */ new Date() });
-          this.sessionTokenUsage.inputTokens += roundUsage.inputTokens;
-          this.sessionTokenUsage.outputTokens += roundUsage.outputTokens;
+          this.addWebSessionUsage(roundUsage);
           session.addTokenUsage(roundUsage);
           return;
         }
@@ -917,8 +931,7 @@ ${summaryResult.content}`,
           message: `Reached maximum tool call rounds (${maxToolRounds}). You can continue by asking the AI to proceed.`
         });
       }
-      this.sessionTokenUsage.inputTokens += roundUsage.inputTokens;
-      this.sessionTokenUsage.outputTokens += roundUsage.outputTokens;
+      this.addWebSessionUsage(roundUsage);
       session.addTokenUsage(roundUsage);
     } catch (err) {
       if (err.name === "AbortError") {
@@ -1034,7 +1047,7 @@ ${summaryResult.content}`,
       case "clear":
         this.saveIfNeeded();
         this.sessions.createSession(this.currentProvider, this.currentModel);
-        this.sessionTokenUsage = { inputTokens: 0, outputTokens: 0 };
+        this.resetWebSessionUsage();
         this.send({ type: "info", message: "Conversation cleared." });
         this.sendStatus();
         this.sendSessionList();
@@ -1068,12 +1081,19 @@ ${summaryResult.content}`,
       }
       case "status": {
         const session = this.sessions.current;
+        const cacheRead = this.sessionTokenUsage.cacheReadTokens;
+        const cacheCreate = this.sessionTokenUsage.cacheCreationTokens;
+        const cost = computeCost(this.currentProvider, this.currentModel, this.sessionTokenUsage);
+        const cacheLine = cacheRead > 0 || cacheCreate > 0 ? `
+Cache:  write=${cacheCreate} read=${cacheRead}` : "";
+        const costLine = cost != null ? `
+Cost:   ${formatCost(cost)}` : "";
         this.send({
           type: "info",
           message: `Provider: ${this.currentProvider}
 Model: ${this.currentModel}
 Session: ${session?.id ?? "none"} (${session?.messages.length ?? 0} messages)
-Tokens: in=${this.sessionTokenUsage.inputTokens} out=${this.sessionTokenUsage.outputTokens}`
+Tokens: in=${this.sessionTokenUsage.inputTokens} out=${this.sessionTokenUsage.outputTokens}${cacheLine}${costLine}`
         });
         break;
       }
@@ -1083,7 +1103,7 @@ Tokens: in=${this.sessionTokenUsage.inputTokens} out=${this.sessionTokenUsage.ou
           this.saveIfNeeded();
           const created = this.sessions.createSession(this.currentProvider, this.currentModel);
           this.unsavedSessions.set(created.id, created);
-          this.sessionTokenUsage = { inputTokens: 0, outputTokens: 0 };
+          this.resetWebSessionUsage();
           this.send({ type: "info", message: "New session created." });
           this.sendStatus();
           this.sendSessionList();
@@ -1094,7 +1114,7 @@ Tokens: in=${this.sessionTokenUsage.inputTokens} out=${this.sessionTokenUsage.ou
           const cached = cachedExact ?? [...this.unsavedSessions.values()].find((s) => s.id.startsWith(targetId));
           if (cached) {
             this.sessions.setCurrent(cached);
-            this.sessionTokenUsage = { inputTokens: 0, outputTokens: 0 };
+            this.resetWebSessionUsage();
             this.send({
               type: "info",
               message: `Loaded session: ${cached.id.slice(0, 8)} "${cached.title ?? ""}" (${cached.messages.length} messages)`
@@ -1108,7 +1128,7 @@ Tokens: in=${this.sessionTokenUsage.inputTokens} out=${this.sessionTokenUsage.ou
           const found = list.find((s) => s.id.startsWith(targetId));
           if (found) {
             this.sessions.loadSession(found.id);
-            this.sessionTokenUsage = { inputTokens: 0, outputTokens: 0 };
+            this.resetWebSessionUsage();
             this.send({ type: "info", message: `Loaded session: ${found.id.slice(0, 8)} "${found.title ?? ""}" (${found.messageCount} messages)` });
             this.sendSessionMessages();
             this.sendStatus();
@@ -1116,7 +1136,7 @@ Tokens: in=${this.sessionTokenUsage.inputTokens} out=${this.sessionTokenUsage.ou
           } else {
             const recreated = this.sessions.createSession(this.currentProvider, this.currentModel);
             this.unsavedSessions.set(recreated.id, recreated);
-            this.sessionTokenUsage = { inputTokens: 0, outputTokens: 0 };
+            this.resetWebSessionUsage();
             this.send({
               type: "info",
               message: `Previous session (${targetId.slice(0, 8)}) is no longer available \u2014 started a new one.`
@@ -1256,16 +1276,41 @@ Tokens: in=${this.sessionTokenUsage.inputTokens} out=${this.sessionTokenUsage.ou
         });
         break;
       case "cost": {
-        const total = this.sessionTokenUsage.inputTokens + this.sessionTokenUsage.outputTokens;
-        this.send({
-          type: "info",
-          message: `\u{1F4CA} Token Usage
-  Provider: ${this.currentProvider}
-  Model: ${this.currentModel}
-  Input:  ${this.sessionTokenUsage.inputTokens.toLocaleString()}
-  Output: ${this.sessionTokenUsage.outputTokens.toLocaleString()}
-  Total:  ${total.toLocaleString()}`
-        });
+        const u = this.sessionTokenUsage;
+        const total = u.inputTokens + u.outputTokens + u.cacheCreationTokens + u.cacheReadTokens;
+        const cost = computeCost(this.currentProvider, this.currentModel, u);
+        let savings = null;
+        if (cost != null && u.cacheReadTokens > 0) {
+          const withoutCache = computeCost(this.currentProvider, this.currentModel, {
+            inputTokens: u.inputTokens + u.cacheReadTokens,
+            outputTokens: u.outputTokens,
+            cacheCreationTokens: u.cacheCreationTokens,
+            cacheReadTokens: 0
+          });
+          if (withoutCache != null) savings = withoutCache - cost;
+        }
+        const lines = [
+          "\u{1F4B0} Session Cost & Token Usage",
+          `  Provider        : ${this.currentProvider}`,
+          `  Model           : ${this.currentModel}`,
+          `  Input (uncached): ${u.inputTokens.toLocaleString()}`,
+          `  Output          : ${u.outputTokens.toLocaleString()}`
+        ];
+        if (u.cacheCreationTokens > 0) lines.push(`  Cache write     : ${u.cacheCreationTokens.toLocaleString()}`);
+        if (u.cacheReadTokens > 0) {
+          const pct = Math.round(u.cacheReadTokens / (u.cacheReadTokens + u.inputTokens) * 100);
+          lines.push(`  Cache read      : ${u.cacheReadTokens.toLocaleString()} (${pct}% hit rate)`);
+        }
+        lines.push(`  Total tokens    : ${total.toLocaleString()}`);
+        if (cost != null) {
+          lines.push(`  Cost            : ${formatCost(cost)}`);
+          if (savings != null && savings > 0) {
+            lines.push(`  Cache savings   : -${formatCost(savings)} (vs no cache)`);
+          }
+        } else {
+          lines.push(`  Cost            : \u2014 (pricing unknown for this model)`);
+        }
+        this.send({ type: "info", message: lines.join("\n") });
         break;
       }
       case "tools":
@@ -1691,7 +1736,7 @@ ${undoResults.map((r) => `  \u2022 ${r}`).join("\n")}` });
       case "test": {
         this.send({ type: "info", message: "\u{1F9EA} Running tests..." });
         try {
-          const { executeTests } = await import("./run-tests-P53FNUJY.js");
+          const { executeTests } = await import("./run-tests-X4PCLXA2.js");
           const argStr = args.join(" ").trim();
           let testArgs = {};
           if (argStr) {

package/dist/{task-orchestrator-TSY7CJE6.js → task-orchestrator-MWO6A4KQ.js}

@@ -4,11 +4,11 @@ import {
   getDangerLevel,
   googleSearchContext,
   truncateOutput
-} from "./chunk-TAR67QTH.js";
+} from "./chunk-YQEIQJ6K.js";
 import "./chunk-4BKXL7SM.js";
 import {
   SUBAGENT_ALLOWED_TOOLS
-} from "./chunk-NP5KZVP6.js";
+} from "./chunk-W7QVBFIJ.js";
 
 // src/hub/task-orchestrator.ts
 import { createInterface } from "readline";

package/dist/web/client/app.js

@@ -496,7 +496,16 @@ function handleStatus(msg) {
     btnPlan.classList.toggle('btn-active-toggle', msg.planMode);
     statusSession.textContent = `📋 ${msg.sessionId?.slice(0, 8) || '—'} (${msg.messageCount} msgs)`;
     if (msg.tokenUsage) {
-      statusTokens.textContent = `📊 in: ${msg.tokenUsage.inputTokens}  out: ${msg.tokenUsage.outputTokens}`;
+      const u = msg.tokenUsage;
+      const cacheRead = u.cacheReadTokens || 0;
+      let line = `📊 in: ${u.inputTokens.toLocaleString()}  out: ${u.outputTokens.toLocaleString()}`;
+      if (cacheRead > 0) line += `  cache: ${cacheRead.toLocaleString()}`;
+      if (msg.costUsd != null) {
+        const cost = msg.costUsd;
+        const costStr = cost === 0 ? '$0' : cost < 0.01 ? `$${cost.toFixed(4)}` : cost < 1 ? `$${cost.toFixed(3)}` : `$${cost.toFixed(2)}`;
+        line += `  💰 ${costStr}`;
+      }
+      statusTokens.textContent = line;
     }
     sessionListEl.querySelectorAll('.session-item').forEach(el => {
       el.classList.toggle('active', el.dataset.sessionId === msg.sessionId);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "jinzd-ai-cli",
-  "version": "0.4.54",
+  "version": "0.4.55",
   "description": "Cross-platform REPL-style AI CLI with multi-provider support",
   "type": "module",
   "main": "./dist/index.js",