jinzd-ai-cli 0.4.261 → 0.4.263

package/dist/electron-server.js

@@ -34,14 +34,14 @@ import {
   touchMemoryReferences,
   updateMemoryApproval,
   updateMemoryEntry
-} from "./chunk-T7R7ILYT.js";
+} from "./chunk-5K2LFD7D.js";
 import {
   redactJson,
   scanString
 } from "./chunk-YVTASHS5.js";
 import {
   runTestsTool
-} from "./chunk-KFNQESEG.js";
+} from "./chunk-7YCA2BD3.js";
 import {
   AGENTIC_BEHAVIOR_GUIDELINE,
   APP_NAME,
@@ -73,7 +73,7 @@ import {
   SUBAGENT_MAX_ROUNDS_LIMIT,
   VERSION,
   buildUserIdentityPrompt
-} from "./chunk-R6DXWVVN.js";
+} from "./chunk-3RHY2HF7.js";
 import {
   hasSemanticIndex,
   semanticSearch
@@ -128,6 +128,8 @@ var ModelParamsSchema = z.object({
   timeout: z.number().int().positive().optional(),
   /** 是否启用深度思考（thinking）模式，Claude Sonnet/Opus、GLM-5 等 */
   thinking: z.boolean().optional(),
+  /** DeepSeek V4 reasoning effort. */
+  thinkingEffort: z.enum(["low", "high", "max"]).optional(),
   /** thinking 模式的 token 预算（最小 1024，仅 Claude Extended Thinking 使用） */
   thinkingBudget: z.number().int().min(1024).optional()
 });
@@ -2784,6 +2786,7 @@ async function runLeanAgentLoop(opts) {
           maxTokens: opts.chatParams?.maxTokens,
           timeout: opts.chatParams?.timeout,
           thinking: opts.chatParams?.thinking,
+          thinkingEffort: opts.chatParams?.thinkingEffort,
           ...extraMessages.length > 0 ? { _extraMessages: extraMessages } : {}
         },
         opts.toolDefsForRound(round)
@@ -2954,6 +2957,14 @@ var OpenAICompatibleProvider = class extends BaseProvider {
     }
     return msgs;
   }
+  /** Provider-specific request fields shared by all four chat paths. */
+  buildChatRequestParams(request) {
+    return {
+      temperature: request.temperature,
+      max_tokens: request.maxTokens,
+      ...request.thinking ? { thinking: { type: "enabled" } } : {}
+    };
+  }
   /**
    * 非流式请求必须把 `request.signal` 交给 SDK —— 否则 Ctrl+C 完全无效。
    *
@@ -2970,10 +2981,8 @@ var OpenAICompatibleProvider = class extends BaseProvider {
       const response = await this.client.chat.completions.create({
         model: request.model,
         messages: this.buildMessages(request),
-        temperature: request.temperature,
-        max_tokens: request.maxTokens,
-        stream: false,
-        ...request.thinking ? { thinking: { type: "enabled" } } : {}
+        ...this.buildChatRequestParams(request),
+        stream: false
       }, {
         timeout: request.timeout ?? this.defaultTimeout,
         signal: request.signal
@@ -2996,12 +3005,10 @@ var OpenAICompatibleProvider = class extends BaseProvider {
       const stream = await this.client.chat.completions.create({
         model: request.model,
         messages: this.buildMessages(request),
-        temperature: request.temperature,
-        max_tokens: request.maxTokens,
+        ...this.buildChatRequestParams(request),
         stream: true,
         // 请求末尾 usage chunk，供 token 统计使用
-        stream_options: { include_usage: true },
-        ...request.thinking ? { thinking: { type: "enabled" } } : {}
+        stream_options: { include_usage: true }
       }, {
         timeout: request.timeout ?? this.defaultTimeout,
         signal: request.signal
@@ -3077,10 +3084,8 @@ var OpenAICompatibleProvider = class extends BaseProvider {
         messages: allMessages,
         tools: openaiTools,
         tool_choice: "auto",
-        temperature: request.temperature,
-        max_tokens: request.maxTokens,
-        stream: false,
-        ...request.thinking ? { thinking: { type: "enabled" } } : {}
+        ...this.buildChatRequestParams(request),
+        stream: false
       }, {
         timeout: request.timeout ?? this.defaultTimeout,
         // 非流式路径同样必须收 signal，否则 Ctrl+C 期间整个请求不可中断（见 chat() 上方注释）
@@ -3178,11 +3183,9 @@ var OpenAICompatibleProvider = class extends BaseProvider {
         messages: allMessages,
         tools: openaiTools,
         tool_choice: "auto",
-        temperature: request.temperature,
-        max_tokens: request.maxTokens,
+        ...this.buildChatRequestParams(request),
         stream: true,
-        stream_options: { include_usage: true },
-        ...request.thinking ? { thinking: { type: "enabled" } } : {}
+        stream_options: { include_usage: true }
       }, {
         timeout: request.timeout ?? this.defaultTimeout,
         signal: request.signal
@@ -3376,6 +3379,18 @@ Rules you MUST follow:
 - The system will detect and reject phantom claims. Each failed detection wastes a round. Be honest.`;
 var DeepSeekProvider = class extends OpenAICompatibleProvider {
   defaultBaseUrl = "https://api.deepseek.com/v1";
+  buildChatRequestParams(request) {
+    const thinkingEnabled = request.thinking === true;
+    return {
+      // DeepSeek ignores sampling controls in thinking mode; omit them to keep the
+      // wire request unambiguous. Explicitly disable thinking otherwise because
+      // V4 defaults it to enabled server-side.
+      temperature: thinkingEnabled ? void 0 : request.temperature,
+      max_tokens: request.maxTokens,
+      thinking: { type: thinkingEnabled ? "enabled" : "disabled" },
+      ...thinkingEnabled ? { reasoning_effort: request.thinkingEffort ?? "high" } : {}
+    };
+  }
   /** 禁用流式工具调用，确保 chatWithTools 覆写（代码块检测）生效 */
   enableStreamingToolCalls = false;
   /**
@@ -3396,14 +3411,14 @@ var DeepSeekProvider = class extends OpenAICompatibleProvider {
       // ── V4 family (2026-04-23+)：1M context，支持 Thinking / Non-Thinking 双模式 ──
       {
         id: "deepseek-v4-pro",
-        displayName: "DeepSeek V4 Pro (1.6T MoE, 49B active)",
+        displayName: "DeepSeek V4 Pro (0813, 1.6T MoE, 49B active)",
         contextWindow: 1048576,
         supportsStreaming: true,
         supportsThinking: true
       },
       {
         id: "deepseek-v4-flash",
-        displayName: "DeepSeek V4 Flash (284B MoE, 13B active)",
+        displayName: "DeepSeek V4 Flash (0731, 284B MoE, 13B active)",
         contextWindow: 1048576,
         supportsStreaming: true,
         supportsThinking: true
@@ -3991,6 +4006,18 @@ var BUILT_IN_PROVIDERS = [
   OpenRouterProvider,
   OllamaProvider
 ];
+function listBuiltInProviderMetadata() {
+  return BUILT_IN_PROVIDERS.map((ProviderClass) => {
+    const { info } = new ProviderClass();
+    return {
+      id: info.id,
+      displayName: info.displayName,
+      defaultModel: info.defaultModel,
+      apiKeyEnvVar: info.apiKeyEnvVar,
+      requiresApiKey: info.requiresApiKey
+    };
+  });
+}
 var ProviderRegistry = class {
   providers = /* @__PURE__ */ new Map();
   customConfigs = [];
@@ -4085,13 +4112,12 @@ var ProviderRegistry = class {
     return [...this.providers.values()];
   }
   listAll() {
-    const builtIn = BUILT_IN_PROVIDERS.map((P) => {
-      const p = new P();
+    const builtIn = listBuiltInProviderMetadata().map((info) => {
       return {
-        id: p.info.id,
-        displayName: p.info.displayName,
-        configured: this.providers.has(p.info.id),
-        defaultModel: p.info.defaultModel,
+        id: info.id,
+        displayName: info.displayName,
+        configured: this.providers.has(info.id),
+        defaultModel: info.defaultModel,
         isCustom: false
       };
     });
@@ -11066,7 +11092,8 @@ async function runSubAgent(task, maxRounds, agentIndex, ctx, agent) {
         temperature: ctx.modelParams.temperature,
         maxTokens: ctx.modelParams.maxTokens,
         timeout: ctx.modelParams.timeout,
-        thinking: ctx.modelParams.thinking
+        thinking: ctx.modelParams.thinking,
+        thinkingEffort: ctx.modelParams.thinkingEffort
       },
       executeTools: (calls) => subExecutor.executeAll(calls),
       systemPromptForRound: () => subSystemPrompt,
@@ -13624,6 +13651,29 @@ function loadMemoryContent(configDir, maxChars) {
 }
 
 // src/core/pricing.ts
+var DEEPSEEK_CNY_PER_USD = 6.9;
+var DEEPSEEK_TIME_PRICING_EFFECTIVE_AT = /* @__PURE__ */ new Date("2026-08-16T16:00:00.000Z");
+function deepSeekCnyPricing(cacheRead, input, output) {
+  return {
+    cacheRead: cacheRead / DEEPSEEK_CNY_PER_USD,
+    input: input / DEEPSEEK_CNY_PER_USD,
+    output: output / DEEPSEEK_CNY_PER_USD
+  };
+}
+function getDeepSeekV4Pricing(model, at) {
+  const isPro = model.startsWith("deepseek-v4-pro");
+  const isFlash = model.startsWith("deepseek-v4-flash");
+  if (!isPro && !isFlash) return null;
+  if (at < DEEPSEEK_TIME_PRICING_EFFECTIVE_AT) {
+    return isPro ? deepSeekCnyPricing(0.025, 3, 6) : deepSeekCnyPricing(0.02, 1, 2);
+  }
+  const beijingHour = (at.getUTCHours() + 8) % 24;
+  const isPeak = beijingHour >= 9 && beijingHour < 12 || beijingHour >= 14 && beijingHour < 18;
+  if (isPro) {
+    return isPeak ? deepSeekCnyPricing(0.3, 9, 27) : deepSeekCnyPricing(0.15, 4.5, 13.5);
+  }
+  return isPeak ? deepSeekCnyPricing(0.1, 3, 9) : deepSeekCnyPricing(0.05, 1.5, 4.5);
+}
 var PRICING_TABLE = {
   // ── Anthropic Claude ──────────────────────────────────────────
   "claude-opus-4-8": { input: 15, output: 75, cacheWrite: 18.75, cacheRead: 1.5 },
@@ -13667,8 +13717,6 @@ var PRICING_TABLE = {
   // 20% 拍脑袋而非抄文档）——`aicli usage` 因此系统性高报 DeepSeek 成本。
   // 待办：官方已公告峰谷定价（北京时间 9-12 点、14-18 点为平时 2 倍），本表仍是单一费率，
   //       届时高峰时段会低报；等该策略实际生效后再决定是否引入时段维度。
-  "deepseek-v4-pro": { input: 0.435, output: 0.87, cacheRead: 3625e-6 },
-  "deepseek-v4-flash": { input: 0.14, output: 0.28, cacheRead: 28e-4 },
   // Legacy aliases：retires 2026-07-24 UTC 15:59，官方 route 到 V4 Flash。
   // 这几条只服务历史 session 的成本回算，保留当时的费率口径，不随现价改动。
   "deepseek-chat": { input: 0.14, output: 0.28, cacheRead: 0.028 },
@@ -13712,11 +13760,15 @@ var PRICING_TABLE = {
   // Handled via provider check below (FREE_PROVIDERS → genuine $0.00, not "unknown").
 };
 var FREE_PROVIDERS = /* @__PURE__ */ new Set(["ollama"]);
-function getPricing(provider, model) {
+function getPricing(provider, model, at = /* @__PURE__ */ new Date()) {
   if (FREE_PROVIDERS.has(provider.toLowerCase())) {
     return { input: 0, output: 0 };
   }
   const key = model.toLowerCase();
+  if (provider.toLowerCase() === "deepseek") {
+    const dynamicPricing = getDeepSeekV4Pricing(key, at);
+    if (dynamicPricing) return dynamicPricing;
+  }
   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) {
@@ -13724,8 +13776,8 @@ function getPricing(provider, model) {
   }
   return null;
 }
-function computeCost(provider, model, usage) {
-  const p = getPricing(provider, model);
+function computeCost(provider, model, usage, at = /* @__PURE__ */ new Date()) {
+  const p = getPricing(provider, model, at);
   if (!p) return null;
   const input = usage.inputTokens * p.input;
   const output = usage.outputTokens * p.output;
@@ -14852,11 +14904,24 @@ async function handleModel(args, ctx) {
   ctx.sendStatus();
 }
 async function handleThink(args, ctx) {
-  const sub = args[0];
+  const sub = args[0]?.toLowerCase() ?? "toggle";
+  if (sub === "status") {
+    ctx.send({
+      type: "info",
+      message: `Thinking mode: ${ctx.runtimeThinking === null ? "CONFIG" : ctx.runtimeThinking ? "ON" : "OFF"}; effort: ${ctx.runtimeThinkingEffort ?? "config/default (high for DeepSeek V4)"}`
+    });
+    return;
+  }
   if (sub === "on") ctx.runtimeThinking = true;
   else if (sub === "off") ctx.runtimeThinking = false;
-  else ctx.runtimeThinking = !ctx.runtimeThinking;
-  ctx.send({ type: "info", message: `Thinking mode: ${ctx.runtimeThinking ? "ON" : "OFF"}` });
+  else if (sub === "low" || sub === "high" || sub === "max") {
+    ctx.runtimeThinking = true;
+    ctx.runtimeThinkingEffort = sub;
+  } else ctx.runtimeThinking = ctx.runtimeThinking !== true;
+  ctx.send({
+    type: "info",
+    message: `Thinking mode: ${ctx.runtimeThinking ? "ON" : "OFF"}${ctx.runtimeThinking ? ` (${ctx.runtimeThinkingEffort ?? "high"})` : ""}`
+  });
   ctx.sendStatus();
 }
 async function handlePlan(args, ctx) {
@@ -15369,7 +15434,7 @@ Cache:  write=${cacheCreate} read=${cacheRead}` : "";
 Cost:   ${formatCost(cost)}` : "";
   let memoryLine = "";
   try {
-    const { countPendingMemories } = await import("./persistent-memory-P3COUZZM.js");
+    const { countPendingMemories } = await import("./persistent-memory-ANIMO5CF.js");
     const pending = countPendingMemories(ctx.config.getConfigDir());
     if (pending > 0) memoryLine = `
 Memory: \u26A0 ${pending} pending approval \u2014 see the Memory panel or /memory`;
@@ -16178,14 +16243,14 @@ async function handleSecurityReview(args, ctx) {
 async function handleTest(args, ctx) {
   ctx.send({ type: "info", message: "\u{1F9EA} Running tests..." });
   try {
-    const { executeTests } = await import("./run-tests-7RYU4BK5.js");
     const argStr = args.join(" ").trim();
     let testArgs = {};
     if (argStr) {
       const isCommand = argStr.includes(" ") || /^(mvn|gradle|npm|pytest|cargo|go)\b/.test(argStr);
       testArgs = isCommand ? { command: argStr } : { filter: argStr };
     }
-    const report = await executeTests(testArgs);
+    const runTests = ctx.runTests ?? (await import("./run-tests-TF7UPUWW.js")).executeTests;
+    const report = await runTests(testArgs);
     ctx.send({ type: "info", message: report });
   } catch (err) {
     ctx.send({ type: "error", message: `Test failed: ${err.message}` });
@@ -16442,7 +16507,7 @@ async function handleMemory(args, ctx) {
     ctx.handleMemoryManage(sub, args[1], sub === "expire" ? args[2] : void 0);
   } else if (sub === "export") {
     if (args[1] === "md") {
-      const { exportMemoryEntries: exportMemoryEntries2 } = await import("./persistent-memory-P3COUZZM.js");
+      const { exportMemoryEntries: exportMemoryEntries2 } = await import("./persistent-memory-ANIMO5CF.js");
       ctx.send({
         type: "export_data",
         format: "md",
@@ -16873,6 +16938,7 @@ var SessionHandler = class {
   currentModel;
   planMode = false;
   runtimeThinking = null;
+  runtimeThinkingEffort = null;
   sessionTokenUsage = { inputTokens: 0, outputTokens: 0, cacheCreationTokens: 0, cacheReadTokens: 0 };
   unknownPricingWarnings = /* @__PURE__ */ new Set();
   /** Accumulate a TokenUsage (with optional cache fields) into sessionTokenUsage. */
@@ -17011,7 +17077,8 @@ var SessionHandler = class {
       cwd: process.cwd(),
       messageCount: this.sessions.current?.messages.length ?? 0,
       planMode: this.planMode,
-      thinkingMode: this.runtimeThinking ?? false,
+      thinkingMode: this.getModelParams().thinking ?? false,
+      thinkingEffort: this.runtimeThinkingEffort ?? this.getModelParams().thinkingEffort,
       permissionProfile: this.config.get("defaultPermissionProfile") ?? "legacy",
       tokenUsage: { ...this.sessionTokenUsage },
       costUsd,
@@ -17376,6 +17443,7 @@ var SessionHandler = class {
         maxTokens: modelParams.maxTokens,
         timeout: modelParams.timeout,
         thinking: modelParams.thinking,
+        thinkingEffort: modelParams.thinkingEffort,
         thinkingBudget: modelParams.thinkingBudget,
         signal: ac.signal
       });
@@ -17450,6 +17518,7 @@ ${mcpBudgetNote}` : "");
             maxTokens: modelParams.maxTokens,
             timeout: modelParams.timeout,
             thinking: modelParams.thinking,
+            thinkingEffort: modelParams.thinkingEffort,
             thinkingBudget: modelParams.thinkingBudget,
             signal: ac.signal,
             ...extraMessages.length > 0 ? { _extraMessages: extraMessages } : {}
@@ -17766,6 +17835,7 @@ ${summaryContent}`,
         maxTokens: modelParams.maxTokens,
         timeout: modelParams.timeout,
         thinking: modelParams.thinking,
+        thinkingEffort: modelParams.thinkingEffort,
         thinkingBudget: modelParams.thinkingBudget,
         signal: ac.signal,
         _extraMessages: teeExtraMessages
@@ -18300,6 +18370,7 @@ Backup saved: ${result.backupPath}` : "";
       ...params,
       maxTokens: params.maxTokens ?? DEFAULT_MAX_TOKENS,
       thinking: this.runtimeThinking ?? params.thinking,
+      thinkingEffort: this.runtimeThinkingEffort ?? params.thinkingEffort,
       thinkingBudget: params.thinkingBudget
     };
   }

package/dist/{hub-RCANK6SK.js → hub-6EYFGJRT.js}

@@ -154,7 +154,7 @@ ${content}`);
   }
 }
 async function runTaskMode(config, providers, configManager, topic) {
-  const { TaskOrchestrator } = await import("./task-orchestrator-ZFL4E6NZ.js");
+  const { TaskOrchestrator } = await import("./task-orchestrator-2RAKLFWN.js");
   const orchestrator = new TaskOrchestrator(config, providers, configManager);
   let interrupted = false;
   const onSigint = () => {