jinzd-ai-cli 0.4.264 → 0.4.266

package/dist/electron-server.js

@@ -34,14 +34,14 @@ import {
   touchMemoryReferences,
   updateMemoryApproval,
   updateMemoryEntry
-} from "./chunk-UQP2BNQT.js";
+} from "./chunk-GWOZET2E.js";
 import {
   redactJson,
   scanString
 } from "./chunk-YVTASHS5.js";
 import {
   runTestsTool
-} from "./chunk-TANTNIGP.js";
+} from "./chunk-LHY7ST6X.js";
 import {
   AGENTIC_BEHAVIOR_GUIDELINE,
   APP_NAME,
@@ -73,7 +73,7 @@ import {
   SUBAGENT_MAX_ROUNDS_LIMIT,
   VERSION,
   buildUserIdentityPrompt
-} from "./chunk-4T4KHPRD.js";
+} from "./chunk-FA3HGXXM.js";
 import {
   hasSemanticIndex,
   semanticSearch
@@ -725,6 +725,16 @@ var BaseProvider = class {
   get supportsContentOnlyTee() {
     return this.reliableContentOnlyTee;
   }
+  /**
+   * 该 provider 是否支持 `reasoning_effort` 档位（low/high/max）。目前只有 DeepSeek V4。
+   *
+   * 2026-08-14 审计 P2-03：提示符标签此前无条件渲染成 `[THINK:high]`，于是 Claude
+   * （用 thinkingBudget 计 token 预算）和 GLM（根本没有 effort 概念）也显示一个
+   * 既不生效、也无从设置的档位——屏幕上的假信息比没有信息更坏。
+   */
+  get supportsReasoningEffort() {
+    return false;
+  }
   /**
    * 将 Message[] 转换为 OpenAI API 格式的消息数组。
    * content 为 string 时直接传递；为 MessageContentPart[] 时保留数组格式（vision 请求）。
@@ -2991,10 +3001,12 @@ var OpenAICompatibleProvider = class extends BaseProvider {
       if (!firstChoice) {
         return { content: "", model: response.model, usage: void 0 };
       }
+      const reasoningContent = firstChoice.message.reasoning_content;
       return {
         content: firstChoice.message.content ?? "",
         model: response.model,
-        usage: toUsage(response.usage)
+        usage: toUsage(response.usage),
+        ...reasoningContent ? { reasoningContent } : {}
       };
     } catch (err) {
       throw this.wrapError(err);
@@ -3391,6 +3403,10 @@ var DeepSeekProvider = class extends OpenAICompatibleProvider {
       ...thinkingEnabled ? { reasoning_effort: request.thinkingEffort ?? "high" } : {}
     };
   }
+  /** V4 支持 low/high/max 三档 reasoning_effort（官方 2026-08 更新公告）。 */
+  get supportsReasoningEffort() {
+    return true;
+  }
   /** 禁用流式工具调用，确保 chatWithTools 覆写（代码块检测）生效 */
   enableStreamingToolCalls = false;
   /**
@@ -13792,6 +13808,17 @@ function formatCost(amount) {
   return `$${amount.toFixed(2)}`;
 }
 
+// src/core/chat-params.ts
+function thinkingParamsFor(allModelParams, model) {
+  const params = allModelParams?.[model];
+  if (!params) return {};
+  const out = {};
+  if (params.thinking !== void 0) out.thinking = params.thinking;
+  if (params.thinkingEffort !== void 0) out.thinkingEffort = params.thinkingEffort;
+  if (params.thinkingBudget !== void 0) out.thinkingBudget = params.thinkingBudget;
+  return out;
+}
+
 // src/session/tool-history.ts
 var SESSION_SIZE_LIMIT = 2 * 1024 * 1024;
 function persistToolRound(session, toolCalls, toolResults, opts) {
@@ -14264,6 +14291,11 @@ var HubAgent = class {
   defaultModel;
   /** External context documents injected into system prompt */
   context;
+  /**
+   * 该 agent 模型的思考参数（审计 P0-01）。HubAgent 自己拿不到 ConfigManager，
+   * 由构造方（hub-server / discuss）按 modelId 解析后注入；不注入则沿用产品默认（不思考）。
+   */
+  thinking;
   constructor(role, providers, defaultProvider, defaultModel, context) {
     this.role = role;
     this.providers = providers;
@@ -14295,7 +14327,8 @@ var HubAgent = class {
       systemPrompt,
       stream: false,
       temperature: 0.7,
-      maxTokens: 4096
+      maxTokens: 4096,
+      ...this.thinking ?? {}
     });
     const { content, passed, converged } = parseTurn(response.content);
     return {
@@ -14328,7 +14361,8 @@ var HubAgent = class {
       systemPrompt,
       stream: true,
       temperature: 0.7,
-      maxTokens: 4096
+      maxTokens: 4096,
+      ...this.thinking ?? {}
     });
     for await (const chunk of stream) {
       if (chunk.delta) {
@@ -14366,7 +14400,8 @@ var HubAgent = class {
       model: this.modelId,
       stream: false,
       temperature: 0.3,
-      maxTokens: 4096
+      maxTokens: 4096,
+      ...this.thinking ?? {}
     });
     return response.content.trim();
   }
@@ -14470,7 +14505,11 @@ var DiscussionOrchestrator = class {
   onRoundReview;
   constructor(config, providers) {
     this.agents = config.roles.map(
-      (role) => new HubAgent(role, providers, config.defaultProvider, config.defaultModel, config.context)
+      (role) => {
+        const agent = new HubAgent(role, providers, config.defaultProvider, config.defaultModel, config.context);
+        agent.thinking = config.thinking;
+        return agent;
+      }
     );
     this.humanSteer = config.humanSteer ?? false;
     this.voteConverge = config.voteConverge ?? false;
@@ -14842,6 +14881,68 @@ async function persistDiscussion(state2, config, defaultProvider, defaultModel)
   return { id: session.id, path: join20(config.getHistoryDir(), `${session.id}.json`) };
 }
 
+// src/core/think-command.ts
+var EFFORTS = /* @__PURE__ */ new Set(["low", "high", "max"]);
+function resolveEffective(state2) {
+  if (state2.runtimeThinking !== null) return { effective: state2.runtimeThinking, origin: "runtime override" };
+  if (state2.configuredThinking !== void 0) return { effective: state2.configuredThinking, origin: "config" };
+  return { effective: false, origin: "default" };
+}
+function effortLine(state2, effort) {
+  if (!state2.supportsEffort) return `Reasoning effort: n/a for ${state2.provider}`;
+  const resolved = effort ?? state2.runtimeEffort ?? state2.configuredEffort;
+  return `Reasoning effort: ${resolved ?? "high (provider default)"}`;
+}
+function applyThinkCommand(sub, state2) {
+  const key = (sub ?? "toggle").toLowerCase();
+  const current = resolveEffective(state2);
+  const stay = (summary, details = []) => ({
+    runtimeThinking: state2.runtimeThinking,
+    runtimeEffort: state2.runtimeEffort,
+    effective: current.effective,
+    origin: current.origin,
+    summary,
+    details,
+    changed: false
+  });
+  const enable = (effort) => ({
+    runtimeThinking: true,
+    runtimeEffort: effort ?? state2.runtimeEffort,
+    effective: true,
+    origin: "runtime override",
+    summary: effort && state2.supportsEffort ? `Extended Thinking: ON (effort: ${effort})` : "Extended Thinking: ON",
+    details: effort && !state2.supportsEffort ? [`Note: ${state2.provider} has no reasoning-effort levels \u2014 the level is ignored.`] : [],
+    changed: true
+  });
+  const disable = () => ({
+    runtimeThinking: false,
+    // 刻意保留档位：重新开启时不该丢掉用户选过的档
+    runtimeEffort: state2.runtimeEffort,
+    effective: false,
+    origin: "runtime override",
+    summary: "Extended Thinking: OFF",
+    details: [],
+    changed: true
+  });
+  if (key === "status") {
+    const details = current.effective ? [effortLine(state2, null)] : [
+      "Enable with /think on | low | high | max, or set",
+      "modelParams.<model>.thinking in ~/.aicli/config.json"
+    ];
+    return stay(
+      `Extended Thinking: ${current.effective ? "ON" : "OFF"} (${current.origin})`,
+      details
+    );
+  }
+  if (key === "on") return enable(null);
+  if (key === "off") return disable();
+  if (EFFORTS.has(key)) return enable(key);
+  if (key === "toggle") {
+    return current.effective ? disable() : enable(null);
+  }
+  return stay(`Unknown /think option "${sub}" \u2014 expected on | off | low | high | max | status`);
+}
+
 // src/web/commands/provider-commands.ts
 async function handleProvider(args, ctx) {
   const id = args[0];
@@ -14904,25 +15005,24 @@ async function handleModel(args, ctx) {
   ctx.sendStatus();
 }
 async function handleThink(args, ctx) {
-  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;
+  const modelParams = ctx.config.get("modelParams")[ctx.currentModel];
+  const outcome = applyThinkCommand(args[0], {
+    runtimeThinking: ctx.runtimeThinking,
+    runtimeEffort: ctx.runtimeThinkingEffort,
+    configuredThinking: modelParams?.thinking,
+    configuredEffort: modelParams?.thinkingEffort,
+    supportsEffort: ctx.providers.get(ctx.currentProvider)?.supportsReasoningEffort ?? false,
+    provider: ctx.currentProvider
+  });
+  if (outcome.changed) {
+    ctx.runtimeThinking = outcome.runtimeThinking;
+    ctx.runtimeThinkingEffort = outcome.runtimeEffort;
   }
-  if (sub === "on") ctx.runtimeThinking = true;
-  else if (sub === "off") ctx.runtimeThinking = false;
-  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 ? ` (${ctx.runtimeThinkingEffort ?? "high"})` : ""}`
+    message: [outcome.summary, ...outcome.details].join(" \u2014 ")
   });
-  ctx.sendStatus();
+  if (outcome.changed) ctx.sendStatus();
 }
 async function handlePlan(args, ctx) {
   const sub = args[0];
@@ -15434,7 +15534,7 @@ Cache:  write=${cacheCreate} read=${cacheRead}` : "";
 Cost:   ${formatCost(cost)}` : "";
   let memoryLine = "";
   try {
-    const { countPendingMemories } = await import("./persistent-memory-QX6C2JAR.js");
+    const { countPendingMemories } = await import("./persistent-memory-YPO25MOP.js");
     const pending = countPendingMemories(ctx.config.getConfigDir());
     if (pending > 0) memoryLine = `
 Memory: \u26A0 ${pending} pending approval \u2014 see the Memory panel or /memory`;
@@ -16249,7 +16349,7 @@ async function handleTest(args, ctx) {
       const isCommand = argStr.includes(" ") || /^(mvn|gradle|npm|pytest|cargo|go)\b/.test(argStr);
       testArgs = isCommand ? { command: argStr } : { filter: argStr };
     }
-    const runTests = ctx.runTests ?? (await import("./run-tests-CK44KRY3.js")).executeTests;
+    const runTests = ctx.runTests ?? (await import("./run-tests-4XRHKHN5.js")).executeTests;
     const report = await runTests(testArgs);
     ctx.send({ type: "info", message: report });
   } catch (err) {
@@ -16507,7 +16607,7 @@ 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-QX6C2JAR.js");
+      const { exportMemoryEntries: exportMemoryEntries2 } = await import("./persistent-memory-YPO25MOP.js");
       ctx.send({
         type: "export_data",
         format: "md",
@@ -17290,6 +17390,7 @@ var SessionHandler = class {
     const roles = resolution.roles;
     for (const w of resolution.warnings) this.send({ type: "info", message: w });
     const config = {
+      thinking: thinkingParamsFor(this.config.get("modelParams"), this.currentModel),
       mode: "discuss",
       roles,
       defaultProvider,

package/dist/{hub-P7D6VDCK.js → hub-4APIRQ73.js}

@@ -9,8 +9,11 @@ import {
   getPreset,
   listPresets,
   resolveRoleProviders
-} from "./chunk-VT6NRDD2.js";
-import "./chunk-3ARPAVNV.js";
+} from "./chunk-23SJVOQD.js";
+import "./chunk-VOIOUJXI.js";
+import {
+  thinkingParamsFor
+} from "./chunk-6YZEZFX5.js";
 
 // src/hub/index.ts
 import { readFileSync, existsSync } from "fs";
@@ -129,6 +132,7 @@ ${content}`);
     maxToolRoundsPerTurn: mode === "task" ? options.taskRounds ?? 30 : void 0,
     context,
     contextFiles: contextFileNames.length > 0 ? contextFileNames : void 0,
+    thinking: thinkingParamsFor(configManager.get("modelParams"), defaultModel),
     humanSteer: options.steer === true,
     voteConverge: options.vote === true
   };
@@ -154,7 +158,7 @@ ${content}`);
   }
 }
 async function runTaskMode(config, providers, configManager, topic) {
-  const { TaskOrchestrator } = await import("./task-orchestrator-5RT5EH2J.js");
+  const { TaskOrchestrator } = await import("./task-orchestrator-WNV6PT6Z.js");
   const orchestrator = new TaskOrchestrator(config, providers, configManager);
   let interrupted = false;
   const onSigint = () => {
@@ -190,7 +194,7 @@ async function runTaskMode(config, providers, configManager, topic) {
   }
 }
 async function runDistributedDiscussion(config, providers, topic, port) {
-  const { HubServer } = await import("./hub-server-QGUSMN22.js");
+  const { HubServer } = await import("./hub-server-RF6LURXY.js");
   const hub = new HubServer(config, providers, port);
   let interrupted = false;
   const onSigint = () => {

package/dist/{hub-server-QGUSMN22.js → hub-server-RF6LURXY.js}

@@ -13,7 +13,7 @@ import {
 } from "./chunk-NR5U6LVG.js";
 import {
   HubAgent
-} from "./chunk-3ARPAVNV.js";
+} from "./chunk-VOIOUJXI.js";
 
 // src/hub/hub-server.ts
 import { WebSocketServer } from "ws";
@@ -261,6 +261,7 @@ var HubServer = class {
   }
   async getBuiltinResponse(role, topic, round, maxRounds) {
     const agent = new HubAgent(role, this.providers, this.config.defaultProvider, this.config.defaultModel, this.config.context);
+    agent.thinking = this.config.thinking;
     return agent.speak(topic, this.history, round, maxRounds);
   }
   async generateSummary(topic) {
@@ -274,6 +275,7 @@ var HubServer = class {
       this.config.defaultModel,
       this.config.context
     );
+    summarizer.thinking = this.config.thinking;
     try {
       console.log(chalk.dim("  Generating summary..."));
       const summary = await summarizer.summarize(topic, this.history);