jinzd-ai-cli 0.4.54 → 0.4.56

package/dist/{server-BQHIMEBH.js → server-MOTFJN6L.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-2DC5ABAM.js";
 import {
   AuthManager
 } from "./chunk-BYNY5JPB.js";
@@ -34,7 +36,7 @@ import {
   spawnAgentContext,
   truncateOutput,
   undoStack
-} from "./chunk-TAR67QTH.js";
+} from "./chunk-FJSEFQ54.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-7FOGK5TM.js";
 
 // src/web/server.ts
 import express from "express";
@@ -483,12 +485,24 @@ 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;
   /** Pending ask_user promises */
   pendingAskUser = /* @__PURE__ */ new Map();
+  /** Pending auto-pause promises */
+  pendingAutoPause = /* @__PURE__ */ new Map();
   /** Active system prompt from context files */
   activeSystemPrompt;
   /** Directories added via /add-dir */
@@ -547,6 +561,7 @@ 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 +572,7 @@ var SessionHandler = class _SessionHandler {
       planMode: this.planMode,
       thinkingMode: this.runtimeThinking ?? false,
       tokenUsage: { ...this.sessionTokenUsage },
+      costUsd,
       providers: providerList
     });
   }
@@ -588,6 +604,14 @@ var SessionHandler = class _SessionHandler {
         }
         return;
       }
+      case "auto_pause_response": {
+        const resolve3 = this.pendingAutoPause.get(msg.requestId);
+        if (resolve3) {
+          this.pendingAutoPause.delete(msg.requestId);
+          resolve3({ action: msg.action, message: msg.message });
+        }
+        return;
+      }
       case "abort":
         if (this.abortController) {
           this.abortController.abort();
@@ -603,6 +627,8 @@ var SessionHandler = class _SessionHandler {
     if (this.abortController) this.abortController.abort();
     for (const resolve3 of this.pendingAskUser.values()) resolve3(null);
     this.pendingAskUser.clear();
+    for (const resolve3 of this.pendingAutoPause.values()) resolve3({ action: "stop" });
+    this.pendingAutoPause.clear();
     this.saveIfNeeded();
   }
   /** 根据当前模型 context window 更新工具输出截断上限 */
@@ -724,8 +750,7 @@ 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;
@@ -749,19 +774,24 @@ var SessionHandler = class _SessionHandler {
     const apiMessages = [...messages];
     const extraMessages = [];
     const maxToolRounds = this.config.get("maxToolRounds") ?? DEFAULT_MAX_TOOL_ROUNDS;
+    const autoPauseIntervalRaw = this.config.get("autoPauseInterval");
+    const autoPauseInterval = typeof autoPauseIntervalRaw === "number" ? autoPauseIntervalRaw : 50;
     const baseSystemPrompt = (this.buildSystemPrompt() ?? "") + TOOL_CALL_REMINDER;
+    const pauseHint = autoPauseInterval > 0 ? `
+- Every ${autoPauseInterval} rounds the user will be asked whether to continue \u2014 use this as a natural checkpoint to report progress.` : "";
     const roundBudgetHint = `
 
 [Tool Round Budget]
 You have a maximum of ${maxToolRounds} tool call rounds for this task. Plan efficiently:
 - Prefer batch operations (e.g. global find-and-replace) over repetitive single edits.
 - Prioritize the most critical tasks first in case rounds run out.
-- When remaining rounds are low, focus on completing the current task and summarizing.`;
+- When remaining rounds are low, focus on completing the current task and summarizing.${pauseHint}`;
     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 roundToolHistory = [];
     const warnNoteAt = Math.max(10, Math.floor(maxToolRounds * 0.2));
     const warnLowAt = Math.max(5, Math.floor(maxToolRounds * 0.1));
     const warnCriticalAt = Math.max(3, Math.floor(maxToolRounds * 0.05));
@@ -829,6 +859,8 @@ 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");
@@ -843,12 +875,15 @@ 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;
         }
         if (result.toolCalls && result.toolCalls.length > 0) {
+          roundToolHistory.push({
+            round: round + 1,
+            tools: result.toolCalls.map((tc) => tc.name)
+          });
           googleSearchContext.configManager = this.config;
           spawnAgentContext.provider = provider;
           spawnAgentContext.model = this.currentModel;
@@ -876,6 +911,39 @@ You have a maximum of ${maxToolRounds} tool call rounds for this task. Plan effi
             extraMessages.push({ role: "user", content: msg });
           }
         }
+        const effectiveRound = round + 1;
+        const remaining = maxToolRounds - effectiveRound;
+        if (autoPauseInterval > 0 && effectiveRound > 0 && effectiveRound % autoPauseInterval === 0 && remaining > 0 && !ac.signal.aborted) {
+          const recentHistory = roundToolHistory.slice(-autoPauseInterval);
+          const toolCounts = /* @__PURE__ */ new Map();
+          for (const rh of recentHistory) {
+            for (const t of rh.tools) toolCounts.set(t, (toolCounts.get(t) || 0) + 1);
+          }
+          const toolSummary = [...toolCounts.entries()].sort((a, b) => b[1] - a[1]).map(([name, count]) => count > 1 ? `${name}\xD7${count}` : name).join(", ");
+          const requestId = `pause_${Date.now()}_${Math.random().toString(36).slice(2, 8)}`;
+          const pauseResp = await new Promise((resolve3) => {
+            this.pendingAutoPause.set(requestId, resolve3);
+            this.send({
+              type: "auto_pause_request",
+              requestId,
+              currentRound: effectiveRound,
+              totalRounds: maxToolRounds,
+              toolSummary
+            });
+          });
+          if (ac.signal.aborted) break;
+          if (pauseResp.action === "stop") {
+            this.send({ type: "info", message: `\u23F8  Stopped by user at ${effectiveRound}/${maxToolRounds}` });
+            extraMessages.push({
+              role: "user",
+              content: `The user has stopped the task at round ${effectiveRound}/${maxToolRounds}. Do not call any more tools. Summarize what has been completed and what remains.`
+            });
+            break;
+          } else if (pauseResp.action === "redirect" && pauseResp.message) {
+            this.send({ type: "info", message: `\u26A1 Redirect: "${pauseResp.message}"` });
+            extraMessages.push({ role: "user", content: pauseResp.message });
+          }
+        }
       }
       try {
         const summaryExtra = [
@@ -917,8 +985,7 @@ ${summaryResult.content}`,
           message: `Reached maximum tool call rounds (${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 +1101,7 @@ ${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 +1135,19 @@ ${summaryResult.content}`,
       }
       case "status": {
         const session = this.sessions.current;
+        const cacheRead = this.sessionTokenUsage.cacheReadTokens;
+        const cacheCreate = this.sessionTokenUsage.cacheCreationTokens;
+        const cost = computeCost(this.currentProvider, this.currentModel, this.sessionTokenUsage);
+        const cacheLine = cacheRead > 0 || cacheCreate > 0 ? `
+Cache:  write=${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 +1157,7 @@ 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 +1168,7 @@ 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 ?? ""}" (${cached.messages.length} messages)`
@@ -1108,7 +1182,7 @@ Tokens: in=${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 ?? ""}" (${found.messageCount} messages)` });
             this.sendSessionMessages();
             this.sendStatus();
@@ -1116,7 +1190,7 @@ Tokens: in=${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 +1330,41 @@ 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 +1790,7 @@ ${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-GISOOQZC.js");
           const argStr = args.join(" ").trim();
           let testArgs = {};
           if (argStr) {

package/dist/{task-orchestrator-TSY7CJE6.js → task-orchestrator-6RGRKSTU.js}

@@ -4,11 +4,11 @@ import {
   getDangerLevel,
   googleSearchContext,
   truncateOutput
-} from "./chunk-TAR67QTH.js";
+} from "./chunk-FJSEFQ54.js";
 import "./chunk-4BKXL7SM.js";
 import {
   SUBAGENT_ALLOWED_TOOLS
-} from "./chunk-NP5KZVP6.js";
+} from "./chunk-7FOGK5TM.js";
 
 // src/hub/task-orchestrator.ts
 import { createInterface } from "readline";

package/dist/web/client/app.js

@@ -176,6 +176,7 @@ function handleServerMessage(msg) {
     case 'confirm_request': handleConfirmRequest(msg); break;
     case 'batch_confirm_request': handleBatchConfirmRequest(msg); break;
     case 'ask_user_request':handleAskUserRequest(msg); break;
+    case 'auto_pause_request': handleAutoPauseRequest(msg); break;
     case 'thinking_start':  handleThinkingStart(); break;
     case 'thinking_delta':  handleThinkingDelta(msg.delta); break;
     case 'thinking_end':    handleThinkingEnd(); break;
@@ -377,6 +378,34 @@ function handleAskUserRequest(msg) {
   setTimeout(() => document.getElementById(`ask-input-${msg.requestId}`)?.focus(), 100);
 }
 
+function handleAutoPauseRequest(msg) {
+  const el = document.createElement('div');
+  el.className = 'confirm-card tool-border-write my-1';
+  el.id = `pause-card-${msg.requestId}`;
+  const summary = msg.toolSummary ? escapeHtml(msg.toolSummary) : '(none)';
+  el.innerHTML = `
+    <div class="flex items-center gap-2 mb-2">
+      <span class="badge badge-warning badge-sm">⏸ Auto-pause</span>
+      <span class="text-xs opacity-70">Round ${msg.currentRound}/${msg.totalRounds} · ${msg.totalRounds - msg.currentRound} remaining</span>
+    </div>
+    <div class="text-xs opacity-70 mb-2">Recent tools: ${summary}</div>
+    <div class="flex gap-2 w-full mb-2">
+      <input type="text" id="pause-input-${msg.requestId}"
+        class="input input-sm input-bordered flex-1"
+        placeholder="Optional: redirect AI with a new instruction..."
+        onkeydown="if(event.key==='Enter')respondAutoPause('${msg.requestId}','redirect')">
+    </div>
+    <div class="flex gap-2">
+      <button class="btn btn-success btn-sm flex-1" onclick="respondAutoPause('${msg.requestId}','continue')">▶ Continue</button>
+      <button class="btn btn-primary btn-sm flex-1" onclick="respondAutoPause('${msg.requestId}','redirect')">↪ Redirect</button>
+      <button class="btn btn-error btn-sm flex-1" onclick="respondAutoPause('${msg.requestId}','stop')">⏹ Stop</button>
+    </div>
+  `;
+  messagesEl.appendChild(el);
+  scrollToBottom();
+  setTimeout(() => document.getElementById(`pause-input-${msg.requestId}`)?.focus(), 100);
+}
+
 function handleRoundProgress(msg) {
   const progressBar = document.getElementById('round-progress');
   const progressBarEl = document.getElementById('round-progress-bar');
@@ -496,7 +525,16 @@ 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);
@@ -559,6 +597,30 @@ window.respondAskUser = function(requestId) {
   }
 };
 
+window.respondAutoPause = function(requestId, action) {
+  const card = document.getElementById(`pause-card-${requestId}`);
+  const input = document.getElementById(`pause-input-${requestId}`);
+  const typed = input ? input.value.trim() : '';
+
+  // 若用户在输入框敲回车，action 是 'redirect'；若内容为空则视作 continue
+  let effective = action;
+  let message;
+  if (action === 'redirect') {
+    if (!typed) { effective = 'continue'; }
+    else { message = typed; }
+  }
+
+  send({ type: 'auto_pause_response', requestId, action: effective, message });
+
+  if (card) {
+    card.className = 'confirm-card tool-border-safe my-1 opacity-60';
+    const label = effective === 'continue' ? '▶ Continued'
+      : effective === 'stop' ? '⏹ Stopped'
+      : `↪ Redirected: ${escapeHtml(message || '')}`;
+    card.innerHTML = `<span class="text-sm">${label}</span>`;
+  }
+};
+
 // DaisyUI light themes → highlight.js light stylesheet; others → dark
 const LIGHT_DAISYUI_THEMES = new Set(['light', 'cupcake', 'bumblebee', 'emerald', 'corporate', 'garden', 'lofi', 'pastel', 'fantasy', 'wireframe', 'cmyk', 'autumn', 'acid', 'lemonade', 'winter', 'nord']);
 const HLJS_CDN = 'https://cdn.jsdelivr.net/gh/highlightjs/cdn-release@11.9.0/build/styles';

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "jinzd-ai-cli",
-  "version": "0.4.54",
+  "version": "0.4.56",
   "description": "Cross-platform REPL-style AI CLI with multi-provider support",
   "type": "module",
   "main": "./dist/index.js",