jinzd-ai-cli 0.4.53 → 0.4.55

package/dist/{chunk-W6AK76UM.js → chunk-DJ342VFS.js}

@@ -6,7 +6,7 @@ import { platform } from "os";
 import chalk from "chalk";
 
 // src/core/constants.ts
-var VERSION = "0.4.53";
+var VERSION = "0.4.55";
 var APP_NAME = "ai-cli";
 var CONFIG_DIR_NAME = ".aicli";
 var CONFIG_FILE_NAME = "config.json";

package/dist/{chunk-6I5FUNPR.js → chunk-JL5NK6AR.js}

@@ -7,7 +7,7 @@ import {
   ProviderNotFoundError,
   RateLimitError,
   schemaToJsonSchema
-} from "./chunk-IXDGWT2Z.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-YIMTDKUW.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-YIMTDKUW.js → chunk-W7QVBFIJ.js}

@@ -8,7 +8,7 @@ import { platform } from "os";
 import chalk from "chalk";
 
 // src/core/constants.ts
-var VERSION = "0.4.53";
+var VERSION = "0.4.55";
 var APP_NAME = "ai-cli";
 var CONFIG_DIR_NAME = ".aicli";
 var CONFIG_FILE_NAME = "config.json";

package/dist/{chunk-IXDGWT2Z.js → chunk-YQEIQJ6K.js}

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

package/dist/{hub-4DNFD6JK.js → hub-AUWP4SWJ.js}

@@ -385,7 +385,7 @@ ${content}`);
   }
 }
 async function runTaskMode(config, providers, configManager, topic) {
-  const { TaskOrchestrator } = await import("./task-orchestrator-C42TNHE6.js");
+  const { TaskOrchestrator } = await import("./task-orchestrator-MWO6A4KQ.js");
   const orchestrator = new TaskOrchestrator(config, providers, configManager);
   let interrupted = false;
   const onSigint = () => {