@saber2pr/ai-agent 0.0.19 → 0.0.21

package/lib/cli-graph.d.ts

@@ -0,0 +1,2 @@
+#!/usr/bin/env node
+export {};

package/lib/cli-graph.js

@@ -0,0 +1,9 @@
+#!/usr/bin/env node
+"use strict";
+var __importDefault = (this && this.__importDefault) || function (mod) {
+    return (mod && mod.__esModule) ? mod : { "default": mod };
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+const agent_graph_1 = __importDefault(require("./core/agent-graph"));
+const agent = new agent_graph_1.default();
+agent.start();

package/lib/core/agent-graph.d.ts

@@ -0,0 +1,52 @@
+import { BaseMessage } from "@langchain/core/messages";
+import { AgentOptions } from "../types/type";
+export declare const CONFIG_FILE: string;
+declare const AgentState: import("@langchain/langgraph").AnnotationRoot<{
+    messages: import("@langchain/langgraph").BinaryOperatorAggregate<BaseMessage[], BaseMessage[]>;
+    auditedFiles: import("@langchain/langgraph").BinaryOperatorAggregate<string[], string[]>;
+    targetCount: import("@langchain/langgraph").BinaryOperatorAggregate<number, number>;
+    mode: import("@langchain/langgraph").BinaryOperatorAggregate<"chat" | "auto", "chat" | "auto">;
+}>;
+export default class McpGraphAgent {
+    private model;
+    private toolNode;
+    private targetDir;
+    private options;
+    private checkpointer;
+    private langchainTools;
+    private spinner;
+    constructor(options?: AgentOptions);
+    private askForConfig;
+    private getModel;
+    chat(query?: string): Promise<void>;
+    start(): Promise<void>;
+    private renderOutput;
+    callModel(state: typeof AgentState.State): Promise<{
+        messages: unknown[];
+    }>;
+    trackProgress(state: typeof AgentState.State): Promise<{
+        auditedFiles: string[];
+    }>;
+    createGraph(): Promise<import("@langchain/langgraph").CompiledStateGraph<import("@langchain/langgraph").StateType<{
+        messages: import("@langchain/langgraph").BinaryOperatorAggregate<BaseMessage[], BaseMessage[]>;
+        auditedFiles: import("@langchain/langgraph").BinaryOperatorAggregate<string[], string[]>;
+        targetCount: import("@langchain/langgraph").BinaryOperatorAggregate<number, number>;
+        mode: import("@langchain/langgraph").BinaryOperatorAggregate<"chat" | "auto", "chat" | "auto">;
+    }>, import("@langchain/langgraph").UpdateType<{
+        messages: import("@langchain/langgraph").BinaryOperatorAggregate<BaseMessage[], BaseMessage[]>;
+        auditedFiles: import("@langchain/langgraph").BinaryOperatorAggregate<string[], string[]>;
+        targetCount: import("@langchain/langgraph").BinaryOperatorAggregate<number, number>;
+        mode: import("@langchain/langgraph").BinaryOperatorAggregate<"chat" | "auto", "chat" | "auto">;
+    }>, "tools" | "agent" | "__start__" | "progress", {
+        messages: import("@langchain/langgraph").BinaryOperatorAggregate<BaseMessage[], BaseMessage[]>;
+        auditedFiles: import("@langchain/langgraph").BinaryOperatorAggregate<string[], string[]>;
+        targetCount: import("@langchain/langgraph").BinaryOperatorAggregate<number, number>;
+        mode: import("@langchain/langgraph").BinaryOperatorAggregate<"chat" | "auto", "chat" | "auto">;
+    }, {
+        messages: import("@langchain/langgraph").BinaryOperatorAggregate<BaseMessage[], BaseMessage[]>;
+        auditedFiles: import("@langchain/langgraph").BinaryOperatorAggregate<string[], string[]>;
+        targetCount: import("@langchain/langgraph").BinaryOperatorAggregate<number, number>;
+        mode: import("@langchain/langgraph").BinaryOperatorAggregate<"chat" | "auto", "chat" | "auto">;
+    }, import("@langchain/langgraph").StateDefinition>>;
+}
+export {};

package/lib/core/agent-graph.js

@@ -0,0 +1,245 @@
+"use strict";
+var __importDefault = (this && this.__importDefault) || function (mod) {
+    return (mod && mod.__esModule) ? mod : { "default": mod };
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.CONFIG_FILE = void 0;
+const openai_1 = require("@langchain/openai");
+const messages_1 = require("@langchain/core/messages");
+const langgraph_1 = require("@langchain/langgraph");
+const prebuilt_1 = require("@langchain/langgraph/prebuilt");
+const prompts_1 = require("@langchain/core/prompts");
+const readline_1 = __importDefault(require("readline"));
+const fs_1 = __importDefault(require("fs"));
+const path_1 = __importDefault(require("path"));
+const os_1 = __importDefault(require("os"));
+const ora_1 = __importDefault(require("ora")); // 用于显示 Loading 动画
+const builtin_1 = require("../tools/builtin");
+const convertToLangChainTool_1 = require("../utils/convertToLangChainTool");
+exports.CONFIG_FILE = path_1.default.join(os_1.default.homedir(), ".saber2pr-agent.json");
+// --- 1. 定义状态 (State) ---
+const AgentState = langgraph_1.Annotation.Root({
+    messages: (0, langgraph_1.Annotation)({
+        reducer: (x, y) => x.concat(y),
+        default: () => [],
+    }),
+    auditedFiles: (0, langgraph_1.Annotation)({
+        reducer: (x, y) => Array.from(new Set([...x, ...y])),
+        default: () => [],
+    }),
+    targetCount: (0, langgraph_1.Annotation)({
+        reducer: (x, y) => y !== null && y !== void 0 ? y : x,
+        default: () => 4,
+    }),
+    mode: (0, langgraph_1.Annotation)({
+        reducer: (x, y) => y !== null && y !== void 0 ? y : x,
+        default: () => "chat",
+    }),
+});
+class McpGraphAgent {
+    constructor(options = {}) {
+        this.checkpointer = new langgraph_1.MemorySaver();
+        this.langchainTools = [];
+        this.spinner = (0, ora_1.default)({ color: "cyan" });
+        this.options = options;
+        this.targetDir = options.targetDir || process.cwd();
+        process.setMaxListeners(50); // 防止 AbortSignal 监听器过多的警告
+        const builtinToolInfos = (0, builtin_1.createDefaultBuiltinTools)({ options });
+        const externalToolInfos = options.tools || [];
+        this.langchainTools = [...builtinToolInfos, ...externalToolInfos].map((t) => (0, convertToLangChainTool_1.convertToLangChainTool)(t));
+        this.toolNode = new prebuilt_1.ToolNode(this.langchainTools);
+    }
+    async askForConfig() {
+        let config = {};
+        if (fs_1.default.existsSync(exports.CONFIG_FILE)) {
+            try {
+                config = JSON.parse(fs_1.default.readFileSync(exports.CONFIG_FILE, "utf-8"));
+            }
+            catch (e) { }
+        }
+        if (!config.baseURL || !config.apiKey) {
+            const rl = readline_1.default.createInterface({ input: process.stdin, output: process.stdout });
+            const question = (q) => new Promise((res) => rl.question(q, res));
+            console.log(`💡 首次运行请配置信息：`);
+            config.baseURL = config.baseURL || await question(`? API Base URL: `);
+            config.apiKey = config.apiKey || await question(`? API Key: `);
+            config.model = config.model || await question(`? Model Name: `) || "gpt-4o";
+            fs_1.default.writeFileSync(exports.CONFIG_FILE, JSON.stringify(config, null, 2));
+            rl.close();
+        }
+        return config;
+    }
+    async getModel() {
+        if (this.model)
+            return this.model;
+        let modelInstance = this.options.apiModel;
+        if (!modelInstance) {
+            const config = await this.askForConfig();
+            modelInstance = new openai_1.ChatOpenAI({
+                openAIApiKey: config.apiKey,
+                configuration: { baseURL: config.baseURL },
+                modelName: config.model,
+                temperature: 0,
+            });
+        }
+        this.model = modelInstance.bindTools(this.langchainTools);
+        return this.model;
+    }
+    async chat(query = "开始代码审计") {
+        await this.getModel();
+        const app = await this.createGraph();
+        const stream = await app.stream({
+            messages: [new messages_1.HumanMessage(query)],
+            mode: "auto",
+            targetCount: 4
+        }, { configurable: { thread_id: "auto_worker" }, recursionLimit: 100 });
+        for await (const output of stream)
+            this.renderOutput(output);
+        console.log("✅ 任务执行完毕。");
+    }
+    async start() {
+        await this.getModel();
+        const app = await this.createGraph();
+        const rl = readline_1.default.createInterface({ input: process.stdin, output: process.stdout });
+        const threadId = `session_${Date.now()}`;
+        console.log(`\n💬 已进入交互审计模式 (Thread: ${threadId})`);
+        const ask = () => {
+            rl.question("> ", async (input) => {
+                if (input.toLowerCase() === "exit") {
+                    rl.close();
+                    return;
+                }
+                const stream = await app.stream({ messages: [new messages_1.HumanMessage(input)], mode: "chat" }, { configurable: { thread_id: threadId }, recursionLimit: 50 });
+                for await (const output of stream)
+                    this.renderOutput(output);
+                ask();
+            });
+        };
+        ask();
+    }
+    renderOutput(output) {
+        var _a, _b;
+        // 每次渲染输出前，确保停止 Spinner
+        if (this.spinner.isSpinning)
+            this.spinner.stop();
+        const agentNode = output.agent;
+        if (agentNode) {
+            const msg = agentNode.messages[0];
+            // 1. 打印思考过程
+            const reasoning = (_a = msg.additional_kwargs) === null || _a === void 0 ? void 0 : _a.reasoning;
+            if (reasoning) {
+                console.log("\n🧠 [思考过程]:\n" + "─".repeat(50) + "\n" + reasoning + "\n" + "─".repeat(50) + "\n");
+            }
+            // 2. 打印正式回答
+            if (msg.content)
+                console.log("🤖 [AI]:", msg.content);
+            // 3. 打印工具调用
+            if ((_b = msg.tool_calls) === null || _b === void 0 ? void 0 : _b.length) {
+                msg.tool_calls.forEach((call) => {
+                    console.log(`🛠️ [调用工具]: ${call.name} 📦 参数: ${JSON.stringify(call.args)}`);
+                });
+            }
+        }
+    }
+    // --- 节点逻辑 ---
+    async callModel(state) {
+        // 处理变量序列化，防止 [object Object]
+        const auditedListStr = state.auditedFiles.length > 0
+            ? state.auditedFiles.map(f => `\n  - ${f}`).join("")
+            : "暂无";
+        const extraPromptStr = typeof this.options.extraSystemPrompt === 'object'
+            ? JSON.stringify(this.options.extraSystemPrompt, null, 2)
+            : (this.options.extraSystemPrompt || "");
+        // 使用变量占位符 {extraPrompt} 避免内容中的 {} 引发模板解析错误
+        const prompt = prompts_1.ChatPromptTemplate.fromMessages([
+            ["system", `你是一个代码专家。工作目录：${this.targetDir}。
+
+# 当前进度状态
+- 审计模式: {mode}
+- 目标任务数: {targetCount}
+- 已完成数量: {doneCount}
+- 已审计文件列表: {auditedList}
+
+# 核心任务准则
+1. 目标导向：如果 {doneCount} >= {targetCount}，说明任务已达标。请直接输出总结，不要再调用任何工具。
+2. 避免死循环：不要反复尝试审计同一个文件或调用同样的工具。如果你发现某个文件修复失败，请尝试审计其他文件。
+3. 严格格式：
+    - 必须先在 <think> 标签内进行推理。
+    - 工具调用必须严格按照 Action: 名称 和 Arguments: {{JSON}} 的格式。
+    - 【重要】Arguments 中的 JSON 字符串，所有的换行符必须转义为 \\n，严禁出现物理换行符。
+
+# 附加指令
+{extraPrompt}`],
+            new prompts_1.MessagesPlaceholder("messages"),
+        ]);
+        // ✅ 显示 Loading
+        this.spinner.start("AI 正在思考并分析代码...");
+        try {
+            const chain = prompt.pipe(this.model);
+            const response = await chain.invoke({
+                messages: state.messages,
+                mode: state.mode,
+                targetCount: state.targetCount,
+                doneCount: state.auditedFiles.length,
+                auditedList: auditedListStr,
+                extraPrompt: extraPromptStr, // 变量方式传入更安全
+            });
+            this.spinner.stop(); // 得到响应即停止
+            return { messages: [response] };
+        }
+        catch (error) {
+            this.spinner.fail("AI 响应异常");
+            throw error;
+        }
+    }
+    // agent-graph.ts 中的 trackProgress 节点
+    async trackProgress(state) {
+        var _a;
+        // 获取最后一条 AI 消息（即发起工具调用的那条）
+        const lastAiMsg = state.messages[state.messages.length - 1];
+        const newFiles = [];
+        if ((_a = lastAiMsg === null || lastAiMsg === void 0 ? void 0 : lastAiMsg.tool_calls) === null || _a === void 0 ? void 0 : _a.length) {
+            for (const tc of lastAiMsg.tool_calls) {
+                // 这里的逻辑要宽容：只要 AI 尝试处理了这些文件，就计入进度
+                const file = tc.args.path || tc.args.filePath || tc.args.file;
+                if (file && typeof file === 'string') {
+                    newFiles.push(file);
+                }
+            }
+        }
+        // 如果这一轮没有任何新文件被处理，且 AI 也没给最终回复，
+        // 我们需要防止它在下一轮条件判断中陷入死循环
+        return { auditedFiles: newFiles };
+    }
+    async createGraph() {
+        const workflow = new langgraph_1.StateGraph(AgentState)
+            .addNode("agent", (state) => this.callModel(state))
+            .addNode("tools", this.toolNode)
+            .addNode("progress", (state) => this.trackProgress(state))
+            .addEdge(langgraph_1.START, "agent")
+            .addConditionalEdges("agent", (state) => {
+            var _a;
+            const lastMsg = state.messages[state.messages.length - 1];
+            // 1. 如果 AI 想要调用工具，去 tools 节点
+            if ((_a = lastMsg.tool_calls) === null || _a === void 0 ? void 0 : _a.length)
+                return "tools";
+            // 2. 如果是自动模式且未达标
+            if (state.mode === "auto") {
+                const isDone = state.auditedFiles.length >= state.targetCount;
+                // 如果还没达标，继续让 agent 思考下一步
+                if (!isDone)
+                    return "agent";
+            }
+            // 3. 其他情况（达标了，或者对话模式 AI 给出了回复）一律结束
+            return langgraph_1.END;
+        }, {
+            tools: "tools",
+            agent: "agent",
+            [langgraph_1.END]: langgraph_1.END,
+        })
+            .addEdge("tools", "progress")
+            .addEdge("progress", "agent");
+        return workflow.compile({ checkpointer: this.checkpointer });
+    }
+}
+exports.default = McpGraphAgent;

package/lib/index.d.ts

@@ -3,3 +3,5 @@ export { default as McpChainAgent } from './core/agent-chain';
 export { default } from './core/agent';
 export { createTool } from './utils/createTool';
 export { AgentChainModel } from './model/AgentChainModel';
+export { AgentGraphModel, AgentGraphLLMResponse } from './model/AgentGraphModel';
+export { default as McpGraphAgent } from './core/agent-graph';

package/lib/index.js

@@ -17,7 +17,7 @@ var __importDefault = (this && this.__importDefault) || function (mod) {
     return (mod && mod.__esModule) ? mod : { "default": mod };
 };
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.AgentChainModel = exports.createTool = exports.default = exports.McpChainAgent = void 0;
+exports.McpGraphAgent = exports.AgentGraphModel = exports.AgentChainModel = exports.createTool = exports.default = exports.McpChainAgent = void 0;
 __exportStar(require("./core/agent"), exports);
 var agent_chain_1 = require("./core/agent-chain");
 Object.defineProperty(exports, "McpChainAgent", { enumerable: true, get: function () { return __importDefault(agent_chain_1).default; } });
@@ -27,3 +27,7 @@ var createTool_1 = require("./utils/createTool");
 Object.defineProperty(exports, "createTool", { enumerable: true, get: function () { return createTool_1.createTool; } });
 var AgentChainModel_1 = require("./model/AgentChainModel");
 Object.defineProperty(exports, "AgentChainModel", { enumerable: true, get: function () { return AgentChainModel_1.AgentChainModel; } });
+var AgentGraphModel_1 = require("./model/AgentGraphModel");
+Object.defineProperty(exports, "AgentGraphModel", { enumerable: true, get: function () { return AgentGraphModel_1.AgentGraphModel; } });
+var agent_graph_1 = require("./core/agent-graph");
+Object.defineProperty(exports, "McpGraphAgent", { enumerable: true, get: function () { return __importDefault(agent_graph_1).default; } });

package/lib/model/AgentChainModel.d.ts

@@ -1,12 +1,12 @@
 import { BaseChatModel } from '@langchain/core/language_models/chat_models';
 import { AIMessage, MessageFieldWithRole } from '@langchain/core/messages';
 interface AgentChainModelImpl {
-    generateAgentChainResponse: (messages: MessageFieldWithRole[]) => Promise<string>;
+    generateResponse: (messages: MessageFieldWithRole[]) => Promise<string>;
 }
 export declare abstract class AgentChainModel extends BaseChatModel implements AgentChainModelImpl {
     bind(args: any): any;
     constructor(fields?: any);
-    generateAgentChainResponse(messages: MessageFieldWithRole[]): Promise<string>;
+    generateResponse(messages: MessageFieldWithRole[]): Promise<string>;
     _generate(messages: any): Promise<{
         generations: {
             text: string;

package/lib/model/AgentChainModel.js

@@ -11,11 +11,11 @@ class AgentChainModel extends chat_models_1.BaseChatModel {
         return (super.bind ? super.bind(args) : this);
     }
     constructor(fields) { super(fields || {}); }
-    async generateAgentChainResponse(messages) {
+    async generateResponse(messages) {
         return '';
     }
     async _generate(messages) {
-        let text = await this.generateAgentChainResponse(messages);
+        let text = await this.generateResponse(messages);
         return { generations: [{ text, message: new messages_1.AIMessage(text) }] };
     }
     _llmType() { return "my_private_llm"; }

package/lib/model/AgentGraphModel.d.ts

@@ -0,0 +1,21 @@
+import { BaseChatModel, BaseChatModelParams } from '@langchain/core/language_models/chat_models';
+import { BaseMessage } from '@langchain/core/messages';
+import { ChatResult } from '@langchain/core/outputs';
+export interface AgentGraphLLMResponse {
+    text: string;
+    reasoning?: string;
+    chatId?: string;
+}
+export declare abstract class AgentGraphModel extends BaseChatModel {
+    protected boundTools?: any[];
+    protected chatId?: string;
+    constructor(fields?: BaseChatModelParams & {
+        chatId?: string;
+    });
+    bindTools(tools: any[]): any;
+    abstract callApi(prompt: string, chatId?: string): Promise<AgentGraphLLMResponse>;
+    _generate(messages: BaseMessage[]): Promise<ChatResult>;
+    private serializeMessages;
+    private parseToolCalls;
+    _llmType(): string;
+}

package/lib/model/AgentGraphModel.js

@@ -0,0 +1,96 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.AgentGraphModel = void 0;
+const chat_models_1 = require("@langchain/core/language_models/chat_models");
+const messages_1 = require("@langchain/core/messages");
+const function_calling_1 = require("@langchain/core/utils/function_calling");
+class AgentGraphModel extends chat_models_1.BaseChatModel {
+    constructor(fields) {
+        super(fields || {});
+        this.chatId = fields === null || fields === void 0 ? void 0 : fields.chatId;
+    }
+    bindTools(tools) {
+        this.boundTools = tools.map(t => (0, function_calling_1.convertToOpenAITool)(t));
+        return this;
+    }
+    async _generate(messages) {
+        const fullPrompt = this.serializeMessages(messages);
+        const response = await this.callApi(fullPrompt, this.chatId);
+        if (response.chatId)
+            this.chatId = response.chatId;
+        let { text, reasoning } = response;
+        // ✅ 通用逻辑：解析 <think> 标签
+        if (!reasoning && text.includes("<think>")) {
+            const match = text.match(/<think>([\s\S]*?)<\/think>/);
+            if (match) {
+                reasoning = match[1].trim();
+                text = text.replace(/<think>[\s\S]*?<\/think>/, "").trim();
+            }
+        }
+        const toolCalls = this.parseToolCalls(text);
+        return {
+            generations: [{
+                    text,
+                    message: new messages_1.AIMessage({
+                        content: text,
+                        tool_calls: toolCalls,
+                        additional_kwargs: { reasoning: reasoning || "" }
+                    })
+                }]
+        };
+    }
+    serializeMessages(messages) {
+        const systemMsg = messages.find(m => m._getType() === 'system');
+        const lastMsg = messages[messages.length - 1];
+        const format = (m) => {
+            const content = typeof m.content === 'string' ? m.content : JSON.stringify(m.content, null, 2);
+            return `${m._getType().toUpperCase()}: ${content}`;
+        };
+        const toolsContext = this.boundTools ? `\n[Tools]\n${JSON.stringify(this.boundTools, null, 2)}` : "";
+        return `
+${format(systemMsg)}
+${toolsContext}
+# Current Progress
+${format(lastMsg)}
+# Output Requirement
+1. Reasoning in <think> tags.
+2. Action: Name
+3. Arguments: {JSON}
+`.trim();
+    }
+    parseToolCalls(text) {
+        const actionMatch = text.match(/Action:\s*(\w+)/);
+        const argsMatch = text.match(/Arguments:\s*({[\s\S]*})/);
+        if (!actionMatch)
+            return [];
+        let args = {};
+        if (argsMatch) {
+            try {
+                // 强力解析逻辑，处理物理换行
+                const rawArgs = argsMatch[1].trim().replace(/\n/g, "\\n");
+                args = JSON.parse(rawArgs);
+                // 参数映射
+                const anyArgs = args;
+                if (anyArgs.file_path && !anyArgs.path)
+                    anyArgs.path = anyArgs.file_path;
+                if (anyArgs.filePath && !anyArgs.path)
+                    anyArgs.path = anyArgs.filePath;
+                if (anyArgs.path && !anyArgs.filePath)
+                    anyArgs.filePath = anyArgs.path;
+                if (anyArgs.file && !anyArgs.filePath)
+                    anyArgs.filePath = anyArgs.file;
+            }
+            catch (e) {
+                console.warn("JSON Parse Error", e);
+            }
+        }
+        return [{
+                name: actionMatch[1],
+                args,
+                id: `call_${Date.now()}`,
+                type: "tool_call",
+            }];
+    }
+    _llmType() { return "agent_graph_model"; }
+}
+exports.AgentGraphModel = AgentGraphModel;

package/lib/types/type.d.ts

@@ -46,4 +46,16 @@ export interface AgentOptions {
      * only for chain agent
      */
     verbose?: boolean;
+    /**
+     * only for graph agent
+     */
+    baseURL?: string;
+    /**
+     * only for graph agent
+     */
+    apiKey?: string;
+    /**
+     * only for graph agent
+     */
+    modelName?: string;
 }

package/lib/utils/convertToLangChainTool.d.ts

@@ -0,0 +1,3 @@
+import { DynamicStructuredTool } from "@langchain/core/tools";
+import { z } from "zod";
+export declare function convertToLangChainTool(info: any): DynamicStructuredTool<z.ZodRecord<z.ZodString, z.ZodAny>>;

package/lib/utils/convertToLangChainTool.js

@@ -0,0 +1,24 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.convertToLangChainTool = convertToLangChainTool;
+const tools_1 = require("@langchain/core/tools");
+const zod_1 = require("zod");
+function convertToLangChainTool(info) {
+    return new tools_1.DynamicStructuredTool({
+        name: info.function.name,
+        description: info.function.description || "",
+        schema: zod_1.z.record(zod_1.z.any()),
+        func: async (args) => {
+            if (info._handler)
+                return await info._handler(args);
+            if (info._client && info._originalName) {
+                const result = await info._client.callTool({
+                    name: info._originalName,
+                    arguments: args,
+                });
+                return JSON.stringify(result);
+            }
+            return "Error: No tool execution handler found.";
+        },
+    });
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@saber2pr/ai-agent",
-  "version": "0.0.19",
+  "version": "0.0.21",
   "description": "AI Assistant CLI.",
   "author": "saber2pr",
   "license": "ISC",
@@ -9,7 +9,8 @@
   ],
   "bin": {
     "sagent": "./lib/cli.js",
-    "sagent-chain": "./lib/cli-chain.js"
+    "sagent-chain": "./lib/cli-chain.js",
+    "sagent-graph": "./lib/cli-graph.js"
   },
   "publishConfig": {
     "access": "public",
@@ -22,18 +23,20 @@
     "prepublishOnly": "tsc"
   },
   "dependencies": {
+    "@langchain/core": "0.3.39",
+    "@langchain/langgraph": "^0.2.39",
+    "@langchain/openai": "0.4.0",
     "@modelcontextprotocol/sdk": "^1.25.3",
     "@saber2pr/ts-context-mcp": "^0.0.8",
     "diff": "^8.0.3",
     "glob": "^10.5.0",
     "js-tiktoken": "^1.0.21",
+    "langchain": "0.3.15",
     "minimatch": "^10.0.1",
     "openai": "^6.16.0",
-    "zod-to-json-schema": "3.23.2",
-    "langchain": "0.3.15",
-    "@langchain/core": "0.3.39",
-    "@langchain/openai": "0.4.0",
-    "zod": "3.23.8"
+    "ora": "^9.3.0",
+    "zod": "3.23.8",
+    "zod-to-json-schema": "3.23.2"
   },
   "resolutions": {
     "@langchain/core": "0.3.39"