@lutery/vision-mcp 1.0.0

package/dist/src/prompts/system.js

@@ -0,0 +1,272 @@
+/**
+ * System Prompt Templates
+ *
+ * @description 内置的系统提示词模板，覆盖常见视觉任务场景
+ */
+/**
+ * 基础系统提示词
+ * 提供通用指导原则和规范
+ */
+export const BASE_SYSTEM_PROMPT = `You are a vision analysis assistant with access to image data.
+
+Your task is to analyze images accurately and provide helpful responses based on what you see.
+
+Guidelines:
+1. Be precise and detailed in your observations
+2. When coordinates or measurements are requested, provide them in the format specified
+3. If an image is unclear or ambiguous, acknowledge the limitations
+4. Convert any complex structured data to the requested format
+5. Be helpful and follow the specific instructions provided by the user
+
+Remember: You are providing visual information to another AI system, so clarity and structure are important.`;
+/**
+ * 通用描述模板
+ */
+export const GENERAL_DESCRIPTION_TEMPLATE = {
+    id: 'general-description',
+    name: 'General Description',
+    description: 'Provide a detailed description of the image',
+    useCases: ['image description', 'scene analysis', 'object identification'],
+    template: `${BASE_SYSTEM_PROMPT}
+
+Analyze this image and provide a comprehensive description. Include:
+- Main objects and their characteristics
+- Setting or environment
+- Colors, lighting, and composition
+- Any text or symbols visible
+- Overall mood or atmosphere
+
+Focus on observable details and be specific.`
+};
+/**
+ * UI 分析模板
+ * 用于分析 UI 原型图、截图等
+ */
+export const UI_ANALYSIS_TEMPLATE = {
+    id: 'ui-analysis',
+    name: 'UI Analysis',
+    description: 'Analyze UI prototypes and screenshots',
+    useCases: ['UI design analysis', 'interface review', 'component identification'],
+    template: `${BASE_SYSTEM_PROMPT}
+
+You are analyzing a UI design or interface screenshot. Your goal is to extract structured information about the UI elements and layout.
+
+Please provide:
+1. **Overall Layout**
+   - Screen/window dimensions (if apparent)
+   - Layout type (grid, flex, absolute positioning, etc.)
+   - Main sections and their relative positions
+
+2. **UI Components**
+   - Identify all interactive elements (buttons, inputs, links, etc.)
+   - For each component, provide:
+     * Type (button, input, card, navigation, etc.)
+     * Text content or labels
+     * Approximate position and size (relative coordinates or percentages)
+     * Visual style (colors, borders, shadows, etc.)
+
+3. **Visual Hierarchy**
+   - Primary actions and elements
+   - Secondary or supporting elements
+   - Navigation structure
+
+4. **Design Patterns**
+   - Common patterns used (cards, lists, modals, etc.)
+   - Responsive design considerations
+   - Accessibility features (if visible)
+
+5. **Content Analysis**
+   - Text content and typography
+   - Images and media elements
+   - Icons and symbols
+
+If you need to provide coordinates or positions, use:
+- Either percentage-based coordinates (x, y, width, height as percentages)
+- Or pixel coordinates if the image dimensions are clear
+Format: {x: 10, y: 20, width: 30, height: 15} (in respective units)
+
+Be precise about relationships between elements and their spatial arrangement.`
+};
+/**
+ * 对象定位和检测模板
+ */
+export const OBJECT_DETECTION_TEMPLATE = {
+    id: 'object-detection',
+    name: 'Object Detection and Localization',
+    description: 'Detect and locate objects in the image with coordinates',
+    useCases: ['object detection', 'element positioning', 'bounding box localization'],
+    template: `${BASE_SYSTEM_PROMPT}
+
+Your task is to identify and locate specific objects or elements in the image. For each detected object, provide:
+
+1. **Object Information**
+   - Object class/type
+   - Description and characteristics
+   - Confidence level (if applicable)
+
+2. **Position and Size**
+   - Bounding box coordinates in this format: {x: 10, y: 20, width: 30, height: 15}
+   - Coordinate system: origin (0,0) at top-left, units are percentage of image dimensions
+   - Alternative: If percentages aren't appropriate, describe relative position (e.g., "center", "top-right")
+
+3. **Additional Details**
+   - Color, size, or other distinguishing features
+   - Relationship to other objects
+   - Any text or labels associated with the object
+
+If multiple objects of the same type exist, list them all with their unique positions.
+
+Please format the response as structured data that can be easily parsed.`
+};
+/**
+ * OCR 和文本提取模板
+ */
+export const OCR_TEMPLATE = {
+    id: 'ocr',
+    name: 'OCR and Text Extraction',
+    description: 'Extract text from images with positioning',
+    useCases: ['text extraction', 'OCR', 'document scanning'],
+    template: `${BASE_SYSTEM_PROMPT}
+
+Extract all visible text from the image with the following details:
+
+1. **Text Content**
+   - All readable text in the image
+   - Maintain the original structure and grouping
+   - Preserve formatting if relevant (headings, paragraphs, lists)
+
+2. **Text Positioning**
+   - For each text element, provide approximate coordinates
+   - Format: {text: "content", x: 10, y: 20, width: 30, height: 15}
+   - Coordinates are optional but helpful for reference
+
+3. **Additional Information**
+   - Font characteristics (if apparent: size, weight, style)
+   - Text color and background
+   - Language detection if multiple languages present
+
+4. **Structured Output**
+   - Organize extracted text logically
+   - Group related text elements
+   - Indicate hierarchy if present (titles, body text, captions)
+
+Present the extracted information in a clear, structured format that preserves the original layout information.`
+};
+/**
+ * 结构化信息提取模板
+ */
+export const STRUCTURED_EXTRACTION_TEMPLATE = {
+    id: 'structured-extraction',
+    name: 'Structured Information Extraction',
+    description: 'Extract structured data according to a specific schema',
+    useCases: ['data extraction', 'form processing', 'structured analysis'],
+    template: `${BASE_SYSTEM_PROMPT}
+
+Extract structured information from the image following the user's specified schema or format. Focus on:
+
+1. **Identifying Key Information**
+   - Extract all relevant data points
+   - Capture relationships between data elements
+   - Include contextual information
+
+2. **Data Validation**
+   - Verify data types and formats
+   - Note any missing or ambiguous information
+   - Flag potential errors or inconsistencies
+
+3. **Output Structure**
+   - Follow the requested JSON schema or format exactly
+   - Include all required fields
+   - Use appropriate data types (string, number, array, object)
+
+4. **Quality Assurance**
+   - Double-check extracted values for accuracy
+   - Provide confidence levels if uncertain
+   - Note any assumptions made
+
+If specific fields or format are requested in the user prompt, prioritize those requirements.`
+};
+/**
+ * 获取系统提示词模板
+ *
+ * @param templateId 模板 ID
+ * @param customInstructions 额外的自定义指令（可选）
+ * @returns 完整的系统提示词
+ */
+export function getSystemPrompt(templateId, customInstructions) {
+    const template = SYSTEM_PROMPT_TEMPLATES[templateId];
+    if (!template) {
+        throw new Error(`Unknown template ID: ${templateId}`);
+    }
+    let prompt = template.template;
+    // 如果有自定义指令，追加到提示词中
+    if (customInstructions) {
+        prompt += `\n\n=== Additional Instructions ===\n${customInstructions}`;
+    }
+    return prompt;
+}
+/**
+ * 所有可用的系统提示词模板
+ */
+export const SYSTEM_PROMPT_TEMPLATES = {
+    'general-description': GENERAL_DESCRIPTION_TEMPLATE,
+    'ui-analysis': UI_ANALYSIS_TEMPLATE,
+    'object-detection': OBJECT_DETECTION_TEMPLATE,
+    'ocr': OCR_TEMPLATE,
+    'structured-extraction': STRUCTURED_EXTRACTION_TEMPLATE
+};
+/**
+ * 获取模板列表
+ */
+export function getAvailableTemplates() {
+    return Object.values(SYSTEM_PROMPT_TEMPLATES).map(t => ({
+        id: t.id,
+        name: t.name,
+        description: t.description,
+        useCases: t.useCases
+    }));
+}
+/**
+ * 自动选择合适的模板
+ *
+ * @param prompt 用户提示词
+ * @returns 推荐的模板 ID
+ */
+export function autoSelectTemplate(prompt) {
+    const lowerPrompt = prompt.toLowerCase();
+    // 关键词匹配
+    if (lowerPrompt.includes('ui') || lowerPrompt.includes('interface') || lowerPrompt.includes('prototype')) {
+        return 'ui-analysis';
+    }
+    if (lowerPrompt.includes('detect') || lowerPrompt.includes('locate') || lowerPrompt.includes('position')) {
+        return 'object-detection';
+    }
+    if (lowerPrompt.includes('ocr') || lowerPrompt.includes('text') || lowerPrompt.includes('extract text')) {
+        return 'ocr';
+    }
+    if (lowerPrompt.includes('json') || lowerPrompt.includes('structured') || lowerPrompt.includes('schema')) {
+        return 'structured-extraction';
+    }
+    if (lowerPrompt.includes('describe') || lowerPrompt.includes('what') || lowerPrompt.includes('explain')) {
+        return 'general-description';
+    }
+    // 默认使用通用描述模板
+    return 'general-description';
+}
+/**
+ * 构建完整的提示词
+ *
+ * @param templateId 模板 ID
+ * @param userPrompt 用户提示词
+ * @returns 构建完成的提示词
+ */
+export function buildPrompt(templateId, userPrompt) {
+    // 如果没有指定模板，自动选择
+    const selectedTemplateId = templateId || autoSelectTemplate(userPrompt);
+    // 构建完整提示词
+    const systemPrompt = getSystemPrompt(selectedTemplateId);
+    // 组合系统提示词和用户提示词
+    // Note: 实际使用时可能在消息数组中分别传递
+    return `${systemPrompt}\n\n=== User Request ===\n${userPrompt}`;
+}
+//# sourceMappingURL=system.js.map

package/dist/src/prompts/system.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"system.js","sourceRoot":"","sources":["../../../src/prompts/system.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAUH;;;GAGG;AACH,MAAM,CAAC,MAAM,kBAAkB,GAAG;;;;;;;;;;;6GAW2E,CAAC;AAE9G;;GAEG;AACH,MAAM,CAAC,MAAM,4BAA4B,GAAyB;IAChE,EAAE,EAAE,qBAAqB;IACzB,IAAI,EAAE,qBAAqB;IAC3B,WAAW,EAAE,6CAA6C;IAC1D,QAAQ,EAAE,CAAC,mBAAmB,EAAE,gBAAgB,EAAE,uBAAuB,CAAC;IAC1E,QAAQ,EAAE,GAAG,kBAAkB;;;;;;;;;6CASY;CAC5C,CAAC;AAEF;;;GAGG;AACH,MAAM,CAAC,MAAM,oBAAoB,GAAyB;IACxD,EAAE,EAAE,aAAa;IACjB,IAAI,EAAE,aAAa;IACnB,WAAW,EAAE,uCAAuC;IACpD,QAAQ,EAAE,CAAC,oBAAoB,EAAE,kBAAkB,EAAE,0BAA0B,CAAC;IAChF,QAAQ,EAAE,GAAG,kBAAkB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;+EAsC8C;CAC9E,CAAC;AAEF;;GAEG;AACH,MAAM,CAAC,MAAM,yBAAyB,GAAyB;IAC7D,EAAE,EAAE,kBAAkB;IACtB,IAAI,EAAE,mCAAmC;IACzC,WAAW,EAAE,yDAAyD;IACtE,QAAQ,EAAE,CAAC,kBAAkB,EAAE,qBAAqB,EAAE,2BAA2B,CAAC;IAClF,QAAQ,EAAE,GAAG,kBAAkB;;;;;;;;;;;;;;;;;;;;;yEAqBwC;CACxE,CAAC;AAEF;;GAEG;AACH,MAAM,CAAC,MAAM,YAAY,GAAyB;IAChD,EAAE,EAAE,KAAK;IACT,IAAI,EAAE,yBAAyB;IAC/B,WAAW,EAAE,2CAA2C;IACxD,QAAQ,EAAE,CAAC,iBAAiB,EAAE,KAAK,EAAE,mBAAmB,CAAC;IACzD,QAAQ,EAAE,GAAG,kBAAkB;;;;;;;;;;;;;;;;;;;;;;;;gHAwB+E;CAC/G,CAAC;AAEF;;GAEG;AACH,MAAM,CAAC,MAAM,8BAA8B,GAAyB;IAClE,EAAE,EAAE,uBAAuB;IAC3B,IAAI,EAAE,mCAAmC;IACzC,WAAW,EAAE,wDAAwD;IACrE,QAAQ,EAAE,CAAC,iBAAiB,EAAE,iBAAiB,EAAE,qBAAqB,CAAC;IACvE,QAAQ,EAAE,GAAG,kBAAkB;;;;;;;;;;;;;;;;;;;;;;;;8FAwB6D;CAC7F,CAAC;AAEF;;;;;;GAMG;AACH,MAAM,UAAU,eAAe,CAAC,UAAkB,EAAE,kBAA2B;IAC7E,MAAM,QAAQ,GAAG,uBAAuB,CAAC,UAAU,CAAC,CAAC;IAErD,IAAI,CAAC,QAAQ,EAAE,CAAC;QACd,MAAM,IAAI,KAAK,CAAC,wBAAwB,UAAU,EAAE,CAAC,CAAC;IACxD,CAAC;IAED,IAAI,MAAM,GAAG,QAAQ,CAAC,QAAQ,CAAC;IAE/B,mBAAmB;IACnB,IAAI,kBAAkB,EAAE,CAAC;QACvB,MAAM,IAAI,wCAAwC,kBAAkB,EAAE,CAAC;IACzE,CAAC;IAED,OAAO,MAAM,CAAC;AAChB,CAAC;AAED;;GAEG;AACH,MAAM,CAAC,MAAM,uBAAuB,GAAyC;IAC3E,qBAAqB,EAAE,4BAA4B;IACnD,aAAa,EAAE,oBAAoB;IACnC,kBAAkB,EAAE,yBAAyB;IAC7C,KAAK,EAAE,YAAY;IACnB,uBAAuB,EAAE,8BAA8B;CACxD,CAAC;AAEF;;GAEG;AACH,MAAM,UAAU,qBAAqB;IAMnC,OAAO,MAAM,CAAC,MAAM,CAAC,uBAAuB,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;QACtD,EAAE,EAAE,CAAC,CAAC,EAAE;QACR,IAAI,EAAE,CAAC,CAAC,IAAI;QACZ,WAAW,EAAE,CAAC,CAAC,WAAW;QAC1B,QAAQ,EAAE,CAAC,CAAC,QAAQ;KACrB,CAAC,CAAC,CAAC;AACN,CAAC;AAED;;;;;GAKG;AACH,MAAM,UAAU,kBAAkB,CAAC,MAAc;IAC/C,MAAM,WAAW,GAAG,MAAM,CAAC,WAAW,EAAE,CAAC;IAEzC,QAAQ;IACR,IAAI,WAAW,CAAC,QAAQ,CAAC,IAAI,CAAC,IAAI,WAAW,CAAC,QAAQ,CAAC,WAAW,CAAC,IAAI,WAAW,CAAC,QAAQ,CAAC,WAAW,CAAC,EAAE,CAAC;QACzG,OAAO,aAAa,CAAC;IACvB,CAAC;IAED,IAAI,WAAW,CAAC,QAAQ,CAAC,QAAQ,CAAC,IAAI,WAAW,CAAC,QAAQ,CAAC,QAAQ,CAAC,IAAI,WAAW,CAAC,QAAQ,CAAC,UAAU,CAAC,EAAE,CAAC;QACzG,OAAO,kBAAkB,CAAC;IAC5B,CAAC;IAED,IAAI,WAAW,CAAC,QAAQ,CAAC,KAAK,CAAC,IAAI,WAAW,CAAC,QAAQ,CAAC,MAAM,CAAC,IAAI,WAAW,CAAC,QAAQ,CAAC,cAAc,CAAC,EAAE,CAAC;QACxG,OAAO,KAAK,CAAC;IACf,CAAC;IAED,IAAI,WAAW,CAAC,QAAQ,CAAC,MAAM,CAAC,IAAI,WAAW,CAAC,QAAQ,CAAC,YAAY,CAAC,IAAI,WAAW,CAAC,QAAQ,CAAC,QAAQ,CAAC,EAAE,CAAC;QACzG,OAAO,uBAAuB,CAAC;IACjC,CAAC;IAED,IAAI,WAAW,CAAC,QAAQ,CAAC,UAAU,CAAC,IAAI,WAAW,CAAC,QAAQ,CAAC,MAAM,CAAC,IAAI,WAAW,CAAC,QAAQ,CAAC,SAAS,CAAC,EAAE,CAAC;QACxG,OAAO,qBAAqB,CAAC;IAC/B,CAAC;IAED,aAAa;IACb,OAAO,qBAAqB,CAAC;AAC/B,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,WAAW,CAAC,UAA8B,EAAE,UAAkB;IAC5E,gBAAgB;IAChB,MAAM,kBAAkB,GAAG,UAAU,IAAI,kBAAkB,CAAC,UAAU,CAAC,CAAC;IAExE,UAAU;IACV,MAAM,YAAY,GAAG,eAAe,CAAC,kBAAkB,CAAC,CAAC;IAEzD,gBAAgB;IAChB,0BAA0B;IAC1B,OAAO,GAAG,YAAY,6BAA6B,UAAU,EAAE,CAAC;AAClE,CAAC"}

package/dist/src/tools/vision-tool.d.ts

@@ -0,0 +1,91 @@
+/**
+ * Vision Tool
+ *
+ * @description MCP 视觉分析工具，提供统一的图片分析接口
+ */
+import { z } from 'zod';
+import { ModelConfig } from '../config/model-config.js';
+/**
+ * 工具参数 Schema
+ */
+export declare const VisionToolInputSchema: z.ZodObject<{
+    /**
+     * 图片输入：支持 URL、base64(data URL) 或本地路径
+     * @example "https://example.com/image.jpg"
+     * @example "data:image/jpeg;base64,/9j/4AAQ..."
+     * @example "/path/to/local/image.png"
+     */
+    image: z.ZodString;
+    /**
+     * 分析提示词：描述要执行的分析任务
+     * @example "Describe this UI design and extract all components"
+     * @example "Detect objects and provide their coordinates"
+     * @example "Extract text using OCR"
+     */
+    prompt: z.ZodString;
+    /**
+     * 输出格式：可选，指定为 "text" 或 "json"
+     * @default "text"
+     */
+    output_format: z.ZodDefault<z.ZodOptional<z.ZodEnum<["text", "json"]>>>;
+    /**
+     * 系统提示词模板：可选，指定使用的模板
+     * @default auto-detected based on prompt
+     * @example "ui-analysis"
+     * @example "object-detection"
+     */
+    template: z.ZodOptional<z.ZodString>;
+}, "strict", z.ZodTypeAny, {
+    image: string;
+    prompt: string;
+    output_format: "text" | "json";
+    template?: string | undefined;
+}, {
+    image: string;
+    prompt: string;
+    output_format?: "text" | "json" | undefined;
+    template?: string | undefined;
+}>;
+export type VisionToolInput = z.infer<typeof VisionToolInputSchema>;
+/**
+ * 工具响应格式
+ */
+export interface VisionToolResponse {
+    content: string;
+    format: 'text' | 'json';
+    metadata: {
+        modelType: string;
+        modelName: string;
+        imageFormat: string;
+        processingTimeMs: number;
+        imageSize: number;
+    };
+}
+/**
+ * 视觉工具类
+ */
+export declare class VisionTool {
+    private modelConfig;
+    private adapter;
+    constructor(config?: ModelConfig);
+    /**
+     * 创建模型适配器
+     */
+    private createAdapter;
+    /**
+     * 执行视觉分析
+     */
+    analyze(input: VisionToolInput): Promise<VisionToolResponse>;
+    /**
+     * 构建元数据
+     */
+    private buildMetadata;
+    /**
+     * 获取工具信息
+     */
+    getToolInfo(): {
+        modelConfig: ModelConfig;
+    };
+}
+export declare function getVisionTool(config?: ModelConfig): VisionTool;
+//# sourceMappingURL=vision-tool.d.ts.map

package/dist/src/tools/vision-tool.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"vision-tool.d.ts","sourceRoot":"","sources":["../../../src/tools/vision-tool.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AACxB,OAAO,EAAE,WAAW,EAAwC,MAAM,2BAA2B,CAAC;AAS9F;;GAEG;AACH,eAAO,MAAM,qBAAqB;IAChC;;;;;OAKG;;IAGH;;;;;OAKG;;IAGH;;;OAGG;;IAGH;;;;;OAKG;;;;;;;;;;;;EAEM,CAAC;AAEZ,MAAM,MAAM,eAAe,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,qBAAqB,CAAC,CAAC;AAEpE;;GAEG;AACH,MAAM,WAAW,kBAAkB;IACjC,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,GAAG,MAAM,CAAC;IACxB,QAAQ,EAAE;QACR,SAAS,EAAE,MAAM,CAAC;QAClB,SAAS,EAAE,MAAM,CAAC;QAClB,WAAW,EAAE,MAAM,CAAC;QACpB,gBAAgB,EAAE,MAAM,CAAC;QACzB,SAAS,EAAE,MAAM,CAAC;KACnB,CAAC;CACH;AAED;;GAEG;AACH,qBAAa,UAAU;IACrB,OAAO,CAAC,WAAW,CAAc;IACjC,OAAO,CAAC,OAAO,CAAqB;gBAExB,MAAM,CAAC,EAAE,WAAW;IAgBhC;;OAEG;IACH,OAAO,CAAC,aAAa;IAarB;;OAEG;IACU,OAAO,CAAC,KAAK,EAAE,eAAe,GAAG,OAAO,CAAC,kBAAkB,CAAC;IA2EzE;;OAEG;IACH,OAAO,CAAC,aAAa;IAUrB;;OAEG;IACI,WAAW,IAAI;QACpB,WAAW,EAAE,WAAW,CAAC;KAC1B;CAKF;AAOD,wBAAgB,aAAa,CAAC,MAAM,CAAC,EAAE,WAAW,GAAG,UAAU,CAK9D"}

package/dist/src/tools/vision-tool.js

@@ -0,0 +1,171 @@
+/**
+ * Vision Tool
+ *
+ * @description MCP 视觉分析工具，提供统一的图片分析接口
+ */
+import { z } from 'zod';
+import { loadModelConfig, validateModelConfig } from '../config/model-config.js';
+import { GLM4VisionAdapter } from '../adapters/glm-adapter.js';
+import { SiliconFlowAdapter } from '../adapters/siliconflow-adapter.js';
+import { normalizeImageInput } from '../utils/image-input.js';
+import { ModelConfigError } from '../utils/errors.js';
+import { logger } from '../utils/logger.js';
+import { buildPrompt } from '../prompts/system.js';
+/**
+ * 工具参数 Schema
+ */
+export const VisionToolInputSchema = z.object({
+    /**
+     * 图片输入：支持 URL、base64(data URL) 或本地路径
+     * @example "https://example.com/image.jpg"
+     * @example "data:image/jpeg;base64,/9j/4AAQ..."
+     * @example "/path/to/local/image.png"
+     */
+    image: z.string().describe('Image URL, base64 data URL, or local file path'),
+    /**
+     * 分析提示词：描述要执行的分析任务
+     * @example "Describe this UI design and extract all components"
+     * @example "Detect objects and provide their coordinates"
+     * @example "Extract text using OCR"
+     */
+    prompt: z.string().describe('Analysis prompt describing the task'),
+    /**
+     * 输出格式：可选，指定为 "text" 或 "json"
+     * @default "text"
+     */
+    output_format: z.enum(['text', 'json']).optional().default('text').describe('Output format preference'),
+    /**
+     * 系统提示词模板：可选，指定使用的模板
+     * @default auto-detected based on prompt
+     * @example "ui-analysis"
+     * @example "object-detection"
+     */
+    template: z.string().optional().describe('System prompt template to use')
+}).strict();
+/**
+ * 视觉工具类
+ */
+export class VisionTool {
+    modelConfig;
+    adapter;
+    constructor(config) {
+        // 加载配置
+        this.modelConfig = config || loadModelConfig();
+        // 验证配置
+        validateModelConfig(this.modelConfig);
+        // 创建适配器
+        this.adapter = this.createAdapter(this.modelConfig);
+        logger.info('Vision tool initialized', {
+            modelType: this.modelConfig.type,
+            modelName: this.modelConfig.name
+        });
+    }
+    /**
+     * 创建模型适配器
+     */
+    createAdapter(config) {
+        switch (config.type) {
+            case 'glm-4.6v':
+                return new GLM4VisionAdapter(config);
+            case 'siliconflow':
+                return new SiliconFlowAdapter(config);
+            default:
+                throw new ModelConfigError(`Unsupported model type: ${config.type}`);
+        }
+    }
+    /**
+     * 执行视觉分析
+     */
+    async analyze(input) {
+        const startTime = Date.now();
+        const requestId = Math.random().toString(36).substring(7);
+        logger.logRequest('Vision analysis started', {
+            requestId,
+            modelType: this.modelConfig.type,
+            imageInputLength: input.image.length,
+            hasTemplate: !!input.template
+        });
+        try {
+            // 1. 规范化图片输入
+            logger.debug('Normalizing image input', { requestId });
+            const normalizedImage = await normalizeImageInput(input.image);
+            logger.debug('Image normalized successfully', {
+                requestId,
+                type: normalizedImage.type,
+                mimeType: normalizedImage.mimeType,
+                dataLength: normalizedImage.dataUrl.length
+            });
+            // 2. 构建完整提示词
+            logger.debug('Building prompt', { requestId });
+            const fullPrompt = buildPrompt(input.template, input.prompt);
+            logger.debug('Prompt built', {
+                requestId,
+                promptLength: fullPrompt.length,
+                templateUsed: input.template || 'auto-detected'
+            });
+            // 3. 调用视觉模型
+            logger.logRequest('Calling vision model', {
+                requestId,
+                modelType: this.modelConfig.type,
+                modelName: this.modelConfig.name
+            });
+            const modelResponse = await this.adapter.analyzeWithResponse(normalizedImage.dataUrl, fullPrompt);
+            // 4. 构建响应
+            const response = {
+                content: modelResponse.content,
+                format: input.output_format,
+                metadata: {
+                    modelType: this.modelConfig.type,
+                    modelName: this.modelConfig.name,
+                    imageFormat: normalizedImage.mimeType,
+                    processingTimeMs: Date.now() - startTime,
+                    imageSize: normalizedImage.dataUrl.length
+                }
+            };
+            logger.logRequest('Vision analysis completed', {
+                requestId,
+                processingTimeMs: response.metadata.processingTimeMs,
+                outputLength: response.content.length
+            });
+            return response;
+        }
+        catch (error) {
+            logger.error('Vision analysis failed', error, {
+                requestId,
+                processingTimeMs: Date.now() - startTime
+            });
+            throw error;
+        }
+    }
+    /**
+     * 构建元数据
+     */
+    buildMetadata(normalizedImage, startTime) {
+        return {
+            modelType: this.modelConfig.type,
+            modelName: this.modelConfig.name,
+            imageFormat: normalizedImage.mimeType,
+            processingTimeMs: Date.now() - startTime,
+            imageSize: normalizedImage.dataUrl.length
+        };
+    }
+    /**
+     * 获取工具信息
+     */
+    getToolInfo() {
+        return {
+            modelConfig: this.modelConfig
+        };
+    }
+}
+/**
+ * 创建工具实例（单例模式）
+ */
+let toolInstance;
+export function getVisionTool(config) {
+    if (!toolInstance) {
+        toolInstance = new VisionTool(config);
+    }
+    return toolInstance;
+}
+//# sourceMappingURL=vision-tool.js.map

package/dist/src/tools/vision-tool.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"vision-tool.js","sourceRoot":"","sources":["../../../src/tools/vision-tool.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AACxB,OAAO,EAAe,eAAe,EAAE,mBAAmB,EAAE,MAAM,2BAA2B,CAAC;AAC9F,OAAO,EAAE,iBAAiB,EAAE,MAAM,4BAA4B,CAAC;AAC/D,OAAO,EAAE,kBAAkB,EAAE,MAAM,oCAAoC,CAAC;AAExE,OAAO,EAAE,mBAAmB,EAAwB,MAAM,yBAAyB,CAAC;AACpF,OAAO,EAAqB,gBAAgB,EAAE,MAAM,oBAAoB,CAAC;AACzE,OAAO,EAAE,MAAM,EAAE,MAAM,oBAAoB,CAAC;AAC5C,OAAO,EAAE,WAAW,EAAE,MAAM,sBAAsB,CAAC;AAEnD;;GAEG;AACH,MAAM,CAAC,MAAM,qBAAqB,GAAG,CAAC,CAAC,MAAM,CAAC;IAC5C;;;;;OAKG;IACH,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,gDAAgD,CAAC;IAE5E;;;;;OAKG;IACH,MAAM,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,qCAAqC,CAAC;IAElE;;;OAGG;IACH,aAAa,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC,QAAQ,CAAC,0BAA0B,CAAC;IAEvG;;;;;OAKG;IACH,QAAQ,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,+BAA+B,CAAC;CAC1E,CAAC,CAAC,MAAM,EAAE,CAAC;AAmBZ;;GAEG;AACH,MAAM,OAAO,UAAU;IACb,WAAW,CAAc;IACzB,OAAO,CAAqB;IAEpC,YAAY,MAAoB;QAC9B,OAAO;QACP,IAAI,CAAC,WAAW,GAAG,MAAM,IAAI,eAAe,EAAE,CAAC;QAE/C,OAAO;QACP,mBAAmB,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;QAEtC,QAAQ;QACR,IAAI,CAAC,OAAO,GAAG,IAAI,CAAC,aAAa,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;QAEpD,MAAM,CAAC,IAAI,CAAC,yBAAyB,EAAE;YACrC,SAAS,EAAE,IAAI,CAAC,WAAW,CAAC,IAAI;YAChC,SAAS,EAAE,IAAI,CAAC,WAAW,CAAC,IAAI;SACjC,CAAC,CAAC;IACL,CAAC;IAED;;OAEG;IACK,aAAa,CAAC,MAAmB;QACvC,QAAQ,MAAM,CAAC,IAAI,EAAE,CAAC;YACpB,KAAK,UAAU;gBACb,OAAO,IAAI,iBAAiB,CAAC,MAAM,CAAC,CAAC;YAEvC,KAAK,aAAa;gBAChB,OAAO,IAAI,kBAAkB,CAAC,MAAM,CAAC,CAAC;YAExC;gBACE,MAAM,IAAI,gBAAgB,CAAC,2BAA2B,MAAM,CAAC,IAAI,EAAE,CAAC,CAAC;QACzE,CAAC;IACH,CAAC;IAED;;OAEG;IACI,KAAK,CAAC,OAAO,CAAC,KAAsB;QACzC,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QAC7B,MAAM,SAAS,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC;QAE1D,MAAM,CAAC,UAAU,CAAC,yBAAyB,EAAE;YAC3C,SAAS;YACT,SAAS,EAAE,IAAI,CAAC,WAAW,CAAC,IAAI;YAChC,gBAAgB,EAAE,KAAK,CAAC,KAAK,CAAC,MAAM;YACpC,WAAW,EAAE,CAAC,CAAC,KAAK,CAAC,QAAQ;SAC9B,CAAC,CAAC;QAEH,IAAI,CAAC;YACH,aAAa;YACb,MAAM,CAAC,KAAK,CAAC,yBAAyB,EAAE,EAAE,SAAS,EAAE,CAAC,CAAC;YACvD,MAAM,eAAe,GAAG,MAAM,mBAAmB,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;YAE/D,MAAM,CAAC,KAAK,CAAC,+BAA+B,EAAE;gBAC5C,SAAS;gBACT,IAAI,EAAE,eAAe,CAAC,IAAI;gBAC1B,QAAQ,EAAE,eAAe,CAAC,QAAQ;gBAClC,UAAU,EAAE,eAAe,CAAC,OAAO,CAAC,MAAM;aAC3C,CAAC,CAAC;YAEH,aAAa;YACb,MAAM,CAAC,KAAK,CAAC,iBAAiB,EAAE,EAAE,SAAS,EAAE,CAAC,CAAC;YAC/C,MAAM,UAAU,GAAG,WAAW,CAAC,KAAK,CAAC,QAAQ,EAAE,KAAK,CAAC,MAAM,CAAC,CAAC;YAE7D,MAAM,CAAC,KAAK,CAAC,cAAc,EAAE;gBAC3B,SAAS;gBACT,YAAY,EAAE,UAAU,CAAC,MAAM;gBAC/B,YAAY,EAAE,KAAK,CAAC,QAAQ,IAAI,eAAe;aAChD,CAAC,CAAC;YAEH,YAAY;YACZ,MAAM,CAAC,UAAU,CAAC,sBAAsB,EAAE;gBACxC,SAAS;gBACT,SAAS,EAAE,IAAI,CAAC,WAAW,CAAC,IAAI;gBAChC,SAAS,EAAE,IAAI,CAAC,WAAW,CAAC,IAAI;aACjC,CAAC,CAAC;YAEH,MAAM,aAAa,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,mBAAmB,CAC1D,eAAe,CAAC,OAAO,EACvB,UAAU,CACX,CAAC;YAEF,UAAU;YACV,MAAM,QAAQ,GAAuB;gBACnC,OAAO,EAAE,aAAa,CAAC,OAAO;gBAC9B,MAAM,EAAE,KAAK,CAAC,aAAa;gBAC3B,QAAQ,EAAE;oBACR,SAAS,EAAE,IAAI,CAAC,WAAW,CAAC,IAAI;oBAChC,SAAS,EAAE,IAAI,CAAC,WAAW,CAAC,IAAI;oBAChC,WAAW,EAAE,eAAe,CAAC,QAAQ;oBACrC,gBAAgB,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,SAAS;oBACxC,SAAS,EAAE,eAAe,CAAC,OAAO,CAAC,MAAM;iBAC1C;aACF,CAAC;YAEF,MAAM,CAAC,UAAU,CAAC,2BAA2B,EAAE;gBAC7C,SAAS;gBACT,gBAAgB,EAAE,QAAQ,CAAC,QAAQ,CAAC,gBAAgB;gBACpD,YAAY,EAAE,QAAQ,CAAC,OAAO,CAAC,MAAM;aACtC,CAAC,CAAC;YAEH,OAAO,QAAQ,CAAC;QAClB,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,MAAM,CAAC,KAAK,CAAC,wBAAwB,EAAE,KAAc,EAAE;gBACrD,SAAS;gBACT,gBAAgB,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,SAAS;aACzC,CAAC,CAAC;YAEH,MAAM,KAAK,CAAC;QACd,CAAC;IACH,CAAC;IAED;;OAEG;IACK,aAAa,CAAC,eAAqC,EAAE,SAAiB;QAC5E,OAAO;YACL,SAAS,EAAE,IAAI,CAAC,WAAW,CAAC,IAAI;YAChC,SAAS,EAAE,IAAI,CAAC,WAAW,CAAC,IAAI;YAChC,WAAW,EAAE,eAAe,CAAC,QAAQ;YACrC,gBAAgB,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,SAAS;YACxC,SAAS,EAAE,eAAe,CAAC,OAAO,CAAC,MAAM;SAC1C,CAAC;IACJ,CAAC;IAED;;OAEG;IACI,WAAW;QAGhB,OAAO;YACL,WAAW,EAAE,IAAI,CAAC,WAAW;SAC9B,CAAC;IACJ,CAAC;CACF;AAED;;GAEG;AACH,IAAI,YAAoC,CAAC;AAEzC,MAAM,UAAU,aAAa,CAAC,MAAoB;IAChD,IAAI,CAAC,YAAY,EAAE,CAAC;QAClB,YAAY,GAAG,IAAI,UAAU,CAAC,MAAM,CAAC,CAAC;IACxC,CAAC;IACD,OAAO,YAAY,CAAC;AACtB,CAAC"}

package/dist/src/utils/errors.d.ts

@@ -0,0 +1,65 @@
+/**
+ * Vision MCP Error Types
+ *
+ * @description 统一错误处理系统，提供清晰的错误分类和信息
+ */
+export declare enum VisionMCPErrorCode {
+    INVALID_INPUT = "INVALID_INPUT",
+    MODEL_CONFIG_ERROR = "MODEL_CONFIG_ERROR",
+    IMAGE_LOAD_ERROR = "IMAGE_LOAD_ERROR",
+    MODEL_API_ERROR = "MODEL_API_ERROR",
+    TIMEOUT_ERROR = "TIMEOUT_ERROR",
+    UNKNOWN_ERROR = "UNKNOWN_ERROR"
+}
+export interface VisionMCPErrorOptions {
+    message: string;
+    code: VisionMCPErrorCode;
+    details?: Record<string, any>;
+    cause?: unknown;
+}
+export declare class VisionMCPError extends Error {
+    readonly code: VisionMCPErrorCode;
+    readonly details?: Record<string, any>;
+    readonly timestamp: Date;
+    constructor({ message, code, details, cause }: VisionMCPErrorOptions);
+    toJSON(): {
+        error: {
+            message: string;
+            code: VisionMCPErrorCode;
+            details: {
+                [x: string]: any;
+            };
+            timestamp: string;
+        };
+    };
+}
+export declare class InvalidInputError extends VisionMCPError {
+    constructor(message: string, details?: Record<string, any>, cause?: unknown);
+}
+export declare class ModelConfigError extends VisionMCPError {
+    constructor(message: string, details?: Record<string, any>, cause?: unknown);
+}
+export declare class ImageLoadError extends VisionMCPError {
+    constructor(message: string, details?: Record<string, any>, cause?: unknown);
+}
+export declare class ModelAPIError extends VisionMCPError {
+    constructor(message: string, details?: Record<string, any>, cause?: unknown);
+}
+export declare class TimeoutError extends VisionMCPError {
+    constructor(message: string, details?: Record<string, any>, cause?: unknown);
+}
+/**
+ * 将未知错误转换为 VisionMCPError
+ */
+export declare function toVisionMCPError(error: unknown, context?: string): VisionMCPError;
+/**
+ * 将错误映射为 MCP 工具响应格式
+ */
+export declare function toMCPErrorResponse(error: VisionMCPError): {
+    content: {
+        type: "text";
+        text: string;
+    }[];
+    isError: boolean;
+};
+//# sourceMappingURL=errors.d.ts.map

package/dist/src/utils/errors.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"errors.d.ts","sourceRoot":"","sources":["../../../src/utils/errors.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,oBAAY,kBAAkB;IAC5B,aAAa,kBAAkB;IAC/B,kBAAkB,uBAAuB;IACzC,gBAAgB,qBAAqB;IACrC,eAAe,oBAAoB;IACnC,aAAa,kBAAkB;IAC/B,aAAa,kBAAkB;CAChC;AAED,MAAM,WAAW,qBAAqB;IACpC,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,kBAAkB,CAAC;IACzB,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC;IAC9B,KAAK,CAAC,EAAE,OAAO,CAAC;CACjB;AAED,qBAAa,cAAe,SAAQ,KAAK;IACvC,SAAgB,IAAI,EAAE,kBAAkB,CAAC;IACzC,SAAgB,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC;IAC9C,SAAgB,SAAS,EAAE,IAAI,CAAC;gBAEpB,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,KAAK,EAAE,EAAE,qBAAqB;IAiBpE,MAAM;;;;;;;;;;CAaP;AAED,qBAAa,iBAAkB,SAAQ,cAAc;gBACvC,OAAO,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,EAAE,KAAK,CAAC,EAAE,OAAO;CAS5E;AAED,qBAAa,gBAAiB,SAAQ,cAAc;gBACtC,OAAO,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,EAAE,KAAK,CAAC,EAAE,OAAO;CAS5E;AAED,qBAAa,cAAe,SAAQ,cAAc;gBACpC,OAAO,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,EAAE,KAAK,CAAC,EAAE,OAAO;CAS5E;AAED,qBAAa,aAAc,SAAQ,cAAc;gBACnC,OAAO,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,EAAE,KAAK,CAAC,EAAE,OAAO;CAS5E;AAED,qBAAa,YAAa,SAAQ,cAAc;gBAClC,OAAO,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,EAAE,KAAK,CAAC,EAAE,OAAO;CAS5E;AAED;;GAEG;AACH,wBAAgB,gBAAgB,CAAC,KAAK,EAAE,OAAO,EAAE,OAAO,CAAC,EAAE,MAAM,GAAG,cAAc,CA+BjF;AAED;;GAEG;AACH,wBAAgB,kBAAkB,CAAC,KAAK,EAAE,cAAc;;;;;;EAUvD"}