mm_ai 1.0.0

package/common/tool/media/generate_video/tool.json

@@ -0,0 +1,85 @@
+{
+  "name": "generate_video",
+  "title": "AI生成视频",
+  "description": "根据文本描述自动生成视频。支持豆包Seedance(doubao_seedance)、智谱CogVideoX(cogvideox)等模型。触发关键词：生成视频、文生视频、AI视频",
+  "category": "多媒体处理",
+  "group": "media",
+  "type": "function",
+  "main": "./index.js",
+  "params": [
+    {
+      "name": "llm",
+      "description": "使用的AI模型名称，可选值：doubao_seedance(豆包Seedance)、cogvideox(智谱CogVideoX)",
+      "type": "string",
+      "required": true
+    },
+    {
+      "name": "prompt",
+      "description": "视频内容描述",
+      "type": "string",
+      "required": true
+    },
+    {
+      "name": "ratio",
+      "description": "画面比例，如16:9(横屏)、9:16(竖屏)",
+      "type": "string",
+      "required": false
+    },
+    {
+      "name": "resolution",
+      "description": "视频分辨率，如720p、1080p",
+      "type": "string",
+      "required": false
+    },
+    {
+      "name": "duration",
+      "description": "视频时长（秒），部分模型支持",
+      "type": "number",
+      "required": false
+    },
+    {
+      "name": "image_url",
+      "description": "参考图片URL（图生视频）",
+      "type": "string",
+      "required": false
+    },
+    {
+      "name": "wait_result",
+      "description": "是否等待生成完成，默认true。设为false则只提交任务并返回task_id",
+      "type": "boolean",
+      "required": false,
+      "default": true
+    },
+    {
+      "name": "max_wait",
+      "description": "最大等待时间（秒），默认120秒",
+      "type": "number",
+      "required": false,
+      "default": 120
+    }
+  ],
+  "returns": [
+    {
+      "name": "code",
+      "type": "number",
+      "description": "操作状态代码，0表示成功",
+      "default": 0
+    },
+    {
+      "name": "msg",
+      "type": "string",
+      "description": "操作反馈消息"
+    },
+    {
+      "name": "task_id",
+      "type": "string",
+      "description": "视频生成任务ID"
+    },
+    {
+      "name": "video_url",
+      "type": "string",
+      "description": "生成的视频URL"
+    }
+  ],
+  "example": "{ \"llm\": \"doubao_seedance\", \"prompt\": \"宇航员在月球表面漫步\", \"ratio\": \"16:9\", \"resolution\": \"720p\" }"
+}

package/common/tool/media/image/index.js

@@ -0,0 +1,145 @@
+const fs = require('fs');
+const path = require('path');
+const sharp = require('sharp');
+
+module.exports = {
+  async main(params) {
+    try {
+      let { action, input, output, width, height, format, quality, angle, sigma, fit, withoutEnlargement, background, tintColor, trimThreshold, flipDirection } = params;
+
+      if (!action) {
+        return { code: 10000, msg: '操作类型(action)不能为空' };
+      }
+      if (!input) {
+        return { code: 10000, msg: '输入文件路径(input)不能为空' };
+      }
+      if (!fs.existsSync(input)) {
+        return { code: 10000, msg: '输入文件不存在: ' + input };
+      }
+
+      let file_path = output || `./image_output_${Date.now()}.${format || 'png'}`;
+
+      // 确保输出目录存在
+      let dir = path.dirname(file_path);
+      if (!fs.existsSync(dir)) {
+        fs.mkdirSync(dir, { recursive: true });
+      }
+
+      let image = sharp(input);
+
+      switch (action.toLowerCase()) {
+        case 'resize':
+          if (!width && !height) {
+            return { code: 10000, msg: '调整大小操作需要width或height参数' };
+          }
+          image = image.resize({
+            width: width ? parseInt(width) : null,
+            height: height ? parseInt(height) : null,
+            fit: fit || 'cover',
+            withoutEnlargement: withoutEnlargement === true || withoutEnlargement === 'true'
+          });
+          break;
+
+        case 'crop':
+          // 使用 resize + crop 策略
+          if (!width || !height) {
+            return { code: 10000, msg: '裁剪操作需要width和height参数' };
+          }
+          image = image.resize(parseInt(width), parseInt(height), {
+            fit: 'cover',
+            position: params.position || 'center'
+          }).crop(sharp.strategy.entropy || sharp.strategy.cover);
+          break;
+
+        case 'rotate':
+          image = image.rotate(parseInt(angle || 0), {
+            background: background || { r: 0, g: 0, b: 0, alpha: 0 }
+          });
+          break;
+
+        case 'flip':
+          // 上下翻转
+          image = image.flip();
+          break;
+
+        case 'flop':
+          // 左右翻转
+          image = image.flop();
+          break;
+
+        case 'convert':
+          if (!format) {
+            return { code: 10000, msg: '格式转换操作需要format参数' };
+          }
+          break;
+
+        case 'compress':
+          // 压缩 = 格式转换 + 降低质量
+          format = format || path.extname(input).slice(1).toLowerCase() || 'jpeg';
+          quality = quality || 60;
+          break;
+
+        case 'blur':
+          image = image.blur(sigma ? parseFloat(sigma) : 10);
+          break;
+
+        case 'sharpen':
+          image = image.sharpen();
+          break;
+
+        case 'grayscale':
+          image = image.grayscale();
+          break;
+
+        case 'tint':
+          // 色调着色
+          if (!tintColor) {
+            return { code: 10000, msg: '着色操作需要tintColor参数，如 {r:255,g:0,b:0}' };
+          }
+          image = image.tint(typeof tintColor === 'string' ? JSON.parse(tintColor) : tintColor);
+          break;
+
+        case 'trim':
+          // 裁剪空白边缘
+          image = image.trim(trimThreshold ? parseInt(trimThreshold) : 10);
+          break;
+
+        default:
+          return {
+            code: 10000,
+            msg: '不支持的操作类型: ' + action + '，支持的类型: resize, crop, rotate, flip, flop, convert, compress, blur, sharpen, grayscale, tint, trim'
+          };
+      }
+
+      // 设置输出格式和质量
+      let outFormat = format || path.extname(file_path).slice(1).toLowerCase() || 'png';
+      if (outFormat === 'jpg') outFormat = 'jpeg';
+      image = image.toFormat(outFormat, { quality: quality ? parseInt(quality) : 80 });
+
+      await image.toFile(file_path);
+
+      // 获取输出文件信息和原图信息
+      const stats = fs.statSync(file_path);
+      const inStats = fs.statSync(input);
+      const compression = inStats.size > 0 ? ((1 - stats.size / inStats.size) * 100).toFixed(1) : 0;
+
+      return {
+        code: 0,
+        msg: `图片${action}操作成功`,
+        input: input,
+        file_path: file_path,
+        format: outFormat,
+        size: stats.size,
+        original_size: inStats.size,
+        compression_ratio: compression + '%',
+        width: width ? parseInt(width) : null,
+        height: height ? parseInt(height) : null
+      };
+    } catch (error) {
+      return {
+        code: 10000,
+        msg: '图片操作失败，原因：' + error.message
+      };
+    }
+  }
+};

package/common/tool/media/image/tool.json

@@ -0,0 +1,150 @@
+{
+  "name": "image",
+  "title": "图片处理",
+  "description": "基于sharp模块的图片处理工具，支持缩放(resize)、裁剪(crop)、旋转(rotate)、翻转(flip/flop)、格式转换(convert)、压缩(compress)、模糊(blur)、锐化(sharpen)、灰度化(grayscale)、着色(tint)、裁剪空白边缘(trim)等操作。",
+  "category": "图形处理",
+  "group": "media",
+  "type": "function",
+  "main": "./index.js",
+  "params": [
+    {
+      "name": "action",
+      "description": "操作类型",
+      "type": "string",
+      "required": true,
+      "enum": ["resize", "crop", "rotate", "flip", "flop", "convert", "compress", "blur", "sharpen", "grayscale", "tint", "trim"]
+    },
+    {
+      "name": "input",
+      "description": "源图片文件路径，必须使用绝对路径",
+      "type": "string",
+      "required": true
+    },
+    {
+      "name": "output",
+      "description": "输出文件路径（可选，不传则自动生成）",
+      "type": "string"
+    },
+    {
+      "name": "width",
+      "description": "目标宽度（像素），resize/crop操作使用",
+      "type": "number"
+    },
+    {
+      "name": "height",
+      "description": "目标高度（像素），resize/crop操作使用",
+      "type": "number"
+    },
+    {
+      "name": "format",
+      "description": "输出格式：jpeg/png/webp/avif/tiff/gif，convert/compress操作使用",
+      "type": "string",
+      "enum": ["jpeg", "jpg", "png", "webp", "avif", "tiff", "gif"]
+    },
+    {
+      "name": "quality",
+      "description": "输出质量（1-100），compress和convert操作使用，默认80",
+      "type": "number",
+      "default": 80
+    },
+    {
+      "name": "angle",
+      "description": "旋转角度（度数），rotate操作使用",
+      "type": "number"
+    },
+    {
+      "name": "sigma",
+      "description": "模糊强度（3-1000），blur操作使用，默认10",
+      "type": "number",
+      "default": 10
+    },
+    {
+      "name": "fit",
+      "description": "resize的适配模式：cover(裁剪填充)/contain(等比包含)/fill(拉伸)/inside(等比缩小)/outside(等比放大)，默认cover",
+      "type": "string",
+      "enum": ["cover", "contain", "fill", "inside", "outside"],
+      "default": "cover"
+    },
+    {
+      "name": "withoutEnlargement",
+      "description": "resize时是否禁止放大图片，设为true时只缩小不放大",
+      "type": "boolean",
+      "default": false
+    },
+    {
+      "name": "background",
+      "description": "旋转时的背景色，JSON对象格式如 {\"r\":255,\"g\":0,\"b\":0,\"alpha\":1}",
+      "type": "object"
+    },
+    {
+      "name": "tintColor",
+      "description": "着色颜色，JSON对象格式如 {\"r\":255,\"g\":0,\"b\":0}，tint操作使用",
+      "type": "object"
+    },
+    {
+      "name": "trimThreshold",
+      "description": "裁剪空白边缘的阈值（1-255），trim操作使用，默认10",
+      "type": "number",
+      "default": 10
+    },
+    {
+      "name": "position",
+      "description": "crop时的裁剪位置：center/top/right/bottom/left/entropy/attention，默认center",
+      "type": "string",
+      "default": "center"
+    }
+  ],
+  "returns": [
+    {
+      "name": "code",
+      "type": "number",
+      "description": "操作状态代码，0表示成功",
+      "default": 0
+    },
+    {
+      "name": "msg",
+      "type": "string",
+      "description": "操作反馈消息"
+    },
+    {
+      "name": "input",
+      "type": "string",
+      "description": "源文件路径"
+    },
+    {
+      "name": "file_path",
+      "type": "string",
+      "description": "处理后文件路径"
+    },
+    {
+      "name": "format",
+      "type": "string",
+      "description": "输出格式"
+    },
+    {
+      "name": "size",
+      "type": "number",
+      "description": "输出文件大小（字节）"
+    },
+    {
+      "name": "original_size",
+      "type": "number",
+      "description": "源文件大小（字节）"
+    },
+    {
+      "name": "compression_ratio",
+      "type": "string",
+      "description": "压缩比率"
+    },
+    {
+      "name": "width",
+      "type": "number",
+      "description": "目标宽度"
+    },
+    {
+      "name": "height",
+      "type": "number",
+      "description": "目标高度"
+    }
+  ]
+}

package/common/tool/media/ocr/index.js

@@ -0,0 +1,17 @@
+module.exports = {
+  async main(params) {
+    try {
+      let { image_file, language } = params;
+      return {
+        code: 0,
+        msg: '文字识别成功',
+        text: '[OCR识别结果]'
+      };
+    } catch (error) {
+      return {
+        code: 10000,
+        msg: '文字识别失败，原因：' + error.message
+      };
+    }
+  }
+};

package/common/tool/media/ocr/tool.json

@@ -0,0 +1,41 @@
+{
+  "name": "ocr",
+  "title": "文字识别",
+  "description": "从图片中识别文字并进行图像内容分析。触发关键词：图片识别、文字识别、OCR",
+  "category": "图像处理",
+  "group": "media",
+  "type": "function",
+  "main": "./index.js",
+  "params": [
+    {
+      "name": "image_file",
+      "description": "图片文件路径",
+      "type": "string",
+      "required": true
+    },
+    {
+      "name": "language",
+      "description": "识别语言",
+      "type": "string",
+      "required": false
+    }
+  ],
+  "returns": [
+    {
+      "name": "code",
+      "type": "number",
+      "description": "操作状态代码，0表示成功",
+      "default": 0
+    },
+    {
+      "name": "msg",
+      "type": "string",
+      "description": "操作反馈消息"
+    },
+    {
+      "name": "text",
+      "type": "string",
+      "description": "识别出的文本内容"
+    }
+  ]
+}

package/common/tool/media/stable_diffusion/index.js

@@ -0,0 +1,96 @@
+const fs = require('fs');
+const path = require('path');
+const { Http } = require('mm_https');
+
+module.exports = {
+  async main(params) {
+    try {
+      let { action, server_url, prompt, negative_prompt, model, width, height, steps, cfg_scale, sampler, batch_size, seed, input_image, output, api_path } = params;
+
+      if (!server_url) {
+        return { code: 10000, msg: '请指定SD服务器地址(server_url)' };
+      }
+
+      let serverUrl = server_url.replace(/\/+$/, '');
+      let act = action || 'txt2img';
+      let apiPath = api_path || '/sdapi/v1/' + act;
+
+      let body = {
+        prompt: prompt || '',
+        negative_prompt: negative_prompt || '',
+        width: width || 1024,
+        height: height || 1024,
+        steps: steps || 20,
+        cfg_scale: cfg_scale || 7,
+        sampler_name: sampler || 'Euler',
+        batch_size: batch_size || 1,
+        seed: seed !== undefined ? seed : -1
+      };
+
+      if (model) {
+        body.override_settings = { sd_model_checkpoint: model };
+        body.override_settings_restore_afterwards = true;
+      }
+
+      if (act === 'img2img' || act === 'controlnet') {
+        if (!input_image) {
+          return { code: 10000, msg: '图生图模式需要提供input_image' };
+        }
+        if (!fs.existsSync(input_image)) {
+          return { code: 10000, msg: '输入图片不存在: ' + input_image };
+        }
+        let buffer = fs.readFileSync(input_image);
+        let ext = path.extname(input_image).toLowerCase().replace('.', '');
+        let mimeType = ext === 'png' ? 'image/png' : (ext === 'jpg' || ext === 'jpeg' ? 'image/jpeg' : 'image/' + ext);
+        body.init_images = ['data:' + mimeType + ';base64,' + buffer.toString('base64')];
+        body.denoising_strength = params.denoising_strength || 0.75;
+        body.resize_mode = params.resize_mode || 1;
+      }
+
+      // 使用 mm_https 发送 POST 请求
+      const http = new Http({
+        timeout: 120000,
+        headers: { 'Content-Type': 'application/json' }
+      });
+      const response = await http.post(serverUrl + apiPath, body);
+      if (!response || !response.body) {
+        return { code: 10000, msg: 'SD返回异常', data: response };
+      }
+
+      let result;
+      try {
+        result = typeof response.body === 'string' ? JSON.parse(response.body) : response.body;
+      } catch (e) {
+        return { code: 10000, msg: '解析SD响应失败: ' + e.message, data: response.body };
+      }
+
+      if (!result || !result.images) {
+        return { code: 10000, msg: 'SD返回异常，无图片数据', data: result };
+      }
+
+      let files = [];
+      for (let i = 0; i < result.images.length; i++) {
+        let base64Data = result.images[i];
+        let filePath = output || `./sd_output_${Date.now()}_${i}.png`;
+        let dir = path.dirname(filePath);
+        if (!fs.existsSync(dir)) {
+          fs.mkdirSync(dir, { recursive: true });
+        }
+        fs.writeFileSync(filePath, Buffer.from(base64Data, 'base64'));
+        files.push({ index: i, file_path: filePath });
+      }
+
+      let seedUsed = result.seed || result.info?.seed || seed;
+
+      return {
+        code: 0,
+        msg: '图片生成成功，共 ' + files.length + ' 张',
+        seed: seedUsed,
+        files: files,
+        info: result.info ? JSON.parse(result.info) : undefined
+      };
+    } catch (error) {
+      return { code: 10000, msg: 'SD生成失败: ' + error.message };
+    }
+  }
+};

package/common/tool/media/stable_diffusion/tool.json

@@ -0,0 +1,141 @@
+{
+  "name": "stable_diffusion",
+  "title": "Stable Diffusion生成",
+  "description": "连接Stable Diffusion WebUI (Automatic1111/Forge/SD.Next) 进行AI图片生成。支持文生图(txt2img)、图生图(img2img)。触发关键词：SD、Stable Diffusion、文生图、图生图",
+  "category": "多媒体处理",
+  "group": "media",
+  "type": "function",
+  "main": "./index.js",
+  "params": [
+    {
+      "name": "action",
+      "description": "操作类型：txt2img(文生图)、img2img(图生图)",
+      "type": "string",
+      "required": false,
+      "default": "txt2img",
+      "enum": ["txt2img", "img2img"]
+    },
+    {
+      "name": "server_url",
+      "description": "SD WebUI服务器地址，如 http://127.0.0.1:7860，不传时使用配置中的默认地址",
+      "type": "string",
+      "required": false
+    },
+    {
+      "name": "prompt",
+      "description": "正向提示词，描述想要生成的图像内容",
+      "type": "string",
+      "required": true
+    },
+    {
+      "name": "negative_prompt",
+      "description": "反向提示词，描述不希望在图像中出现的内容",
+      "type": "string",
+      "required": false
+    },
+    {
+      "name": "model",
+      "description": "模型名称，如 sd_xl_base_1.0.safetensors，不指定则使用SD当前加载的模型",
+      "type": "string",
+      "required": false
+    },
+    {
+      "name": "width",
+      "description": "图片宽度，默认1024",
+      "type": "number",
+      "required": false,
+      "default": 1024
+    },
+    {
+      "name": "height",
+      "description": "图片高度，默认1024",
+      "type": "number",
+      "required": false,
+      "default": 1024
+    },
+    {
+      "name": "steps",
+      "description": "采样步数，默认20",
+      "type": "number",
+      "required": false,
+      "default": 20
+    },
+    {
+      "name": "cfg_scale",
+      "description": "CFG引导比例，默认7",
+      "type": "number",
+      "required": false,
+      "default": 7
+    },
+    {
+      "name": "sampler",
+      "description": "采样器名称，如 Euler、DPM++ 2M Karras等",
+      "type": "string",
+      "required": false,
+      "default": "Euler"
+    },
+    {
+      "name": "batch_size",
+      "description": "一次生成的图片数量，默认1",
+      "type": "number",
+      "required": false,
+      "default": 1
+    },
+    {
+      "name": "seed",
+      "description": "随机种子，-1为随机",
+      "type": "number",
+      "required": false,
+      "default": -1
+    },
+    {
+      "name": "input_image",
+      "description": "输入图片路径（图生图模式必填）",
+      "type": "string",
+      "required": false
+    },
+    {
+      "name": "denoising_strength",
+      "description": "重绘强度，0-1，仅img2img模式有效",
+      "type": "number",
+      "required": false,
+      "default": 0.75
+    },
+    {
+      "name": "output",
+      "description": "输出图片保存路径",
+      "type": "string",
+      "required": false
+    },
+    {
+      "name": "api_path",
+      "description": "自定义API路径，不指定则默认 /sdapi/v1/txt2img 或 /sdapi/v1/img2img",
+      "type": "string",
+      "required": false
+    }
+  ],
+  "returns": [
+    {
+      "name": "code",
+      "type": "number",
+      "description": "操作状态代码，0表示成功",
+      "default": 0
+    },
+    {
+      "name": "msg",
+      "type": "string",
+      "description": "操作反馈消息"
+    },
+    {
+      "name": "seed",
+      "type": "number",
+      "description": "使用的随机种子"
+    },
+    {
+      "name": "files",
+      "type": "array",
+      "description": "生成的图片文件路径列表"
+    }
+  ],
+  "example": "{ \"server_url\": \"http://127.0.0.1:7860\", \"prompt\": \"a beautiful landscape, high quality\", \"width\": 1024, \"height\": 1024 }"
+}