@make-u-free/migi 0.3.1 → 0.3.3

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@make-u-free/migi",
-  "version": "0.3.1",
+  "version": "0.3.3",
   "description": "Your AI right-hand agent. Works anywhere, with any LLM API.",
   "type": "module",
   "bin": {
@@ -10,10 +10,12 @@
     "start": "node bin/migi.js"
   },
   "dependencies": {
+    "adm-zip": "^0.5.16",
     "chalk": "^5.3.0",
     "dotenv": "^16.4.0",
     "glob": "^11.0.0",
     "openai": "^4.0.0",
+    "pdf-parse": "^2.4.5",
     "xlsx": "^0.18.5"
   },
   "engines": {

package/src/agent.js

@@ -7,8 +7,9 @@ import { httpsAgent } from './tls.js'
 
 export class MigiAgent {
   constructor({ context = '', promptFn = null, apiKey = null, model = 'gpt-4.1-2025-04-14', name = 'Migi', userName = '', teamsWebhookUrl = '' } = {}) {
+    this.apiKey = apiKey || process.env.OPENAI_API_KEY
     this.client = new OpenAI({
-      apiKey: apiKey || process.env.OPENAI_API_KEY,
+      apiKey: this.apiKey,
       ...(httpsAgent ? { httpAgent: httpsAgent } : {})
     })
     this.model = model
@@ -123,7 +124,11 @@ ${userNameLine}
 
           if (approved) {
             try {
-              result = await executeTool(name, args, { teamsWebhookUrl: this.teamsWebhookUrl })
+              result = await executeTool(name, args, {
+              teamsWebhookUrl: this.teamsWebhookUrl,
+              apiKey: this.apiKey,
+              model: this.model
+            })
             } catch (err) {
               result = `エラー: ${err.message}`
             }

package/src/tools.js

@@ -4,9 +4,16 @@ import { dirname, extname } from 'path'
 import { request } from 'https'
 import { glob } from 'glob'
 import xlsxPkg from 'xlsx'
+import pdfParse from 'pdf-parse'
+import AdmZip from 'adm-zip'
+import OpenAI from 'openai'
 import { httpsAgent } from './tls.js'
 const { readFile: xlsxReadFile, utils: xlsxUtils } = xlsxPkg
 
+const IMAGE_EXTS = new Set(['.jpg', '.jpeg', '.png', '.gif', '.webp'])
+const OFFICE_EXTS = new Set(['.pdf', '.pptx', '.ppt', '.docx', '.doc', '.odp', '.odt'])
+const MIME = { '.jpg': 'image/jpeg', '.jpeg': 'image/jpeg', '.png': 'image/png', '.gif': 'image/gif', '.webp': 'image/webp' }
+
 // ---- OpenAI ツールスキーマ定義 ----
 
 export const toolSchemas = [
@@ -14,7 +21,7 @@ export const toolSchemas = [
     type: 'function',
     function: {
       name: 'read_file',
-      description: 'ファイルの内容を読み込む',
+      description: 'ファイルの内容を読み込む。テキスト・Excel・PDF・Word・PowerPoint・画像に対応',
       parameters: {
         type: 'object',
         properties: {
@@ -124,6 +131,8 @@ export async function executeTool(name, args, opts = {}) {
     case 'read_file': {
       if (!existsSync(args.path)) return `エラー: ファイルが見つかりません: ${args.path}`
       const ext = extname(args.path).toLowerCase()
+
+      // Excel
       if (ext === '.xlsx' || ext === '.xls') {
         const workbook = xlsxReadFile(args.path)
         const result = []
@@ -134,6 +143,64 @@ export async function executeTool(name, args, opts = {}) {
         }
         return result.join('\n\n')
       }
+
+      // PDF
+      if (ext === '.pdf') {
+        try {
+          const buf = readFileSync(args.path)
+          const data = await pdfParse(buf)
+          return data.text?.trim() || '(テキストが抽出できませんでした)'
+        } catch (err) {
+          return `エラー: PDFの解析に失敗しました: ${err.message}`
+        }
+      }
+
+      // PowerPoint（PPTX）/ Word（DOCX）→ ZIPを展開してXMLからテキスト抽出
+      if (['.pptx', '.ppt', '.docx', '.doc', '.odp', '.odt'].includes(ext)) {
+        try {
+          const zip = new AdmZip(args.path)
+          const entries = zip.getEntries()
+          const xmlTexts = []
+          for (const entry of entries) {
+            const name = entry.entryName
+            const isSlide = name.startsWith('ppt/slides/slide') && name.endsWith('.xml')
+            const isDoc = name === 'word/document.xml'
+            const isOdp = name === 'content.xml'
+            if (isSlide || isDoc || isOdp) {
+              const xml = entry.getData().toString('utf-8')
+              const text = xml.replace(/<[^>]+>/g, ' ').replace(/\s+/g, ' ').trim()
+              if (text) xmlTexts.push(text)
+            }
+          }
+          return xmlTexts.join('\n\n') || '(テキストが抽出できませんでした)'
+        } catch (err) {
+          return `エラー: ファイルの解析に失敗しました: ${err.message}`
+        }
+      }
+
+      // 画像 → Vision API で内容を説明させる
+      if (IMAGE_EXTS.has(ext)) {
+        if (!opts.apiKey) return 'エラー: 画像読み込みにはAPIキーが必要です'
+        const base64 = readFileSync(args.path).toString('base64')
+        const mimeType = MIME[ext] || 'image/jpeg'
+        const client = new OpenAI({
+          apiKey: opts.apiKey,
+          ...(httpsAgent ? { httpAgent: httpsAgent } : {})
+        })
+        const res = await client.chat.completions.create({
+          model: opts.model || 'gpt-4.1-2025-04-14',
+          messages: [{
+            role: 'user',
+            content: [
+              { type: 'image_url', image_url: { url: `data:${mimeType};base64,${base64}` } },
+              { type: 'text', text: 'この画像の内容を詳しく説明してください。テキストが含まれている場合はすべて書き起こしてください。' }
+            ]
+          }],
+          max_tokens: 2000
+        })
+        return res.choices[0].message.content
+      }
+
       return readFileSync(args.path, 'utf-8')
     }