@nogataka/imgen 0.1.0 → 0.3.0

package/README.md

@@ -10,25 +10,32 @@ npm install -g @nogataka/imgen
 
 ## セットアップ
 
-### 環境変数（推奨）
+### 環境変数
 
 ```bash
 export AZURE_OPENAI_ENDPOINT="https://your-resource.openai.azure.com"
 export AZURE_OPENAI_API_KEY="your-api-key"
 export AZURE_OPENAI_DEPLOYMENT_NAME="gpt-5.1"
 export AZURE_OPENAI_DEPLOYMENT_NAME_IMAGE="gpt-image-1.5"
-export AZURE_OPENAI_API_VERSION="2024-02-15-preview"
-export AZURE_OPENAI_IMAGE_API_VERSION="2025-04-01-preview"
+export AZURE_OPENAI_API_VERSION="2024-02-15-preview"        # 省略可
+export AZURE_OPENAI_IMAGE_API_VERSION="2025-04-01-preview"  # 省略可
 ```
 
-### 対話式設定
+### .env ファイル
+
+環境変数の代わりに `.env` ファイルで設定できます。探索順:
+
+1. `cwd/.env`（カレントディレクトリ）
+2. `~/.imgen/.env`
 
 ```bash
-imgen configure
+# .env
+AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com
+AZURE_OPENAI_API_KEY=your-api-key
+AZURE_OPENAI_DEPLOYMENT_NAME=gpt-5.1
+AZURE_OPENAI_DEPLOYMENT_NAME_IMAGE=gpt-image-1.5
 ```
 
-Azure OpenAI の接続情報やデフォルト値を対話的に設定できます。設定は `~/.imgen/config.json` に保存されます。
-
 環境変数が設定されている場合、環境変数が優先されます。
 
 ## 使い方
@@ -54,7 +61,7 @@ imgen image gen "商品写真" -p builtin:landscape
 | `--output <path>` | `-o` | 出力先ファイルまたはディレクトリ | カレントディレクトリ |
 | `--json` | | JSON 形式で出力 | - |
 | `--dry-run` | | API を呼ばずに設定を確認 | - |
-| `--debug` | `-d` | デバッグログを有効化 | - |
+| `--debug` | `-d` | デバッグ情報を表示 | - |
 
 ### 画像編集
 
@@ -95,11 +102,7 @@ imgen image explain chart.png -c "Q4 sales report" -f json -o description.json
 
 対応言語: `ja`(日本語), `en`(英語), `zh`(中国語), `ko`(韓国語), `es`(スペイン語), `fr`(フランス語), `de`(ドイツ語), `it`(イタリア語), `ru`(ロシア語), `vi`(ベトナム語)
 
-## プリセット
-
-よく使う設定をプリセットとして保存・呼び出しできます。
-
-### ビルトインプリセット
+## ビルトインプリセット
 
 | 名前 | サイズ | 品質 | 用途 |
 |------|--------|------|------|
@@ -109,32 +112,95 @@ imgen image explain chart.png -c "Q4 sales report" -f json -o description.json
 | `builtin:draft` | 1024x1024 | low | 下書き、プロトタイプ |
 | `builtin:photo` | 1536x1024 | high | 商品写真 |
 
-### カスタムプリセット
+## 設定の優先順位
+
+1. CLI オプション（最優先）
+2. プリセット値（`-p` 指定時）
+3. デフォルト値
+
+## SDK として使う
+
+imgen は CLI だけでなく、Node.js ライブラリとしても利用できます。スライドジェネレーターや LLM ツールなど、他のアプリケーションから画像生成・編集・説明機能を呼び出せます。
 
 ```bash
-imgen preset save myhd -s 1536x1024 -q high -f png
-imgen preset list
-imgen preset delete myhd
+npm install @nogataka/imgen
 ```
 
-プリセットは `~/.imgen/presets.json` に保存されます。
+### 認証設定
 
-## ログ
+`getAzureConfig()` は以下の優先順位で設定を解決します：
 
-操作ログは `~/.imgen/logs/` に JSON Lines 形式で保存されます。
+1. **環境変数**（最優先）
+2. **`.env` ファイル**（`cwd/.env` → `~/.imgen/.env`）
+
+アプリ側で `.env` ファイルを読み込んでいる場合（dotenv, Next.js 等）、そこに追記するだけで動きます：
 
 ```bash
-imgen log                    # 直近 20 件
-imgen log -n 50 -l debug    # 直近 50 件（DEBUG 以上）
-imgen log -l error           # エラーのみ
+# アプリ側の .env
+AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com
+AZURE_OPENAI_API_KEY=your-api-key
+AZURE_OPENAI_DEPLOYMENT_NAME=gpt-5.1
+AZURE_OPENAI_DEPLOYMENT_NAME_IMAGE=gpt-image-1.5
 ```
 
-## 設定の優先順位
+### 使用例
+
+```typescript
+import {
+  AzureImageClient,
+  AzureChatClient,
+  getAzureConfig,
+  saveFileWithUniqueNameIfExists,
+} from "@nogataka/imgen/sdk";
+
+// Azure OpenAI 設定を取得（環境変数 or .envファイル）
+const config = await getAzureConfig();
+
+// 画像生成
+const imageClient = new AzureImageClient(config);
+const imageBytes = await imageClient.generateImage("夕日の海辺", {
+  size: "1536x1024",
+  quality: "high",
+});
+const savedPath = await saveFileWithUniqueNameIfExists("sunset.png", imageBytes);
+
+// プロンプト拡張・画像説明
+const chatClient = new AzureChatClient(config);
+const prompt = await chatClient.generatePrompt("可愛い猫のマスコット");
+const fileName = await chatClient.generateFileName("可愛い猫のマスコット");
+
+// 画像編集
+import * as fs from "node:fs/promises";
+const photo = Buffer.from(await fs.readFile("photo.jpg"));
+const edited = await imageClient.editImage(photo, "背景を青空に変更");
+
+// 画像説明
+import { readImageFile } from "@nogataka/imgen/sdk";
+const imgData = await readImageFile("screenshot.png");
+const explanation = await chatClient.generateExplanation(imgData, "ja");
+```
 
-1. CLI オプション（最優先）
-2. プリセット値（`-p` 指定時）
-3. 設定ファイル（`~/.imgen/config.json`）
-4. デフォルト値
+### LLM ツールとしての組み込み例
+
+```typescript
+import { AzureImageClient, AzureChatClient, getAzureConfig, saveFileWithUniqueNameIfExists } from "@nogataka/imgen/sdk";
+
+const generateImageTool = {
+  name: "generate_image",
+  description: "テキストから画像を生成する",
+  execute: async ({ theme }: { theme: string }) => {
+    const config = await getAzureConfig();
+    const chat = new AzureChatClient(config);
+    const image = new AzureImageClient(config);
+
+    const prompt = await chat.generatePrompt(theme);
+    const fileName = await chat.generateFileName(theme);
+    const bytes = await image.generateImage(prompt, { size: "1024x1024", quality: "high" });
+    const path = await saveFileWithUniqueNameIfExists(`${fileName}.png`, bytes);
+    return { path };
+  },
+};
+```
 
 ## 開発
 

package/dist/chunk-UZXMJEHL.js

@@ -0,0 +1,371 @@
+// src/utils/config.ts
+import * as fs from "fs/promises";
+import * as os from "os";
+import * as path from "path";
+function parseEnvFile(content) {
+  const result = {};
+  for (const line of content.split("\n")) {
+    const trimmed = line.trim();
+    if (!trimmed || trimmed.startsWith("#")) continue;
+    const eqIndex = trimmed.indexOf("=");
+    if (eqIndex === -1) continue;
+    const key = trimmed.slice(0, eqIndex).trim();
+    let value = trimmed.slice(eqIndex + 1).trim();
+    if (value.startsWith('"') && value.endsWith('"') || value.startsWith("'") && value.endsWith("'")) {
+      value = value.slice(1, -1);
+    }
+    result[key] = value;
+  }
+  return result;
+}
+async function loadEnvVars() {
+  const candidates = [path.join(process.cwd(), ".env"), path.join(os.homedir(), ".imgen", ".env")];
+  for (const envPath of candidates) {
+    try {
+      const content = await fs.readFile(envPath, "utf-8");
+      return parseEnvFile(content);
+    } catch {
+    }
+  }
+  return {};
+}
+async function getAzureConfig() {
+  const envVars = await loadEnvVars();
+  const endpoint = process.env.AZURE_OPENAI_ENDPOINT || envVars.AZURE_OPENAI_ENDPOINT;
+  const apiKey = process.env.AZURE_OPENAI_API_KEY || envVars.AZURE_OPENAI_API_KEY;
+  const deploymentName = process.env.AZURE_OPENAI_DEPLOYMENT_NAME || envVars.AZURE_OPENAI_DEPLOYMENT_NAME;
+  const imageDeploymentName = process.env.AZURE_OPENAI_DEPLOYMENT_NAME_IMAGE || envVars.AZURE_OPENAI_DEPLOYMENT_NAME_IMAGE;
+  const apiVersion = process.env.AZURE_OPENAI_API_VERSION || envVars.AZURE_OPENAI_API_VERSION || "2024-02-15-preview";
+  const imageApiVersion = process.env.AZURE_OPENAI_IMAGE_API_VERSION || envVars.AZURE_OPENAI_IMAGE_API_VERSION || "2025-04-01-preview";
+  if (!endpoint || !apiKey || !deploymentName || !imageDeploymentName) {
+    throw new Error(
+      "Azure OpenAI \u306E\u8A2D\u5B9A\u304C\u898B\u3064\u304B\u308A\u307E\u305B\u3093\u3002\u74B0\u5883\u5909\u6570\u307E\u305F\u306F .env \u30D5\u30A1\u30A4\u30EB\u3067\u8A2D\u5B9A\u3057\u3066\u304F\u3060\u3055\u3044\u3002"
+    );
+  }
+  return { endpoint, apiKey, deploymentName, imageDeploymentName, apiVersion, imageApiVersion };
+}
+
+// src/utils/file.ts
+import * as fs2 from "fs/promises";
+async function generateUniqueFilePath(outputPath, maxRetries = 3) {
+  let finalPath = outputPath;
+  let retryCount = 0;
+  while (retryCount < maxRetries) {
+    try {
+      await fs2.stat(finalPath);
+      const baseName = finalPath.slice(0, finalPath.lastIndexOf("."));
+      const ext = finalPath.slice(finalPath.lastIndexOf("."));
+      const rand = Math.floor(Math.random() * 1e4).toString().padStart(4, "0");
+      finalPath = `${baseName}-${rand}${ext}`;
+      retryCount++;
+    } catch (error) {
+      if (error.code === "ENOENT") {
+        return finalPath;
+      }
+      throw error;
+    }
+  }
+  throw new Error(
+    `\u30D5\u30A1\u30A4\u30EB\u540D\u306E\u751F\u6210\u306B\u5931\u6557\u3057\u307E\u3057\u305F\u3002${maxRetries}\u56DE\u8A66\u884C\u3057\u307E\u3057\u305F\u304C\u3001\u3059\u3079\u3066\u65E2\u5B58\u306E\u30D5\u30A1\u30A4\u30EB\u540D\u3068\u885D\u7A81\u3057\u3066\u3044\u307E\u3059\u3002`
+  );
+}
+async function saveFileWithUniqueNameIfExists(outputPath, data, maxRetries = 3) {
+  const finalPath = await generateUniqueFilePath(outputPath, maxRetries);
+  await fs2.writeFile(finalPath, data);
+  return finalPath;
+}
+async function loadContextFile(contextPath) {
+  if (!contextPath) return "";
+  try {
+    return await fs2.readFile(contextPath, "utf-8");
+  } catch (error) {
+    if (error.code === "ENOENT") {
+      throw new Error(`\u30B3\u30F3\u30C6\u30AD\u30B9\u30C8\u30D5\u30A1\u30A4\u30EB\u304C\u898B\u3064\u304B\u308A\u307E\u305B\u3093: ${contextPath}`);
+    }
+    if (error instanceof Error) {
+      throw new Error(`\u30B3\u30F3\u30C6\u30AD\u30B9\u30C8\u30D5\u30A1\u30A4\u30EB\u306E\u8AAD\u307F\u8FBC\u307F\u306B\u5931\u6557\u3057\u307E\u3057\u305F: ${error.message}`);
+    }
+    throw new Error(`\u30B3\u30F3\u30C6\u30AD\u30B9\u30C8\u30D5\u30A1\u30A4\u30EB\u306E\u8AAD\u307F\u8FBC\u307F\u306B\u5931\u6557\u3057\u307E\u3057\u305F: ${String(error)}`);
+  }
+}
+async function fileExists(filePath) {
+  try {
+    await fs2.access(filePath);
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+// src/utils/azure-chat.ts
+import { AzureOpenAI } from "openai";
+var AzureChatClient = class {
+  client;
+  deploymentName;
+  constructor(config) {
+    this.client = new AzureOpenAI({
+      endpoint: config.endpoint,
+      apiKey: config.apiKey,
+      apiVersion: config.apiVersion,
+      deployment: config.deploymentName
+    });
+    this.deploymentName = config.deploymentName;
+  }
+  /**
+   * Generates a detailed image-generation prompt from a short theme description.
+   * Optionally accepts additional context to guide prompt generation.
+   */
+  async generatePrompt(theme, context = "") {
+    if (!theme) throw new Error("\u30C6\u30FC\u30DE\u304C\u7A7A\u3067\u3059");
+    const prompt = `
+Generate a detailed image generation prompt based on the following information.
+
+Theme: ${theme}
+${context ? `Context:
+${context}
+` : ""}
+Please generate a prompt that meets the following criteria:
+1. Include specific and detailed descriptions
+2. Clearly specify the image style and atmosphere
+3. Include all necessary elements
+4. Output in English
+5. Focus on visual elements and composition
+6. Include lighting and color descriptions
+7. Specify the mood and emotional tone
+8. Limit the output to approximately 1500 characters
+
+Prompt:
+`;
+    try {
+      const response = await this.client.chat.completions.create({
+        model: this.deploymentName,
+        messages: [{ role: "user", content: prompt }]
+      });
+      return response.choices[0]?.message?.content ?? "";
+    } catch {
+      throw new Error("\u30D7\u30ED\u30F3\u30D7\u30C8\u306E\u751F\u6210\u306B\u5931\u6557\u3057\u307E\u3057\u305F");
+    }
+  }
+  /**
+   * Generates a sanitized file name (lowercase alphanumeric + hyphens only) from a theme.
+   */
+  async generateFileName(theme, maxLength = 40) {
+    if (!theme) throw new Error("\u30C6\u30FC\u30DE\u304C\u7A7A\u3067\u3059");
+    try {
+      const response = await this.client.chat.completions.create({
+        model: this.deploymentName,
+        messages: [
+          {
+            role: "user",
+            content: `\u4EE5\u4E0B\u306E\u30C6\u30FC\u30DE\u304B\u3089\u753B\u50CF\u306E\u30D5\u30A1\u30A4\u30EB\u540D\u3092\u751F\u6210\u3057\u3066\u304F\u3060\u3055\u3044\u3002\u82F1\u5C0F\u6587\u5B57\u3068\u30CF\u30A4\u30D5\u30F3\u306E\u307F\u3001${maxLength}\u6587\u5B57\u4EE5\u5185\u3002\u62E1\u5F35\u5B50\u306A\u3057\u3002
+
+\u30C6\u30FC\u30DE: ${theme}
+
+\u30D5\u30A1\u30A4\u30EB\u540D:`
+          }
+        ]
+      });
+      let fileName = (response.choices[0]?.message?.content ?? "").trim();
+      fileName = fileName.toLowerCase().replace(/[^a-z0-9-]/g, "").replace(/-+/g, "-").replace(/^-|-$/g, "");
+      if (fileName.length > maxLength) fileName = fileName.substring(0, maxLength);
+      return fileName || "image";
+    } catch {
+      throw new Error("\u30D5\u30A1\u30A4\u30EB\u540D\u306E\u751F\u6210\u306B\u5931\u6557\u3057\u307E\u3057\u305F");
+    }
+  }
+  /**
+   * Generates a detailed explanation of an image using multimodal (vision) input.
+   * The explanation language is controlled by the `lang` parameter.
+   */
+  async generateExplanation(imageData, lang = "ja", context) {
+    try {
+      const response = await this.client.chat.completions.create({
+        model: this.deploymentName,
+        messages: [
+          {
+            role: "user",
+            content: [
+              {
+                type: "image_url",
+                image_url: { url: `data:${imageData.mimeType};base64,${imageData.data}` }
+              },
+              {
+                type: "text",
+                text: `\u3053\u306E\u753B\u50CF\u306B\u3064\u3044\u3066\u3001${lang}\u3067\u8A73\u7D30\u306A\u8AAC\u660E\u3092\u751F\u6210\u3057\u3066\u304F\u3060\u3055\u3044\u3002${context ? `
+
+\u30B3\u30F3\u30C6\u30AD\u30B9\u30C8\u60C5\u5831:
+${context}` : ""}`
+              }
+            ]
+          }
+        ]
+      });
+      return response.choices[0]?.message?.content ?? "";
+    } catch {
+      throw new Error("\u753B\u50CF\u306E\u8AAC\u660E\u751F\u6210\u306B\u5931\u6557\u3057\u307E\u3057\u305F");
+    }
+  }
+};
+
+// src/utils/azure-image.ts
+import { AzureOpenAI as AzureOpenAI2 } from "openai";
+var AzureImageClient = class {
+  client;
+  config;
+  constructor(config) {
+    this.config = config;
+    this.client = new AzureOpenAI2({
+      endpoint: config.endpoint,
+      apiKey: config.apiKey,
+      apiVersion: config.imageApiVersion,
+      deployment: config.imageDeploymentName
+    });
+  }
+  /**
+   * Generates an image from a text prompt using the Azure OpenAI SDK.
+   * Returns raw image bytes as a Uint8Array.
+   */
+  async generateImage(prompt, options) {
+    const { size = "1024x1024", quality = "high" } = options;
+    try {
+      const response = await this.client.images.generate({
+        model: this.config.imageDeploymentName,
+        prompt,
+        n: 1,
+        size,
+        quality,
+        output_format: "png"
+      });
+      if (!response.data || response.data.length === 0 || !response.data[0].b64_json) {
+        throw new Error("\u753B\u50CF\u30C7\u30FC\u30BF\u304C\u898B\u3064\u304B\u308A\u307E\u305B\u3093");
+      }
+      const b64 = response.data[0].b64_json;
+      const binary = atob(b64);
+      const bytes = new Uint8Array(binary.length);
+      for (let i = 0; i < binary.length; i++) {
+        bytes[i] = binary.charCodeAt(i);
+      }
+      return bytes;
+    } catch (error) {
+      if (error instanceof Error && error.message === "\u753B\u50CF\u30C7\u30FC\u30BF\u304C\u898B\u3064\u304B\u308A\u307E\u305B\u3093") throw error;
+      throw new Error(
+        `\u753B\u50CF\u751F\u6210\u306B\u5931\u6557\u3057\u307E\u3057\u305F: ${error instanceof Error ? error.message : String(error)}`
+      );
+    }
+  }
+  /**
+   * Edits an existing image using the Azure OpenAI REST API.
+   * Uses fetch + FormData because the SDK's image editing support is unreliable.
+   * Returns raw image bytes as a Uint8Array.
+   */
+  async editImage(imageBuffer, prompt, options = {}) {
+    const { size = "1024x1024" } = options;
+    const url = `${this.config.endpoint}/openai/deployments/${this.config.imageDeploymentName}/images/edits?api-version=${this.config.imageApiVersion}`;
+    const blob = new Blob([imageBuffer], { type: "image/png" });
+    const formData = new FormData();
+    formData.append("image", blob, "image.png");
+    formData.append("prompt", prompt);
+    formData.append("size", size);
+    try {
+      const response = await fetch(url, {
+        method: "POST",
+        headers: { "api-key": this.config.apiKey },
+        body: formData
+      });
+      if (!response.ok) {
+        const errorText = await response.text();
+        throw new Error(`Azure API error (${response.status}): ${errorText}`);
+      }
+      const json = await response.json();
+      if (!json.data || json.data.length === 0 || !json.data[0].b64_json) {
+        throw new Error("\u753B\u50CF\u30C7\u30FC\u30BF\u304C\u898B\u3064\u304B\u308A\u307E\u305B\u3093");
+      }
+      const b64 = json.data[0].b64_json;
+      const binary = atob(b64);
+      const bytes = new Uint8Array(binary.length);
+      for (let i = 0; i < binary.length; i++) {
+        bytes[i] = binary.charCodeAt(i);
+      }
+      return bytes;
+    } catch (error) {
+      if (error instanceof Error && error.message === "\u753B\u50CF\u30C7\u30FC\u30BF\u304C\u898B\u3064\u304B\u308A\u307E\u305B\u3093") throw error;
+      throw new Error(
+        `\u753B\u50CF\u7DE8\u96C6\u306B\u5931\u6557\u3057\u307E\u3057\u305F: ${error instanceof Error ? error.message : String(error)}`
+      );
+    }
+  }
+};
+
+// src/lang.ts
+var LANGUAGE_DESCRIPTIONS = {
+  ja: "\u65E5\u672C\u8A9E",
+  en: "\u82F1\u8A9E",
+  zh: "\u4E2D\u56FD\u8A9E",
+  ko: "\u97D3\u56FD\u8A9E",
+  es: "\u30B9\u30DA\u30A4\u30F3\u8A9E",
+  fr: "\u30D5\u30E9\u30F3\u30B9\u8A9E",
+  de: "\u30C9\u30A4\u30C4\u8A9E",
+  it: "\u30A4\u30BF\u30EA\u30A2\u8A9E",
+  ru: "\u30ED\u30B7\u30A2\u8A9E",
+  vi: "\u30D9\u30C8\u30CA\u30E0\u8A9E"
+};
+
+// src/utils/image.ts
+import * as fs3 from "fs/promises";
+async function readImageFile(filePath) {
+  try {
+    const buffer = await fs3.readFile(filePath);
+    return {
+      data: buffer.toString("base64"),
+      mimeType: getMimeType(filePath)
+    };
+  } catch (error) {
+    if (error instanceof Error && error.message.startsWith("\u30B5\u30DD\u30FC\u30C8\u3055\u308C\u3066\u3044\u306A\u3044")) {
+      throw error;
+    }
+    throw new Error(
+      `\u753B\u50CF\u30D5\u30A1\u30A4\u30EB\u306E\u8AAD\u307F\u8FBC\u307F\u306B\u5931\u6557\u3057\u307E\u3057\u305F: ${error instanceof Error ? error.message : String(error)}`
+    );
+  }
+}
+function getMimeType(filePath) {
+  const ext = filePath.toLowerCase().split(".").pop();
+  const map = {
+    jpg: "image/jpeg",
+    jpeg: "image/jpeg",
+    png: "image/png",
+    gif: "image/gif",
+    webp: "image/webp"
+  };
+  if (ext && map[ext]) return map[ext];
+  throw new Error(`\u30B5\u30DD\u30FC\u30C8\u3055\u308C\u3066\u3044\u306A\u3044\u30D5\u30A1\u30A4\u30EB\u5F62\u5F0F\u3067\u3059: .${ext}`);
+}
+
+// src/utils/preset.ts
+var BUILTIN_PRESETS = {
+  "builtin:square": { size: "1024x1024", quality: "high" },
+  "builtin:landscape": { size: "1536x1024", quality: "high" },
+  "builtin:portrait": { size: "1024x1536", quality: "high" },
+  "builtin:draft": { size: "1024x1024", quality: "low" },
+  "builtin:photo": { size: "1536x1024", quality: "high" }
+};
+function getPreset(name) {
+  if (name.startsWith("builtin:") && BUILTIN_PRESETS[name]) {
+    return BUILTIN_PRESETS[name];
+  }
+  return BUILTIN_PRESETS[name] || null;
+}
+
+export {
+  getAzureConfig,
+  saveFileWithUniqueNameIfExists,
+  loadContextFile,
+  fileExists,
+  AzureChatClient,
+  AzureImageClient,
+  LANGUAGE_DESCRIPTIONS,
+  readImageFile,
+  getMimeType,
+  BUILTIN_PRESETS,
+  getPreset
+};