Add PDF translation API, utilities, docs, and config
Introduces core backend and frontend infrastructure for a PDF translation interface. Adds API endpoints for translation, PDF testing, and AI provider testing; implements PDF text extraction, cost tracking, and pricing logic in the lib directory; adds reusable UI components; and provides comprehensive documentation (SDD, environment setup, Claude instructions). Updates Tailwind and global styles, and includes a sample test PDF and configuration files.
This commit is contained in:
@@ -1,30 +1,133 @@
|
||||
import { type NextRequest, NextResponse } from "next/server"
|
||||
import { createOpenAI } from "@ai-sdk/openai"
|
||||
import { openai } from "@ai-sdk/openai"
|
||||
import { generateText } from "ai"
|
||||
import { extractTextFromPDF, generateTranslatedPDF, processImageFile, processPDFWithOCR, ocrLanguageMap, isImageFile, isPDFFile } from "@/lib/pdf-processor"
|
||||
import { calculateCost, estimateTokens, formatTokenCount, MODEL_PRICING } from "@/lib/pricing"
|
||||
import { costTracker } from "@/lib/cost-tracker"
|
||||
|
||||
export async function POST(request: NextRequest) {
|
||||
try {
|
||||
// Select AI provider based on environment variable
|
||||
const aiProvider = process.env.AI_PROVIDER || "deepseek"
|
||||
|
||||
let model
|
||||
let modelName: string
|
||||
|
||||
if (aiProvider === "openai") {
|
||||
// Use OpenAI
|
||||
modelName = process.env.OPENAI_MODEL || "gpt-4o-mini"
|
||||
model = openai(modelName)
|
||||
} else {
|
||||
// Use DeepSeek (default)
|
||||
modelName = process.env.DEEPSEEK_MODEL || "deepseek-chat"
|
||||
const deepseek = createOpenAI({
|
||||
apiKey: process.env.DEEPSEEK_API_KEY,
|
||||
baseURL: process.env.DEEPSEEK_BASE_URL || "https://api.deepseek.com",
|
||||
})
|
||||
model = deepseek(modelName)
|
||||
}
|
||||
|
||||
const formData = await request.formData()
|
||||
const file = formData.get("file") as File
|
||||
const targetLanguage = formData.get("targetLanguage") as string
|
||||
const sourceLanguage = formData.get("sourceLanguage") as string
|
||||
const returnPDF = formData.get("returnPDF") === "true"
|
||||
|
||||
if (!file || !targetLanguage) {
|
||||
return NextResponse.json({ error: "缺少必要參數" }, { status: 400 })
|
||||
}
|
||||
|
||||
// Extract text from PDF
|
||||
// Validate file type
|
||||
if (!isPDFFile(file.type) && !isImageFile(file.type)) {
|
||||
return NextResponse.json({ error: "請上傳 PDF 檔案或圖片檔案" }, { status: 400 })
|
||||
}
|
||||
|
||||
// Check file size (10MB limit)
|
||||
const maxSize = parseInt(process.env.MAX_FILE_SIZE || "10485760")
|
||||
if (file.size > maxSize) {
|
||||
return NextResponse.json({ error: "檔案太大，請上傳小於 10MB 的檔案" }, { status: 413 })
|
||||
}
|
||||
|
||||
// Process file based on type
|
||||
const arrayBuffer = await file.arrayBuffer()
|
||||
const buffer = Buffer.from(arrayBuffer)
|
||||
|
||||
let extractedText = ""
|
||||
let metadata: any = {}
|
||||
|
||||
try {
|
||||
if (isImageFile(file.type)) {
|
||||
// Image file - OCR功能已停用
|
||||
console.log("Image file detected - OCR功能已停用")
|
||||
extractedText = "目前僅支援包含文字的 PDF 文件，不支援圖片檔案。"
|
||||
metadata = { title: file.name, type: 'image' }
|
||||
} else if (isPDFFile(file.type)) {
|
||||
// Process PDF file
|
||||
console.log("Processing PDF file...")
|
||||
const result = await extractTextFromPDF(buffer)
|
||||
metadata = result.metadata
|
||||
|
||||
if (result.isScanned) {
|
||||
// PDF is scanned or text extraction failed
|
||||
console.log("Detected scanned PDF or text extraction failed")
|
||||
const message = result.metadata?.message || "此 PDF 為掃描檔案，目前僅支援包含文字的 PDF 文件。"
|
||||
|
||||
return NextResponse.json({
|
||||
error: `${message}
|
||||
|
||||
// For demo purposes, we'll simulate PDF text extraction
|
||||
// In production, you'd use a library like pdf-parse
|
||||
const pdfText = `這是從PDF提取的示例文本。在實際應用中，這裡會是真實的PDF內容。
|
||||
📋 可能的原因：
|
||||
• PDF 是掃描的圖片檔案
|
||||
• PDF 文件已加密或受保護
|
||||
• PDF 內容格式特殊，無法提取文字
|
||||
• PDF 文件損壞
|
||||
|
||||
這個應用展示了如何使用AI來翻譯文檔內容。您可以上傳任何PDF文件，選擇目標語言，然後獲得翻譯結果。
|
||||
💡 建議：
|
||||
• 嘗試其他包含純文字的 PDF 文件
|
||||
• 確認 PDF 可以在其他軟體中複製文字
|
||||
• 如果是掃描檔案，建議轉換為圖片格式`,
|
||||
details: {
|
||||
pageCount: result.pageCount,
|
||||
textLength: result.metadata?.extractedTextLength || 0,
|
||||
hasTextContent: result.metadata?.hasTextContent || false
|
||||
}
|
||||
}, { status: 400 })
|
||||
} else {
|
||||
// PDF has extractable text, use it directly
|
||||
console.log("PDF contains extractable text, using direct extraction")
|
||||
extractedText = result.text
|
||||
}
|
||||
}
|
||||
|
||||
if (!extractedText || extractedText.trim().length === 0) {
|
||||
extractedText = "無法從檔案擷取文字內容。請確認檔案包含可讀取的文字或清晰的圖像。"
|
||||
}
|
||||
} catch (error) {
|
||||
console.error("File processing error:", error)
|
||||
|
||||
// Provide helpful error message for PDF conversion issues
|
||||
if (error instanceof Error && error.message.includes('PDF 轉圖片失敗')) {
|
||||
return NextResponse.json({
|
||||
error: `📄 掃描 PDF 需要額外工具支援
|
||||
|
||||
主要功能包括：
|
||||
- 支持多種語言翻譯
|
||||
- 清爽的用戶介面
|
||||
- 簡單易用的操作流程`
|
||||
🎯 建議解決方案：
|
||||
1. 💡 立即可用：將 PDF 轉換為圖片格式（JPG/PNG）後上傳
|
||||
- 使用 PDF 閱讀器截圖
|
||||
- 或使用線上 PDF 轉圖片工具
|
||||
|
||||
2. 🔧 安裝系統工具：
|
||||
• Windows: 下載安裝 ImageMagick (https://imagemagick.org/script/download.php#windows)
|
||||
• Mac: brew install imagemagick
|
||||
• Linux: apt-get install imagemagick
|
||||
|
||||
📸 提示：圖片格式的 OCR 識別效果通常比掃描 PDF 更好！`,
|
||||
suggestion: "convert_to_image",
|
||||
downloadLink: "https://imagemagick.org/script/download.php#windows"
|
||||
}, { status: 400 })
|
||||
}
|
||||
|
||||
extractedText = `檔案處理過程中發生錯誤：${error instanceof Error ? error.message : '未知錯誤'}`
|
||||
}
|
||||
|
||||
// Get language name for better translation context
|
||||
const languageNames: Record<string, string> = {
|
||||
@@ -40,23 +143,137 @@ export async function POST(request: NextRequest) {
|
||||
pt: "Português",
|
||||
ru: "Русский",
|
||||
ar: "العربية",
|
||||
hi: "हिन्दी",
|
||||
th: "ไทย",
|
||||
vi: "Tiếng Việt",
|
||||
}
|
||||
|
||||
const targetLanguageName = languageNames[targetLanguage] || targetLanguage
|
||||
const sourceLanguageName = languageNames[sourceLanguage] || "自動偵測"
|
||||
|
||||
// Translate using AI SDK
|
||||
const { text: translatedText } = await generateText({
|
||||
model: "openai/gpt-4o-mini",
|
||||
prompt: `請將以下文本翻譯成${targetLanguageName}。保持原文的格式和結構，只翻譯內容：
|
||||
// Prepare translation prompt
|
||||
const prompt = `You are a professional translator. Translate the following text from ${sourceLanguageName} to ${targetLanguageName}.
|
||||
Keep the original format and structure. Only translate the content, preserving line breaks and paragraphs.
|
||||
If the text appears to be an error message or system message, translate it appropriately.
|
||||
|
||||
${pdfText}`,
|
||||
Text to translate:
|
||||
|
||||
${extractedText}`
|
||||
|
||||
// Estimate input tokens
|
||||
const estimatedInputTokens = estimateTokens(prompt)
|
||||
|
||||
let translatedText: string
|
||||
let usage: any = null
|
||||
|
||||
try {
|
||||
// Try to translate using selected AI provider
|
||||
const result = await generateText({
|
||||
model: model,
|
||||
prompt: prompt,
|
||||
temperature: 0.3, // Lower temperature for more accurate translation
|
||||
maxTokens: 4000,
|
||||
})
|
||||
translatedText = result.text
|
||||
usage = result.usage
|
||||
} catch (error) {
|
||||
console.error("AI API Error:", error)
|
||||
// Fallback to a simple mock translation for demo purposes
|
||||
translatedText = `[模擬翻譯結果]\n\n原文內容: ${extractedText}\n\n注意: 這是模擬翻譯結果，因為 AI API 連接失敗。請檢查 API 金鑰配置。\n\n目標語言: ${targetLanguageName}\n來源語言: ${sourceLanguageName}\n\n實際應用中，這裡會顯示真正的 AI 翻譯結果。`
|
||||
}
|
||||
|
||||
// Calculate token usage and cost
|
||||
const tokenUsage = {
|
||||
promptTokens: usage?.promptTokens || estimatedInputTokens,
|
||||
completionTokens: usage?.completionTokens || estimateTokens(translatedText),
|
||||
totalTokens: (usage?.promptTokens || estimatedInputTokens) + (usage?.completionTokens || estimateTokens(translatedText))
|
||||
}
|
||||
|
||||
const costCalculation = calculateCost(modelName, tokenUsage)
|
||||
const modelDisplayName = MODEL_PRICING[modelName as keyof typeof MODEL_PRICING]?.displayName || modelName
|
||||
|
||||
// Track cost in accumulator (client-side will handle storage)
|
||||
const costSession = {
|
||||
model: modelName,
|
||||
provider: aiProvider,
|
||||
tokenUsage: {
|
||||
promptTokens: tokenUsage.promptTokens,
|
||||
completionTokens: tokenUsage.completionTokens,
|
||||
totalTokens: tokenUsage.totalTokens
|
||||
},
|
||||
cost: {
|
||||
inputCost: costCalculation.inputCost,
|
||||
outputCost: costCalculation.outputCost,
|
||||
totalCost: costCalculation.totalCost,
|
||||
currency: costCalculation.currency
|
||||
}
|
||||
}
|
||||
|
||||
// Generate translated PDF if requested
|
||||
let pdfBase64 = ""
|
||||
if (returnPDF) {
|
||||
try {
|
||||
const pdfBytes = await generateTranslatedPDF(translatedText, metadata, targetLanguage)
|
||||
pdfBase64 = Buffer.from(pdfBytes).toString('base64')
|
||||
} catch (error) {
|
||||
console.error("PDF generation error:", error)
|
||||
// Continue without PDF generation
|
||||
}
|
||||
}
|
||||
|
||||
return NextResponse.json({
|
||||
translatedText,
|
||||
pdfBase64: pdfBase64,
|
||||
hasPDF: pdfBase64.length > 0,
|
||||
tokenUsage: {
|
||||
promptTokens: tokenUsage.promptTokens,
|
||||
completionTokens: tokenUsage.completionTokens,
|
||||
totalTokens: tokenUsage.totalTokens,
|
||||
formattedCounts: {
|
||||
prompt: formatTokenCount(tokenUsage.promptTokens),
|
||||
completion: formatTokenCount(tokenUsage.completionTokens),
|
||||
total: formatTokenCount(tokenUsage.totalTokens)
|
||||
}
|
||||
},
|
||||
cost: {
|
||||
inputCost: costCalculation.inputCost,
|
||||
outputCost: costCalculation.outputCost,
|
||||
totalCost: costCalculation.totalCost,
|
||||
formattedCost: costCalculation.formattedCost,
|
||||
currency: costCalculation.currency
|
||||
},
|
||||
model: {
|
||||
name: modelName,
|
||||
displayName: modelDisplayName,
|
||||
provider: aiProvider
|
||||
},
|
||||
costSession: costSession
|
||||
})
|
||||
|
||||
return NextResponse.json({ translatedText })
|
||||
} catch (error) {
|
||||
console.error("翻譯錯誤:", error)
|
||||
return NextResponse.json({ error: "翻譯過程中發生錯誤" }, { status: 500 })
|
||||
|
||||
// Check for specific error types
|
||||
if (error instanceof Error) {
|
||||
if (error.message.includes("API key")) {
|
||||
return NextResponse.json({ error: "API 金鑰配置錯誤，請聯繫管理員" }, { status: 500 })
|
||||
}
|
||||
if (error.message.includes("rate limit")) {
|
||||
return NextResponse.json({ error: "API 請求過於頻繁，請稍後再試" }, { status: 429 })
|
||||
}
|
||||
}
|
||||
|
||||
return NextResponse.json({ error: "翻譯過程中發生錯誤，請稍後再試" }, { status: 500 })
|
||||
}
|
||||
}
|
||||
|
||||
// OPTIONS method for CORS
|
||||
export async function OPTIONS(request: NextRequest) {
|
||||
return new NextResponse(null, {
|
||||
status: 200,
|
||||
headers: {
|
||||
'Access-Control-Allow-Origin': '*',
|
||||
'Access-Control-Allow-Methods': 'POST, OPTIONS',
|
||||
'Access-Control-Allow-Headers': 'Content-Type',
|
||||
},
|
||||
})
|
||||
}
|
||||
Reference in New Issue
Block a user