fix(sync,rag): keep restored rows, reconcile after tombstone pruning, reject partial knowledge docs, surface RAG indexing failures

apps/desktop/src/main/services/rag/pdf-text.ts

@@ -0,0 +1,94 @@
+// src/main/services/rag/pdf-text.ts
+// PDF 바이너리에서 텍스트를 뽑는다 — 순수 JS + zlib, 외부 의존 없음(RAGService에서 옮김, 동작 동일).
+// FlateDecode 스트림을 풀고 BT...ET 블록 안의 Tj/TJ/' 연산자 문자열을 모은다.
+
+import { inflateSync } from 'zlib'
+
+const STREAM_MARKER_CRLF = 'stream\r\n'
+const STREAM_MARKER_LF = 'stream\n'
+const END_MARKER = 'endstream'
+
+/** 텍스트 블록(BT...ET) 하나에서 문자열 조각을 모은다. */
+export function extractTextOperators(content: string, textParts: string[]): void {
+  const btRegex = /BT([\s\S]*?)ET/g
+  let btMatch: RegExpExecArray | null = null
+  while ((btMatch = btRegex.exec(content)) !== null) {
+    const block = btMatch[1]
+
+    // Tj: (text) Tj
+    const tjRegex = /\(([^)]*)\)\s*Tj/g
+    let tjMatch: RegExpExecArray | null = null
+    while ((tjMatch = tjRegex.exec(block)) !== null) {
+      if (tjMatch[1].trim()) textParts.push(tjMatch[1])
+    }
+
+    // TJ: [(text) num (text)] TJ
+    const tjArrayRegex = /\[(.*?)\]\s*TJ/g
+    let tjArrayMatch: RegExpExecArray | null = null
+    while ((tjArrayMatch = tjArrayRegex.exec(block)) !== null) {
+      const items = tjArrayMatch[1]
+      const itemRegex = /\(([^)]*)\)/g
+      let itemMatch: RegExpExecArray | null = null
+      while ((itemMatch = itemRegex.exec(items)) !== null) {
+        if (itemMatch[1].trim()) textParts.push(itemMatch[1])
+      }
+    }
+
+    // ' 연산자: (text) '
+    const quoteRegex = /\(([^)]*)\)\s*'/g
+    let quoteMatch: RegExpExecArray | null = null
+    while ((quoteMatch = quoteRegex.exec(block)) !== null) {
+      if (quoteMatch[1].trim()) textParts.push(quoteMatch[1])
+    }
+  }
+}
+
+/** PDF 문자열 이스케이프를 풀고 표시할 수 없는 문자를 공백으로 정리한다. */
+export function normalizePdfText(parts: readonly string[]): string {
+  return parts
+    .join(' ')
+    .replace(/\\n/g, '\n')
+    .replace(/\\r/g, '\r')
+    .replace(/\\t/g, '\t')
+    .replace(/\\\(/g, '(')
+    .replace(/\\\)/g, ')')
+    .replace(/\\\\/g, '\\')
+    .replace(/[^\x20-\x7E\u00A0-\u00FF\u3000-\u9FFF\uAC00-\uD7AF\n]/g, ' ')
+    .replace(/\s+/g, ' ')
+    .trim()
+}
+
+/** PDF 바이너리에서 텍스트 추출. 읽을 텍스트가 없으면 빈 문자열. */
+export function extractPdfText(buffer: Buffer): string {
+  const raw = buffer.toString('binary')
+  const textParts: string[] = []
+
+  let pos = 0
+  while (pos < raw.length) {
+    let streamStart = raw.indexOf(STREAM_MARKER_CRLF, pos)
+    let offset = STREAM_MARKER_CRLF.length
+    if (streamStart === -1) {
+      streamStart = raw.indexOf(STREAM_MARKER_LF, pos)
+      offset = STREAM_MARKER_LF.length
+    }
+    if (streamStart === -1) break
+
+    const dataStart = streamStart + offset
+    const streamEnd = raw.indexOf(END_MARKER, dataStart)
+    if (streamEnd === -1) break
+
+    const streamData = Buffer.from(raw.slice(dataStart, streamEnd), 'binary')
+    pos = streamEnd + END_MARKER.length
+
+    let decoded: string
+    try {
+      decoded = inflateSync(streamData).toString('binary')
+    } catch {
+      // 압축 안 된 스트림
+      decoded = streamData.toString('binary')
+    }
+    extractTextOperators(decoded, textParts)
+  }
+
+  return normalizePdfText(textParts)
+}