fix(ocr): fix PaddleOCR crashes, add multi-image and auto-detect language

- Pin PaddlePaddle to 3.0.0 on ARM64 to fix segfault in PIR inference
  engine (3.1+ crashes on aarch64 Debian Bookworm)
- Fix text extraction for PaddleOCR 3.4.x result format (rec_texts)
- Add Node.js-level fallback chain (best -> balanced -> fast) when
  Python subprocess crashes
- Add multi-image OCR: processes all uploaded files sequentially with
  per-file progress and filename headers in combined output
- Convert input images to PNG via Sharp before OCR so HEIC, AVIF, WebP,
  TIFF all work transparently
- Implement real auto-detect language using Tesseract multi-lang script
  detection (analyzes Unicode ranges for Hangul, CJK, Kana, Latin)
- Default enhance to off (hurts clean digital images)

packages/ai/src/ocr.ts

@@ -1,5 +1,6 @@
 import { writeFile } from "node:fs/promises";
 import { join } from "node:path";
+import sharp from "sharp";
 import { type ProgressCallback, runPythonWithProgress } from "./bridge.js";
 
 export type OcrQuality = "fast" | "balanced" | "best";
@@ -24,7 +25,11 @@ export async function extractText(
 ): Promise<OcrResult> {
   const inputPath = join(outputDir, "input_ocr.png");
 
-  await writeFile(inputPath, inputBuffer);
+  // Convert any input format (HEIC, AVIF, WebP, TIFF, etc.) to PNG
+  // so Tesseract and PaddleOCR can read it reliably.
+  const pngBuffer = await sharp(inputBuffer).png().toBuffer();
+  await writeFile(inputPath, pngBuffer);
+
   const { stdout } = await runPythonWithProgress("ocr.py", [inputPath, JSON.stringify(options)], {
     onProgress,
     timeout: 600_000, // 10 min timeout for VLM on CPU