pi-ocr 1.1.3 → 1.2.1

package/extensions/index.ts

@@ -43,7 +43,7 @@ import { homedir } from "node:os";
 
 import type { Backend, Task, OcrConfig } from "./types";
 import { TASKS, BACKENDS } from "./types";
-import { isImage, isPdf, getPdfPageCount, ollamaOcr, ollamaCheckModel, ollamaPullModel } from "./ollama";
+import { isImage, isPdf, ollamaOcr, ollamaCheckModel, ollamaPullModel } from "./ollama";
 import { mineruOcr } from "./mineru";
 import { tesseractOcr } from "./tesseract";
 import { pix2textOcr } from "./pix2text";
@@ -55,7 +55,16 @@ const SETTINGS_PATH = join(homedir(), ".pi", "agent", "settings.json");
 function loadOcrConfig(): Partial<OcrConfig> {
   try {
     if (!existsSync(SETTINGS_PATH)) return {};
-    return (JSON.parse(readFileSync(SETTINGS_PATH, "utf8")) as any).minimodelOcr || {};
+    const settings = JSON.parse(readFileSync(SETTINGS_PATH, "utf8"));
+    // Migrate from old key (pi-minimodel-ocr) to new key (pi-ocr)
+    const old = (settings as any).minimodelOcr;
+    const current = (settings as any).piOcr;
+    if (old && !current) {
+      (settings as any).piOcr = old;
+      delete (settings as any).minimodelOcr;
+      writeFileSync(SETTINGS_PATH, JSON.stringify(settings, null, 2) + "\n", "utf8");
+    }
+    return (settings as any).piOcr || {};
   } catch { return {}; }
 }
 
@@ -66,7 +75,7 @@ function saveOcrConfig(updates: Partial<OcrConfig>) {
     const settings = existsSync(SETTINGS_PATH)
       ? JSON.parse(readFileSync(SETTINGS_PATH, "utf8"))
       : {};
-    settings.minimodelOcr = { ...(settings.minimodelOcr || {}), ...updates };
+    settings.piOcr = { ...(settings.piOcr || {}), ...updates };
     writeFileSync(SETTINGS_PATH, JSON.stringify(settings, null, 2) + "\n", "utf8");
   } catch { /* best effort */ }
 }

package/extensions/mineru.ts

@@ -186,6 +186,19 @@ async function mineruProcessFile(
   return markdown;
 }
 
+// ── Image → PDF wrapper (so MinerU applies language=\"ch\" pipeline) ──────────
+
+const IMG2PDF_SCRIPT = `
+import sys
+from PIL import Image
+img = Image.open(sys.argv[1])
+img.save(sys.argv[2], "PDF")
+`;
+
+async function wrapImageAsPdf(imagePath: string, pdfPath: string): Promise<void> {
+  await execPy(IMG2PDF_SCRIPT, [imagePath, pdfPath]);
+}
+
 // ── Public API ───────────────────────────────────────────────────────────────
 
 export async function mineruOcr(
@@ -195,13 +208,23 @@ export async function mineruOcr(
   const ext = extname(filePath).toLowerCase();
   const fileName = basename(filePath);
 
-  // For images (non-PDF): process as a single individual request
+  // For images (non-PDF): wrap in PDF so MinerU applies language=\"ch\" pipeline
   if (ext !== ".pdf") {
     if (![".png", ".jpg", ".jpeg", ".gif", ".webp", ".bmp", ".tiff", ".tif"].includes(ext)) {
       throw new Error(`MinerU does not support this file type: ${ext}. Use PDF, PNG, JPG, Docx, PPTx, or Xlsx.`);
     }
-    const markdown = await mineruProcessFile(filePath, fileName, "[1/1]", onProgress);
-    return { text: markdown, details: { backend: "mineru", fileName, pages: 1 } };
+
+    // Wrap image as 1-page PDF so language/chinese OCR works
+    onProgress("[1/1] converting image to PDF…");
+    const pdfPath = join(tmpdir(), `pi-mineru-img-${Date.now()}.pdf`);
+    try {
+      await wrapImageAsPdf(filePath, pdfPath);
+      const pdfName = fileName.replace(/\.[^.]+$/, "") + ".pdf";
+      const markdown = await mineruProcessFile(pdfPath, pdfName, "[1/1]", onProgress);
+      return { text: markdown, details: { backend: "mineru", fileName, pages: 1 } };
+    } finally {
+      try { unlinkSync(pdfPath); } catch { /* cleanup */ }
+    }
   }
 
   // ── PDF handling ──

package/extensions/ollama.ts

@@ -6,7 +6,7 @@
  */
 
 import { readFileSync, existsSync, mkdtempSync, readdirSync, unlinkSync, rmdirSync } from "node:fs";
-import { basename, extname, join } from "node:path";
+import { extname, join } from "node:path";
 import { tmpdir } from "node:os";
 import { spawn } from "node:child_process";
 import type { Task, OcrResult, OcrProgressCallback } from "./types";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-ocr",
-  "version": "1.1.3",
+  "version": "1.2.1",
   "description": "Pi extension: Zero-setup multi-backend OCR — MinerU (free cloud), Ollama (local GPU, LaTeX formulas), Pix2Text (local Python). Extract text, formulas, and tables from images and PDFs. Default: zero config, works out of the box.",
   "keywords": [
     "pi-package",