pi-ocr 1.0.2 → 1.1.1

package/README.md

@@ -29,7 +29,8 @@ Switch anytime with `/ocr` (no args).
 |---|---|---|---|
 | ☁️ | **MinerU** (default) | PDFs, tables, general docs | None — works instantly |
 | 🦙 | Ollama | Math formulas → LaTeX, offline | `brew install ollama && ollama pull glm-ocr` |
-| 📐 | Pix2Text | Math + text, CPU-only | `pip install pix2text` |
+| 🔤 | Tesseract | Plain text, ultra-light (~30MB) | `brew install tesseract` |
+| 📐 | Pix2Text | Math + text, CPU Python | `pip install pix2text` |
 
 ---
 
@@ -63,7 +64,20 @@ Switch with `/ocr` → "OCR Backend" → ollama.
 
 ---
 
-## Pix2Text (optional, for offline CPU)
+## Tesseract (optional, for plain text)
+
+Classic OCR engine. Ultra-lightweight (~30MB), CPU-only, fast. System package, zero Python.
+
+```bash
+brew install tesseract              # macOS
+sudo apt install tesseract-ocr      # Linux
+```
+
+Switch with `/ocr` → "OCR Backend" → tesseract.
+
+---
+
+## Pix2Text (optional, for math + text on CPU)
 
 Local Python OCR. Mathpix alternative — handles text + formulas on CPU.
 
@@ -103,6 +117,8 @@ First run downloads ONNX models (~50MB). Switch with `/ocr` → "OCR Backend"
 
 **"python3 not found" (Pix2Text)** → `python3 -m pip install pix2text`
 
+**"tesseract not found"** → `brew install tesseract` (macOS) / `sudo apt install tesseract-ocr` (Linux)
+
 **"pdftoppm not found" (Ollama multi-page)** → `brew install poppler` (macOS) / `sudo apt install poppler-utils` (Linux)
 
 ---

package/extensions/index.ts

@@ -45,6 +45,7 @@ import type { Backend, Task, OcrConfig } from "./types";
 import { TASKS, BACKENDS } from "./types";
 import { isImage, isPdf, getPdfPageCount, ollamaOcr, ollamaCheckModel, ollamaPullModel } from "./ollama";
 import { mineruOcr } from "./mineru";
+import { tesseractOcr } from "./tesseract";
 import { pix2textOcr } from "./pix2text";
 
 // ── Config persistence ───────────────────────────────────────────────────────
@@ -138,7 +139,7 @@ const ocrTool = defineTool({
       throw new Error(`Unsupported file type "${extname(filePath)}". Supported: PNG, JPG, GIF, WEBP, BMP, TIFF, PDF.`);
     }
 
-    const backendLabel = { ollama: "🦙 Ollama", mineru: "☁️ MinerU", pix2text: "📐 Pix2Text" }[config.backend];
+    const backendLabel = { mineru: "☁️ MinerU", ollama: "🦙 Ollama", tesseract: "🔤 Tesseract", pix2text: "📐 Pix2Text" }[config.backend];
     onUpdate?.({ content: [{ type: "text", text: `🔍 OCR ${basename(filePath)} via ${backendLabel} (${resolvedTask})…` }], details: {} });
 
     const onProgress = (msg: string) => onUpdate?.({ content: [{ type: "text", text: msg }], details: {} });
@@ -159,6 +160,9 @@ const ocrTool = defineTool({
           result = await mineruOcr(filePath, resolvedTask, config.mineruSplitPdf, signal, onProgress);
           break;
         }
+        case "tesseract":
+          result = await tesseractOcr(filePath, resolvedTask, signal, onProgress);
+          break;
         case "pix2text":
           result = await pix2textOcr(filePath, resolvedTask, signal, onProgress);
           break;
@@ -175,6 +179,7 @@ const ocrTool = defineTool({
       const msg = e.message || String(e);
       let hint = "";
       if (config.backend === "ollama" && (msg.includes("fetch failed") || msg.includes("ECONNREFUSED"))) hint = "\n\n💡 Is Ollama running? Start: `ollama serve`";
+      else if (config.backend === "tesseract" && msg.includes("not found")) hint = "\n\n💡 Install: `brew install tesseract` (macOS) or `sudo apt install tesseract-ocr` (Linux)";
       else if (config.backend === "pix2text" && msg.includes("python3")) hint = "\n\n💡 Install: `pip install pix2text`";
       else if (config.backend === "mineru" && msg.includes("429")) hint = "\n\n💡 MinerU rate limit. Wait a minute or switch backend with /ocr.";
       else if (config.backend === "mineru" && msg.includes("too large")) hint = "\n\n💡 Compress at https://ilovepdf.com/compress_pdf or switch backend.";
@@ -292,7 +297,9 @@ export default function ocrExtension(pi: ExtensionAPI) {
                     ".\nLarge files? Compress at https://ilovepdf.com/compress_pdf",
                     "info",
                   );
-                } else if (backend === "pix2text") {
+                } else if (backend === "tesseract") {
+                ctx.ui.notify("🔤 Tesseract: `brew install tesseract` (macOS) or `sudo apt install tesseract-ocr` (Linux). ~30MB, CPU-only.", "warning");
+              } else if (backend === "pix2text") {
                   ctx.ui.notify("🐍 Pix2Text: needs `pip install pix2text`", "warning");
                 }
                 break;

package/extensions/tesseract.ts

@@ -0,0 +1,132 @@
+/**
+ * pi-ocr — Tesseract backend
+ *
+ * Uses Tesseract OCR (https://github.com/tesseract-ocr/tesseract) —
+ * the classic open-source OCR engine. Ultra-lightweight (~30MB),
+ * zero Python deps, CPU-only, fast on plain text.
+ *
+ * Prerequisites:
+ *   brew install tesseract        # macOS
+ *   sudo apt install tesseract-ocr # Linux
+ *
+ * For non-English languages, install the corresponding lang pack:
+ *   brew install tesseract-lang   # macOS (all languages)
+ *   sudo apt install tesseract-ocr-chi-sim  # Chinese simplified
+ */
+
+import { mkdtempSync, readdirSync, unlinkSync, rmdirSync } from "node:fs";
+import { basename, join } from "node:path";
+import { tmpdir } from "node:os";
+import { spawn } from "node:child_process";
+import type { Task, OcrResult, OcrProgressCallback } from "./types";
+import { isImage, isPdf, getPdfPageCount } from "./ollama";
+
+// ── Helpers ──────────────────────────────────────────────────────────────────
+
+async function execCapture(cmd: string, args: string[]): Promise<{ stdout: string; stderr: string; code: number }> {
+  return new Promise((resolve) => {
+    const child = spawn(cmd, args, { stdio: ["ignore", "pipe", "pipe"] });
+    const out: Buffer[] = [];
+    const err: Buffer[] = [];
+    child.stdout.on("data", (d) => out.push(d));
+    child.stderr.on("data", (d) => err.push(d));
+    child.on("error", () => resolve({
+      stdout: "", stderr: "tesseract not found. Install: brew install tesseract", code: 1,
+    }));
+    child.on("close", (code) => resolve({
+      stdout: Buffer.concat(out).toString("utf8").trim(),
+      stderr: Buffer.concat(err).toString("utf8").trim(),
+      code: code ?? 1,
+    }));
+  });
+}
+
+function cleanupDir(dir: string) {
+  try {
+    for (const f of readdirSync(dir)) unlinkSync(join(dir, f));
+    rmdirSync(dir);
+  } catch { /* best effort */ }
+}
+
+// ── PDF → image conversion (reuses ollama helpers approach) ──────────────────
+
+async function convertPdfPage(pdfPath: string, pageIndex: number, outPath: string): Promise<void> {
+  if (process.platform === "darwin") {
+    if (pageIndex === 0) {
+      await execCapture("sips", ["-s", "format", "png", pdfPath, "--out", outPath]);
+      return;
+    }
+  }
+  await execCapture("pdftoppm", [
+    "-png", "-r", "200", "-f", String(pageIndex + 1), "-l", String(pageIndex + 1),
+    "-singlefile", pdfPath, outPath.replace(/\.png$/, ""),
+  ]);
+}
+
+// ── Tesseract OCR ────────────────────────────────────────────────────────────
+
+async function tesseractImage(imagePath: string, _task: Task): Promise<string> {
+  const { stdout, stderr, code } = await execCapture("tesseract", [
+    imagePath, "stdout",
+    "-l", "eng+chi_sim",   // English + Chinese simplified
+    "--psm", "3",  // Auto page segmentation
+  ]);
+
+  if (code !== 0) {
+    const msg = stderr || "tesseract failed";
+    if (msg.includes("not found") || msg.includes("ENOENT")) {
+      throw new Error("tesseract not found. Install: brew install tesseract");
+    }
+    throw new Error(msg.slice(0, 500));
+  }
+
+  return stdout;
+}
+
+// ── Public API ───────────────────────────────────────────────────────────────
+
+export async function tesseractOcr(
+  filePath: string, task: Task,
+  signal: AbortSignal | undefined, onProgress: OcrProgressCallback,
+): Promise<OcrResult> {
+  let resultText = "";
+  let tmpDir: string | null = null;
+
+  try {
+    if (isPdf(filePath)) {
+      onProgress("📄 Converting PDF pages to images…");
+      tmpDir = mkdtempSync(join(tmpdir(), "pi-tesseract-"));
+      const pageCount = await getPdfPageCount(filePath);
+
+      const pageResults: string[] = [];
+      for (let i = 0; i < pageCount; i++) {
+        if (signal?.aborted) throw new Error("Aborted");
+        const pageOut = join(tmpDir, `page_${i + 1}.png`);
+
+        try {
+          await convertPdfPage(filePath, i, pageOut);
+        } catch (e: any) {
+          pageResults.push(`## Page ${i + 1}\n\n> ⚠️ Skipped: ${e.message}`);
+          continue;
+        }
+
+        onProgress(`📄 Page ${i + 1}/${pageCount}`);
+        const pageText = await tesseractImage(pageOut, task);
+        if (!pageText.trim()) {
+          pageResults.push(`## Page ${i + 1}\n\n> ⚠️ No text detected`);
+        } else {
+          pageResults.push(`## Page ${i + 1}\n\n${pageText}`);
+        }
+      }
+      resultText = pageResults.join("\n\n");
+    } else if (isImage(filePath)) {
+      resultText = await tesseractImage(filePath, task);
+    } else {
+      throw new Error(`Unsupported file type: ${basename(filePath)}`);
+    }
+
+    return { text: resultText, details: { backend: "tesseract", task } };
+  } finally {
+    if (tmpDir) cleanupDir(tmpDir);
+  }
+}

package/extensions/types.ts

@@ -6,7 +6,7 @@ export const TASKS = ["text", "formula", "table", "figure", "auto"] as const;
 export type Task = (typeof TASKS)[number];
 
 /** All supported OCR backends */
-export const BACKENDS = ["ollama", "mineru", "pix2text"] as const;
+export const BACKENDS = ["mineru", "ollama", "tesseract", "pix2text"] as const;
 export type Backend = (typeof BACKENDS)[number];
 
 export interface OcrConfig {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-ocr",
-  "version": "1.0.2",
+  "version": "1.1.1",
   "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",