OCR Scanner

😫 The Pain Point

Boss hands you a printed paper: “Type this into Word for me.” Or you have a PDF that is actually just a picture of text. You can’t copy-paste anything. Retyping 10 pages manually is painful.

🚀 Agentic Solution

Optical Character Recognition (OCR): The computer “reads” the pixels and converts them back to text letters.

Key Features:

• Language Support: Can read English, Vietnamese, or any language (if pack is installed).
• Layout: Preserves paragraphs.

⚔️ Phase 1: Commander (Quick Fix)

For a single page conversion.

Prompt:

“Use pytesseract to read text from scan.jpg. Save the content to output.txt.”

Result: The text content extracted.

🏗️ Phase 2: Architect (Permanent Tool)

For Librarians/Data Entry.

Engineering Prompt:

**Role:** Python AI Developer
**Task:** Create an "OCR Tool".
**Requirements:**
1.  **Prerequisite:** User must install **Tesseract-OCR** engine separately. Check for installation.
2.  **GUI:**
    *   Select Source (Image or PDF).
    *   Language selection (eng/vie).
    *   "Convert to Word" button.
3.  **Logic:**
    *   If PDF: Convert to images first (`pdf2image`).
    *   Run `pytesseract.image_to_string(img)`.
    *   Save text to `.docx`.
4.  **Deliverables:** `ocr_tool.py`, `run.bat` (Windows), `run.sh` (Mac).

🧠 Prompt Decoding

• Dependency Hell: OCR is tricky because it needs an external engine (Tesseract) installed on the OS. The prompt warns the user about this expectation to prevent “Command Not Found” errors.

🛠️ Instructions

1. Install Tesseract-OCR.
2. Copy Prompt -> Paste -> Run.
3. Select Image -> Convert.