📖 Polytonic Proofreader (But easily convertible to other scripts, see below)

A browser-based polytonic Greek OCR proofreader.

Originally created for proofreading ΤΟ ΣΥΜΠΑΝ (THE UNIVERSE) (Smyrna, 1888) — a family heirloom the author wanted to digitise and read. The TCP port 1888 is the year of printing of this specific book. Although it is still a Greek polytonic proofreader, it can easily (almost instantly with an AI) be converted to another script

💡 Rationale

Already in the research phase digitising the book, it was obvious that a dedicated proofreader/editor was needed:

• Web-based (inside the browser) to reduce the friction of switching tasks to a minimum. One tab for the AI OCR extractor next to the proofreader tab.
• Only one script (polytonic Greek in this case). It would add unnecessary friction for such a hard job to support more at the same time.
• Single file so it can easily be converted (with AI) to another script
• Having specific advanced capabilities, allowing for example to look for misspelled words or words spanning two lines (common in old books), or a magnifying glass. At the same time, very common functionality (search and replace for example) is omitted, as not useful for this task.
• Minimal and easy customisation.
• Preventing common and detrimental mistakes, like saving modern Greek text to the polytonic ocr folder, or saving polytonic text to the el folder.
• Allowing to write Greek characters without even switching to a Greek keyboard. Additionally the character palette allows typing accents and the rest.
• Oriented only for proofreading; the bulk of the text comes from OCR (AI). The project includes the gemini-prompt.txt I have used.
• Written by the person who used it. This feedback loop can create software no generic program can match. Of course this means another person might dislike it, but as we have said it is a single-file program. Open an AI (web-based, or a coding agent) dialogue and say what you want.

🚀 Setup & Workflow

• Download the repo locally on your PC. Inside the repo folder, create a folder my_book (the book title).
• Place page images — inside my_book/scans/, put your scanned page images (JPG, PNG, WebP, GIF, TIF).
• Install Go:

  # Debian / Ubuntu
  sudo apt update && sudo apt install golang-go

• Run the proofreader:

  go run proofreader.go my_book

• Open the link the proofreader shows in a browser tab.
• Get OCR text from an AI vision model — in a nearby browser tab, feed each scan image to an AI vision model (e.g. Gemini, GPT‑4o, Claude) together with a prompt instructing it to output polytonic Greek. Copy the returned OCR text.
• Navigate to the page — use the ← Previous / Next → toolbar buttons to go to the page you just processed.
• Paste the OCR — paste the AI's output into the text pane on the right.
• Proofread — compare the scan image (left) against the text (right). Use the accent‑ignoring search (Ctrl+F) to spot-check dubious words, and the Greek character palette (Ctrl+P) to type corrections. The AI vision model can also suggest what to check again — follow its recommendations.
• Read and annotate — the purpose is not just to correct OCR errors, but to actually read the book as you go. Add comments for small errors in the original text, noteworthy topics, or anything that catches your interest — inside brackets like [λέξη: επεξήγηση] or as plain notes. These annotations persist in the saved text.
• Save and move on — press Save (Ctrl+S) and repeat for the next page. Unsaved changes show a * in the browser tab title.

📁 Book structure

The program discovers scans and text directories by scanning the project directory you pass as argument. Here's how it works:

📂 Scan directories

Any subdirectory whose name starts with "scans" is treated as a scan directory:

• scans — primary scan folder
• scans2, scans_backup, etc. — additional scan sources

The folder named exactly scans is always listed first. You can switch between them using the source dropdown in the toolbar.

🖼️ Accepted image formats

Only files with these extensions are recognised as page images: .jpg, .jpeg, .png, .webp, .gif, .tif, .tiff

Non-image files and nested subdirectories are ignored.

📝 Text (OCR) directories

The program looks for folders named exactly ocr and el (no prefix matching) under the project directory. These store the OCR text files — one .txt file per page.

If neither ocr nor el exists, the program automatically creates ocr/ after validating that the scans structure is present.

📐 Example layout

The repo comes with ready-to-use examples in the examples/ directory:

project/
├── scans/                # page images (jpg, png, webp, gif, tif)
│    ├── 048.jpg
│    ├── ...
├── ocr/                  # OCR text files (one per page)
│    ├── 048.txt
│    └── ...
└── el/                   # modern Greek OCR text (optional)
     └── 048.txt

Run go run proofreader.go examples to try it immediately (use -p 1889 to pick a different port).

🔧 Custom layout

Pages sharing the same 3‑character prefix (e.g. 011.jpg, 011b.jpg) are grouped together. Use the source dropdown to switch between image variants while keeping the same text.

⚙️ Configuration

There are no environment variables, no config files for paths — it's all filesystem-based with the hardcoded "scans" prefix as described above.

🔌 Port

The server runs on port 1888 by default. Use the -p / --port flag to change it:

go run proofreader.go -p 1889 examples
# Open http://localhost:1889/

⚡ Quick start

# Try it with the built-in examples
go run proofreader.go examples
# Open http://localhost:1888/

# Or use your own project
go run proofreader.go /path/to/project

# Use a different port, for tests, or having 2 books in different tabs
go run proofreader.go -p 1889 /path/to/project

✨ Features

🖥️ Side-by-side view

• Left pane — scan image (wheel to scroll vertically, Ctrl+wheel to zoom, click-drag to pan, double-click to fit/zoom)
• Right pane — editable OCR text with wheel to scroll, Ctrl+wheel to change font size

✏️ Edit mode

Click ☐ Edit (or Ctrl+E) to toggle between:

Mode	Behaviour
View	Drag‑scroll text, font resize via scrollwheel, read‑only
Edit	Type corrections, click for cursor, undo/redo work

🎨 Polytonic Greek character palette

Click Greek Char (or Ctrl+P) to open a popup with all accented/breathing‑mark variants. Click any character to insert it at the cursor position. The leftmost column (the bare letter) is also clickable — you can insert α without switching your keyboard.

Undo (Ctrl+Z) works for palette insertions and direct typing.

🔤 Latin → Greek transliteration ("Force Greek")

When ☐ Force Greek is checked (default), both the search box and the text area automatically convert Latin letters to their Greek QWERTY equivalents as you type:

English key	→	Greek		English key	→	Greek
a	→	α		n	→	ν
b	→	β		x	→	χ
g	→	γ		o	→	ο
d	→	δ		p	→	π
e	→	ε		r	→	ρ
z	→	ζ		s	→	σ
h	→	η		t	→	τ
u	→	θ		y	→	υ
i	→	ι		f	→	φ
k	→	κ		c	→	ψ
l	→	λ		v	→	ω
m	→	μ		q	→	ς
				j	→	ξ

Uncheck to type Latin letters normally. The setting persists across page changes and browser refreshes via localStorage.

🔍 Search

The search field finds polytonic Greek text, ignoring accents and breathing marks. Typing καλος also matches καλός, καλῶς, etc. Search is hyphenation‑aware: line‑break hyphens (-\n) are skipped during matching.

📋 Copy All

Click Copy All to copy the entire current page text to your clipboard. Shows a confirmation in the status bar.

🔀 Digraph matching (experimental)

When ☐ Digraph matching is checked (default: on), the search engine also matches historical spelling variants:

• ει matches η, ι, υ (all pronounced similarly in later Greek)
• αι matches ε This helps find words that may have been spelled with different vowel letters in the original text vs. the OCR output.

🔎 Magnifier

Hover over the scan image to see a 3× magnified lens that follows your cursor — useful for checking small diacritics and breathing marks in polytonic Greek. The magnifier can be toggled on/off:

• Settings popup — check ☐ Always-on magnifier
• Middle-click on the image pane — quickly toggle the magnifier on or off
• Ctrl+M — keyboard shortcut to toggle

When the magnifier is off, the cursor changes back to a grab-hand for panning.

🔢 Line numbers

In Settings, toggle ☐ Lines to show/hide line numbers on the left side of the text pane. Lines that are blank or contain only metadata (e.g. [header]) are skipped by the numbering.

🗂️ Image source selector

Use the source dropdown (between the page selector and the text source) to switch between multiple scan directories — e.g. scans (primary) and scans2 (alternative/cleaner images of the same pages).

📄 Text source selector

Use the text source dropdown (next to the image source) to switch between multiple OCR sources — e.g. ocr (polytonic Greek) and el (modern Greek). This allows comparing or saving to different text directories.

⚙️ Settings

Click Settings (no shortcut) to open a popup with the following toggles:

Setting	Default	Description
☐ Lines	On	Show/hide line numbers in the text pane
☐ Always-on magnifier	Off	Keep the magnifier visible while hovering the image
☐ Force Greek	On	Auto-convert Latin letters to Greek in the text area and search box
☐ Digraph matching	On	Match historical spelling variants during search
☐ Restart button	On	Show/hide the Restart button in the toolbar
☐ Edit enabled by default	Off	Start every page in edit mode instead of view mode
☐ Auto-save (30s)	On	Automatically save the current page every 30 seconds when modified

All settings persist across page changes and browser refreshes via localStorage.

❓ Help

Click Help (no shortcut) to view a popup with a full reference of image-pane interactions, text-pane interactions, and global keyboard shortcuts.

🔄 Server restart (hot‑reload)

After editing the Go source, click the Restart button (or visit /restart) to recompile and restart the server in-place. The browser automatically refreshes after 1 second — no need to restart the terminal command. The button is disabled if Go or the source file is not found.

⌨️ Keyboard shortcuts

Shortcut	Action
Ctrl+E	Toggle edit mode
Ctrl+F	Focus search box
Ctrl+G	Toggle Force Greek
Ctrl+M	Toggle magnifier
Ctrl+D	Toggle digraph matching
Ctrl+P	Open Greek character palette
Ctrl+S	Save current page
Ctrl+Z / Y / Shift+Z	Undo / redo (native browser, always works)
Escape	Close Greek palette

🗂️ Project structure

├── proofreader.go       # single‑file Go server + embedded HTML/CSS/JS
├── README.md
├── .gitignore
├── reasonix.toml        # AI coding agent configuration
├── gemini-prompt.txt    # prompt used with AI vision models for OCR
├── screenshot.jpg       # screenshot for the README
└── examples/            # ready‑to‑run sample scans + OCR text
    ├── scans/
    ├── ocr/
    └── el/

No dependencies beyond the Go standard library. No go.mod, no build step — just go run.

🤔 Why there is no need for binaries

This program is distributed as source code only — no precompiled binaries, no installers. You get the Go compiler, and you're ready to go.

🎯 The real reason: customisation

The book you are digitising is unique. Its script, its layout, its quirks — no binary could anticipate them all. By keeping the program as a single editable source file, you — or any AI coding agent — can:

• Change the character palette — add or remove diacritic combinations specific to your book's orthography
• Adapt the transliteration map — re-map keyboard keys for your script's letters
• Tweak the search behaviour — add or remove digraph equivalences for historical spellings
• Modify the save logic — add book-specific metadata headers or annotation conventions
• Add entirely new buttons — anything from "Insert page number" to "Export to Markdown"

Everything is in one file, every feature is a few lines of JavaScript or Go away from being changed. No build tools, no recompilation step — just edit, save, and click Restart.

🌍 Forking for another script

The Go server is script-agnostic — only the embedded frontend (character palette, transliteration map, text detection) is script-specific. To adapt this tool for Arabic, Cyrillic, Hebrew, or any other writing system:

1. Copy proofreader.go → your-fork.go
2. Give it to an AI — paste the file into any code agent (Claude, GPT, Gemini) or web AI chat tool and tell it what script you need. Since everything is in one self-contained file with clear section markers, the AI will understand the structure and make the changes.
3. Run with go run your-fork.go /path/to/project.

No Go dependencies beyond the standard library, no build step — just Go installed.

---

Developed with the assistance of Reasonix Code (Deepseek V4), an AI coding agent. The design, workflow, and feature decisions were made by the author; the agent translated them into code.

⚖️ License

MIT. Do what you want with it.