crossword-solver

OpenCV + Tesseract + LLM-based crossword puzzle solver.

Overview

This project converts a photographed crossword into a structured JSON puzzle, queries LLM solvers for answers, and validates/visualizes the results. The pipeline is modular: image processing → JSON construction → LLM integration → solution validation.

Key features

• Grid extraction and per-cell OCR (Tesseract).
• JSON puzzle builder with word lengths and intersections.
• Pluggable LLM solver integrations (multiple providers supported).

Requirements

• Python 3.8+
• OpenCV (cv2), Pillow, pytesseract
• API keys for any LLM providers used (configured in project)

Quick start

1. Install dependencies (example): pip install -r requirements.txt
2. Ensure Tesseract is installed and pytesseract.tesseract_cmd points to the binary.
3. Run the main program: python crossword.py

Project layout (high level)

• crossword.py — main pipeline and orchestration
• digit_recogniser.py — cell extraction and OCR helpers
• digit_classifiers/ — per-detector implementations (Tesseract-based, etc.) - contains multiple digit-classifier implementations for comparison and testing
• generate_json.py — builds puzzle JSON (lengths, intersections)
• llm/ — LLM solver wrappers and prompts
• overlay_grid.py — visualization/overlay of solutions