shotext

Your screenshots are a goldmine, stop losing what's in them.

Every day you capture code snippets, meeting notes, error messages, and diagrams as screenshots. They pile up on your Desktop and become a black hole: impossible to search, impossible to organise. shotext fixes that.

It watches your screenshots folder, automatically runs OCR to extract the text from every image, and stores it in a full-text search index. Seconds after you press ⌘⇧3, the content of that screenshot is searchable from your terminal.

How It Works

Screenshot taken ──▶ Detected ──▶ OCR (Tesseract) ──▶ Indexed (Tantivy) ──▶ Searchable

Capture: Take a screenshot as you normally would on macOS.
Detect: shotext watches ~/Desktop (or a custom folder) and picks up new .png files automatically.
Extract: Tesseract OCR reads the text from the image. Each file is blake3-hashed for deduplication.
Index: The extracted text, file path, and timestamp are stored in a Sled database and a Tantivy full-text search index.
Search: Find any screenshot by what was in it, not by its filename.

Installation

Prerequisites

Rust toolchain (cargo)
Tesseract OCR engine and language data

# Install Tesseract via Homebrew
brew install tesseract

# Clone and build
git clone https://github.com/cladam/shotext.git
cd shotext
cargo build --release

# The binary is at target/release/shotext

Quick Start

# 1. Initialise config (creates ~/.config/shotext/config.toml)
shotext config

# 2. Index all existing screenshots on your Desktop
shotext ingest

# 3. Start watching for new screenshots in real-time
shotext watch

# 4. Search, what was in that error message?
shotext search "connection refused"

# 5. Or launch the interactive fuzzy finder
shotext search

Commands

Command	Description
`shotext ingest`	Scan the screenshots folder and index all new images. Use `-f` to force re-index everything.
`shotext watch`	Watch the folder for new screenshots and index them the moment they appear.
`shotext search <query>`	Full-text search across all extracted text. Supports stemming ("develop" → "developer", "developing").
`shotext search`	Interactive fuzzy finder (powered by skim) over all indexed screenshots.
`shotext list`	List all indexed screenshots. Add `-v` for dates and text snippets.
`shotext view <target>`	Open a screenshot in a native GUI viewer alongside its extracted text. Accepts a file path or a blake3 hash.
`shotext config`	Show the current configuration. Use `-e` to open it in your `$EDITOR`.

Configuration

Config lives at ~/.config/shotext/config.toml and is created automatically on first run:

[paths]
screenshots = "/Users/you/Desktop"   # Folder to watch / ingest
database = "/Users/you/Library/Application Support/shotext/index.db"

[ocr]
language = "eng"   # Tesseract language code (eng, swe, deu, …)

Tech Stack

Component	Crate	Role
CLI	clap	Command parsing with derive macros
OCR	tesseract	Text extraction via libtesseract
Hashing	blake3	Fast deduplication of screenshot files
Database	sled	Embedded key-value store for metadata
Search	tantivy	Full-text search index with stemming
Watcher	notify	Filesystem polling for new screenshots
Fuzzy find	skim	Interactive terminal search UI
Viewer	eframe/egui	Native GUI window for image + text

License

MIT

shotext 0.1.2