A fast terminal UI for searching and sharing the ebooks you already have on disk.

booksearch walks the current directory, builds a local index of every ebook it finds, and gives you instant typeahead search over titles, authors, and filenames. Pick a book and hit Enter to send it to someone via magic-wormhole without uploading anywhere or signing up for anything.

Why this exists

Calibre and Calibre Web are great if you want a managed library: a single canonical "library folder" that Calibre owns, with files renamed and reorganised into its own structure. That model is wrong if you already have a folder tree of ebooks you've curated yourself, or if your collection lives on a NAS that other tools also touch.

booksearch is built around a different premise:

• Your files stay exactly where they are. Nothing is moved, copied, or renamed.
• Your files are never modified. Metadata is read, not written back.
• The index is a sidecar. A single SQLite file (.booksearch.db) and a cover cache (.booksearch_covers/) live in the directory you ran it in. Delete those two and the tool leaves no trace.
• Search is local and instant. No server, no account, no network unless you explicitly choose to share a file.

Compared to Calibre / Calibre Web

If you want a complete library manager, use Calibre. If you want fast search over an existing folder tree without anything being modified, use this.

Prerequisites

• Rust toolchain (1.92 or newer) to build from source.
• A terminal. Cover images render best in kitty, WezTerm, iTerm2, or any terminal that supports the kitty graphics protocol, iTerm2 inline images, or sixel. Other terminals fall back to half-block rendering or skip covers.
• Calibre's ebook-meta CLI is required only if you want to index non-EPUB formats (.azw3, .azw, .mobi, .pdf, .fb2, .lit, .lrf, .kfx). EPUB files are parsed in-process with rbook and need no external tools.
  • macOS: brew install --cask calibre
  • Debian/Ubuntu: sudo apt install calibre
  • Or download from calibre-ebook.com
  • Make sure ebook-meta is on your PATH.

The recipient on the other end of a magic-wormhole transfer needs a wormhole client — install instructions at https://github.com/magic-wormhole/magic-wormhole.

Install

git clone https://github.com/rahimnathwani/booksearch
cd booksearch
cargo install --path .

This puts a booksearch binary in ~/.cargo/bin/.

Usage

cd /path/to/your/ebooks
booksearch

The first run scans the directory tree and indexes every supported file in the background. The TUI starts immediately; you can search while indexing is still in progress and books appear as soon as they're parsed. Subsequent runs only re-parse files whose size or mtime has changed.

Keys

Search behaviour

Filter terms are split on whitespace and matched as a logical AND. Each term is a case-insensitive substring match against the file name (with _ treated as a space), the book's title, and its author(s).

Typing is incremental: each new character only re-filters the previous result set, so search stays responsive even with hundreds of thousands of books.

Sharing a book

Hit Enter on the selected book. A modal appears with a wormhole code like 7-crossover-clockwork. Tell the recipient that code (over any channel — chat, voice, smoke signals). On their end:

wormhole receive 7-crossover-clockwork

The transfer is end-to-end encrypted. There's no upload step: the bytes flow directly between the two machines (or via a relay if direct connection isn't possible). Esc cancels at any time.

What gets stored

In the directory you run booksearch in:

• .booksearch.db — SQLite database with the index (FTS5 enabled).
• .booksearch_covers/ — extracted cover images, one per book, named by hash of the relative path.

Both are safe to delete; the next run will rebuild from scratch.

Add this to your .gitignore if your ebook directory is under version control:

.booksearch.db
.booksearch.db-shm
.booksearch.db-wal
.booksearch_covers/

Limitations

• No metadata editing. Use Calibre or a tag editor for that.
• No format conversion.
• No remote access. It's a local terminal app.
• Non-EPUB formats need Calibre's ebook-meta on PATH. Without it, those files are still listed (by filename) but title/author/cover won't be populated.
• A single very large or malformed file can take time to parse. The indexer enforces a 60 second timeout and records a placeholder so the same file isn't retried on every run; bump PARSE_TIMEOUT in the source if your environment legitimately needs longer.

Supported formats

EPUB (parsed in-process), plus the following via ebook-meta: AZW3, AZW, MOBI, PDF, FB2, LIT, LRF, KFX.

License

See repository for license details.