Buzee Tauri

⚠️ I have just open sourced Buzee. The documentation is lacking. Please open an issue and I'd be happy to help! ⚠️

The OCR capabilities in Buzee are built on top of Textra on Mac and WinOCR on Windows. Do check these repos out! I feel the native OCR capabilities of Mac and Windows are really good and should be used more!

What is Buzee?

Buzee (pronounced boozey) is a eight-year-old labrador retriever who can't play fetch but can love you like no other.

Buzee is also a full-text search application for your life. It helps you find your files, effortlessly.

Download v0.2.0 from this Releases page.

Screenshots

Features

• Fast, full-text search for all your documents, images, audio, video, folders, and browser history.
• Search all local documents and folders by keyword, time, type, or any combination of these.
• Ignore specific files or folders from being indexed. Or ignore only their content.
• Global shortcut. Press ⎇ / Alt + Space anywhere to show/hide the app. Modify it in the settings.
• In the app, press ⌘ / Ctrl + F or K to go to the search bar from any screen.
• In the app, press ⌘ / Ctrl + Shift + S to go to the scratchpad from any screen.
• View statistics about your files and get your Unique Document Profile.
• Sub-features:
  • Extract text from PDFs and Images using OCR.
  • Use a Scratch Pad to quickly jot down notes.
• Automatically syncs with changes on your filesystem.
• Lightweight installation package and low memory usage.
• Supports these default file types:

  Documents: csv, docx, key, md, numbers, pages, pdf, pptx, txt, xlsx, xls
  Images: jpg, jpeg, png, gif
  Books: epub, mobi, azw3, pdf
  Audio: mp3, wav, aac, flac, ogg
  Video: mp4, mkv, avi, mov, wmv
  

Semantic Search (somewhat)

Include/Exclude from Search

• Use the Filetype filter or simply type it in the search (like invoice pdf)
• Put quotes around keyword(s) to search for the exact phrase (like "annual report")
• Put a hyphen in front of the keyword to exclude it from search (like "annual report" -2022 -pdf)

Search using Time

Use the Date Range filter or simply mention the date/time period that you are looking for in your search. For example:

• last month pdf invoice
• annual report ppt this year
• q2 2023 to q3 2023 retail report xlsx
• prelim findings from 2017 to 2022
• cv docx from march 2 2020 to aug 15 2020 -pdf
• scope study 14/02/2015 to 10/08/2015

If you want to use a phrase as a keyword for search and not time, simply wrap it in quotes like this: invoice "March 2022"

Building from Source

Buzee works best on Mac. Windows may throw up some issues because I haven't had a change to properly test it. Linux is untested entirely, so you're on your own there.

1. Clone the repository.
2. Install Rust and NodeJS.
3. Run npm install in the root directory.
4. Run cargo install in the src-tauri directory.
5. Run npm run tauri dev in the root directory to run the app in development mode.
6. Run npm run tauri build in the root directory to build the app for production.

Building on Windows

Building on Windows requires a few changes. Follow these steps:

1. Remove drag = { path = "./crates/drag", version = "0.4.0", features = [ "serde" ] } from Cargo.toml
2. Comment out crate::drag::start_drag, from ipc.rs
3. Comment out mod drag; from main.rs

Finally, replace binaries/textra with binaries/winocr in tauri.conf.json.

---

TODO / Known Issues

Index:
(~) : partly implemented
(+) : has to be built from scratch
(?) : not sure if it will add great value

• (~) Show matching text for search results by reading from the body table.
• (~) Browser history search should support complex queries the way document search does.
• (~) Icon view should load thumbnails in an efficient, non-blocking manner. Thumbnails should show up on the page as they are loaded.
• (~) Enable adding 'comments' to documents.
• (~) Enable pinning documents/folders to the top of search results.
• (~) Allow user to add or remove supported file types.
• (~) Allow user to switch between profiles on Arc and Chrome. (Currently uses the default profile)
• (~) Test for Linux.
• (~) Improve the speed of parsing PDFs, Images and XLSX files. Especially OCR operations.
• (+) Enable adding 'tags' to documents.
• (+) Create a 'Dashboard' view that shows statistics, pinned documents, and recent searches.
• (+) Add tests to the codebase.
• (?) Record frecency of documents and use it to sort search results.

Tech Stack

Dependencies

Back-end:

• Rust
• Tauri v2
• SQLite
• Tantivy

Front-end:

• Svelte 4 using TypeScript
• shadcn-svelte
• TailwindCSS

See all dependencies in the Cargo.toml and package.json files.

Architecture

• All file metadata is stored in SQLite in the document table. A central metadata table stores the metadata from files and eventually cloud services, emails etc.
• A full-text index is created on metadata and stored as the metadata_fts table.
• Parsed text from documents is stored in the body table.
• A full-text index is created in Tantivy at the same time.
• The Firefox, Chrome and Arc history is searched using their respective history databases directly.
• All front-end code is in the src directory. All back-end code is in the src-tauri directory.

Contributing / Way Forward

Read the Vision and Roadmap.

I have spent two years building this project. It started as an Electron app, then I switched to Tauri for performance gains. When I started I barely new JavaScript and Svelte. Over the course of development, I learned NodeJS, TypeScript, SQLite, Rust, Tauri, Tantivy, and many other technologies. I learned so much about managing a project of this size and complexity. I am proud of what I have built but I am more proud of what I have learned.

I am now letting go of this project because I have other priorities. Please feel free to do with this project as you wish. I am happy to help you get started with the codebase.

If nothing else, this project can serve as an example of how to build a full-text search engine using Tauri and Tantivy. There are several tiny features and performance workarounds that I have implemented that you might find useful.

If you do do something with this project, please let me know. I would love to see what you build!

License

MIT