Arrows is an experimental frontend for creative writing with large language models (LLMs).
video.mp4
It is designed around three principles:
- Generate whole paragraphs: Paragraphs are the natural semantic units of prose. Instead of generating a fixed number of tokens like some other AI writing tools, Arrows stops requesting more tokens at the end of the paragraph, presenting a piece of text to the user that makes sense on its own.
- Give the user two options to choose from: Arrows generates two versions of the next paragraph, and presents them to the user as two possible continuations. With a correctly configured backend (see llama.cpp arguments below), this can be done as a batched operation, which is essentially free, doubling the effective generation speed!
- Zero UI: Arrows has no buttons, labels, or menus. Its user interface consists of the text you are working on, and nothing else. All LLM operations are performed using only six intuitive keyboard shortcuts (see below).
Arrows supports any backend that exposes an OpenAI-compatible text completion API. This includes text-generation-webui, the llama.cpp server, and of course the OpenAI API itself.
Note that Arrows is intended for a completion-based writing workflow like the one I described in this Reddit post. If you want an instruction-based workflow where you explicitly tell the AI what to write next, then this is not the frontend you are looking for.
Arrows is a browser-based text editor and supports all standard text editing shortcuts. In addition, the following shortcuts are provided to use the LLM writing features:
Ctrl+Enter (on Windows/Linux) or ⌘+Enter (on macOS): Generate two options for the paragraph following the current cursor position.
← Left Arrow: Choose left option and continue generating from there.
→ Right Arrow: Choose right option and continue generating from there.
↓ Down Arrow: Generate new options.
↑ Up Arrow: Undo last insertion and generate from there.
Escape: Return to normal editing.
If you have Git and Node.js installed, installing Arrows is straightforward:
git clone https://github.com/p-e-w/arrows.git
cd arrows
npm install
The easiest way to use Arrows locally is to just run
npm run dev
from the installation directory and then open the displayed URL in the browser. You can also use
npm run build
to build a minified production version, and then serve the dist directory
using the web server of your choice.
To set the base URL, API key, and generation parameters, simply edit the file
src/config.ts.
Arrows has a fully asynchronous architecture and is capable of streaming two completion options in parallel, as demonstrated in the video above. This can save valuable user and GPU time.
However, in order for parallel generation to work, the backend server has to support concurrent requests, ideally with automatic batching. As of June 2024, text-generation-webui does not support concurrent request handling. When used as a backend for Arrows, the left/right choices will be generated sequentially rather than in parallel.
A backend that does support concurrent requests is the llama.cpp server. To enable this feature, simply run the server with the arguments
--parallel 2 --cont-batching
Copyright © 2024 Philipp Emanuel Weidmann ([email protected])
This program is free software: you can redistribute it and/or modify it under the terms of the GNU Affero General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License along with this program. If not, see https://www.gnu.org/licenses/.
By contributing to this project, you agree to release your contributions under the same license.