SLYE: Speak like you eat
SLYE adds a plain-language rewrite of the agent's answer in a separate card under the original, on /slye or automatically.
- 140 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The agent answer text goes to the selected model through the host model registry.
- Automatic mode spends requests after every eligible answer.
Install
Manual install
pi install npm:speak-like-you-eatFirst README command: for all projects. The `-l` flag installs for the current project only.
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The agent answer text goes to the selected model through the host model registry.
- Automatic mode spends requests after every eligible answer.
This is third-party code. Review the repository files before installing.
What it does
SLYE, a package for Pi and Oh My Pi, adds a card with a jargon-free rewrite after a completed answer. The original stays on screen unchanged, and the card never enters the model context. `/slye` rewrites the latest answer on request, `/slye on` enables automatic rewriting, `/slye off` disables it and `/slye model` picks the model. On Pi a card has two tabs: ELI5 opens with the conclusion and required actions, while Light stays close to the original wording. A custom system prompt goes into a `slye-prompt.md` file.
Who it is for. People who read long agent answers in Pi and lose time decoding jargon and lists of steps.
Good fit when
- The agent produced a long technical report and a two-sentence conclusion is needed.
- An answer is a pile of terms and the reader must work out what to do next.
- Automatic rewriting is enabled for long answers so nobody has to ask each time.
- A cheap fast model handles only the rewrite while the main model does the work.
Not a fit when
- A short answer needs a rewrite, while automatic mode starts at 200 characters of prose.
- Work happens in Claude Code or Cursor, but the package targets Pi and Oh My Pi.
- The original answer must not be sent to a second model.
Example request
Rewrite the latest answer in plain language with `/slye` and keep the code and commands unchanged.Limitations
A current Pi release is required, and older ones are unsupported. On Pi each rewrite makes two provider requests in parallel and costs roughly twice a single one, while Oh My Pi makes one. Automatic mode works only for a normally completed answer of at least 200 characters of prose outside code blocks. The model ranking in the README rests on a small manual trial and is no guarantee: compare the result with the original answer, especially code examples.
How to disable. Run `/slye off` to stop automatic rewriting, or remove the package with Pi or Oh My Pi tooling. The saved model and settings live in the Pi configuration.
FAQ
How do I install SLYE in Pi?
Run `pi install npm:speak-like-you-eat` for all projects or `pi install -l npm:speak-like-you-eat` for the current one. Oh My Pi uses another command: `omp plugin install speak-like-you-eat`.
What happens on the first `/slye`?
With no usable model, a model and scope picker opens. The choice is saved with automatic rewriting off, and the first rewrite runs immediately.
How do the ELI5 and Light tabs differ?
ELI5 rebuilds the answer for a reader who wants the point: conclusion, actions, steps, facts and open decisions. Light stays close to the original wording or to your prompt. Switch tabs with `ctrl+\`.
Which models does the author suggest?
The README names Claude Haiku 5.5, GPT-6 Luna, DeepSeek V4.1 Flash and Gemma 4 31B in that order. These are observations from a small manual trial, not guarantees.
How do I cancel a rewrite?
Press Escape. Cancellation, a timeout or a provider failure leaves the original unchanged, and you can retry manually.
Discussion
No questions or reviews yet.
Related
PluginsTwo skills for condensed explanations of AI output: eli5-succinct answers briefly in chat, summary builds a single HTML page.
/bro
bro
The /bro command re-explains the assistant's last reply in plain language while commands and numbers stay as they were.
Skill silently rewrites your request for the model family that answers it, then answers without showing the rewrite.
A Claude Code skill that adapts any explanation to its audience: a child, a manager, an engineer or your parents
Skill that reads your prompt history, finds the asks you repeat and turns each into a short command such as /sent
A skill that analyzes materials and subtitles, checks facts against sources and keeps a personal library in a local folder.
README in short
The README explains the commands, ELI5 and Light tabs, recommended models and guarantees: the original is untouched and the card stays out of context. Details live in the specification and benchmark report under `backlog/docs`.