From 598873529a4f3483a5ef1635476f4fa90808c9d2 Mon Sep 17 00:00:00 2001 From: Brock Wilcox Date: Mon, 28 Oct 2024 20:44:24 -0400 Subject: [PATCH] README updates --- README.md | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index cdc01f5..a6e5f04 100644 --- a/README.md +++ b/README.md @@ -85,12 +85,19 @@ Draw some stuff on your screen, and then trigger the assistant by *touching/tapp * External stuff * Let it look things up * Let it send me stuff ... emails, slacks +* Evaluation + * Create a set of screenshots for inputs + * Represent different use-cases + * Some of these, such as TODO-extraction, might have specific expectations for output or execution, but most of them won't + * Run through the system to get example output -- text, svg, actions + * Write a test suite to judge the results .... somewhat human powered? Separate VLM judge? ## References * Adapted screen capture from [reSnap](https://github.com/cloudsftp/reSnap) -* Techniques for screen-drawing inspired from [lamp](https://github.com/rmkit-dev/rmkit/blob/master/src/lamp/main.cpy) +* Techniques for screen-drawing inspired from [rmkit lamp](https://github.com/rmkit-dev/rmkit/blob/master/src/lamp/main.cpy) * Super cool SVG-to-png done with [resvg](https://github.com/RazrFalcon/resvg) * Make the keyboard input device even without a keyboard via [rM-input-devices](https://github.com/pl-semiotics/rM-input-devices) +* Not quite the same, but I recently found [reMarkableAI](https://github.com/nickian/reMarkableAI) that does OCR→OpenAI→PDF→Device ## Development