Repository navigation
Conversation
ai -i opens a full-width TUI. You type the prompt; mode, model, count, aspect ratio and duration are picked from the footer with the keyboard, or with / commands. Images and video frames show as thumbnails, clicking one opens a full-screen slider, and @last feeds the last output into the next prompt. - Bare ai and every one-shot command behave as before. - text errors now throw instead of calling process.exit, so they can run in-process. - The Gateway model catalog is cached per URL and headers, so image and video commands stop refetching it on every call.
- Split interactive.ts into input, settings, rendering, terminal and formatting modules; behavior is unchanged. - Tab and Shift+Tab now step through the setting chips and wrap back to the prompt. - Tests drive real key sequences through the TUI and pin the one-shot Kitty preview bytes. - Document ai -i in the README, a new website page, the agent skill and the changelog.
- Restore mouse mode, cursor and raw mode on SIGINT, SIGTERM and SIGHUP, and close the gallery cleanly when an image file cannot be read. - Show every failure from a turn, not only lines that mention errors. - Keep the gallery and clicks closed until a turn has fully rendered. - Fit live lines to the terminal width so redraws and thumbnail click positions stay aligned on narrow windows. - Buffer mouse reports split across reads, clamp the slash menu selection, and edit emoji as whole characters. - Keep the indentation of code in text replies. - Load the model list from the same Gateway URL and team as generation. - @last now always points at the last image.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
- Strip control characters from model text, errors and notes before printing, so a reply cannot clear the screen or write the clipboard. - Keep the full wrap width for every line of a reply. - Show a clear error when the model list cannot load instead of "No matches". - Name the model that actually ran in the result line. - Keep a /image, /video or /text prompt typed mid-turn in the prompt bar. - Cut the loading area to the terminal height. - Edit emoji, flags and accents as whole characters. - ai -i now needs stdin, stdout and stderr to be terminals, and says when it is given extra arguments. - Test the TUI against the real image, video and text flag definitions, the result path and @last, plus the catalog cache keys.
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
ai -iopens an interactive terminal UI where you type the prompt and pick everything else, with thumbnails, a click-to-open slider and@lastchaining.Demo
A text reply, then two images with settings picked by Tab, and a click on a thumbnail that opens it full screen (played at 2x).
demo.mp4