luxe 0.1.4: remembered models, wheel-driven lists, and a steadier phone app

- luxe starts on the model you last chose in this project, then the last one anywhere. Signing in
  opens the model picker instead of choosing for you.
- Every list in a modal takes the mouse wheel and a draggable scrollbar.
- The phone app shows reasoning in full, stays put while you read back through a long session, and
  shows a question's answer on its card.
- On Windows, luxe update replaces the running luxe, and a signed-in install no longer sees setup
  again.
- The file tree no longer shows "engine busy" in a large project, loop output is easier to follow,
  and the name is written lowercase.

The CHANGELOG has the full list.
This commit is contained in:
Evan Whittaker
2026-10-20 13:51:34 -05:00
parent 6184aa21fb
commit a0666d6c81
591 changed files with 4926 additions and 2942 deletions

View File

@@ -1,4 +1,4 @@
//! Session claims — one Luxe process writes a session's log at a time.
//! Session claims — one luxe process writes a session's log at a time.
//!
//! Two processes appending to one log interleave entries that each parent on their own leaf, so a
//! resume walks one branch and silently drops the other's turns. A claim is an OS lock on a
@@ -70,7 +70,7 @@ mod tests {
file.try_lock().is_ok()
}
/// The break this catches: a claim that takes no OS lock, so a second Luxe writes beside it.
/// The break this catches: a claim that takes no OS lock, so a second luxe writes beside it.
#[test]
fn a_claim_keeps_other_holders_out_until_it_drops() {
let dir = tempfile::tempdir().expect("tempdir");

View File

@@ -34,7 +34,7 @@ pub enum EntryKind {
/// What one cache-warming call cost ([`crate::cache_warming`]).
///
/// A kind of its own rather than a [`EntryKind::Usage`] with a new source, because a source an
/// older Luxe can't parse fails that entry as corrupt and takes the whole session load with it,
/// older luxe can't parse fails that entry as corrupt and takes the whole session load with it,
/// while an unknown *kind* degrades to [`EntryKind::Unknown`] and is skipped. Inert for context
/// assembly, and counted under [`crate::usage::UsageSource::CacheWarm`] on resume.
CacheWarm {
@@ -71,7 +71,7 @@ pub enum EntryKind {
/// Where the worktree was, relative to the project
path: String,
},
/// An entry written by a newer Luxe than this one.
/// An entry written by a newer luxe than this one.
///
/// A session log is append-only and shared across versions, so an unrecognized kind must not
/// make the whole transcript unreadable. It is kept (never rewritten) so the parent chain —
@@ -98,7 +98,7 @@ pub struct Entry {
/// Hand-written because the derive cannot deliver what [`EntryKind::Unknown`] promises.
/// `#[serde(other)]` may only sit on a *unit* variant, and in an adjacently tagged enum serde will
/// only match one when the content field is absent — so `{"kind":"future","data":{…}}`, which is
/// what a newer Luxe actually writes, failed the whole load and took every earlier entry with it.
/// what a newer luxe actually writes, failed the whole load and took every earlier entry with it.
///
/// The retry below is what separates "a kind we have never heard of" from "a kind we know, written
/// wrong". Every known variant carries fields, so stripping `data` and trying again can only

View File

@@ -24,7 +24,7 @@ fn break_first_line(path: &Path) {
fs::write(path, bytes).unwrap();
}
// The break this catches: a cache that never looks at the file again, so another Luxe's entries
// The break this catches: a cache that never looks at the file again, so another luxe's entries
// (or a write this store didn't make) never show up
#[test]
fn entries_another_store_appends_are_read_on() {

View File

@@ -373,7 +373,7 @@ fn a_known_kind_written_wrong_still_fails_loud() {
);
}
/// [`EntryKind::Unknown`] makes three promises about a log a newer Luxe has written to: it
/// [`EntryKind::Unknown`] makes three promises about a log a newer luxe has written to: it
/// stays readable, the unrecognized entry is kept rather than dropped, and the parent chain
/// still walks through it. All three were documented; none were tested.
#[test]