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

@@ -1,7 +1,7 @@
//! Warm, process-wide service singletons — the LSP client, code graph, DAP debugger, and
//! diagnostics cache — plus the per-project session store.
//!
//! Luxe runs one project per process, so each service is a `OnceLock` singleton reused across
//! luxe runs one project per process, so each service is a `OnceLock` singleton reused across
//! every tool call and hook instead of being spawned per request.
use std::sync::Arc;
@@ -56,7 +56,7 @@ pub(crate) fn tool_statuses() -> Vec<luxe_core::ToolStatus> {
/// Three inputs, and the judgement is here rather than in `luxe-lsp` so it can be read and tested
/// in one place: the user's preference, whether a front-end can prompt, and `--trust-project`.
///
/// `interactive` carries the whole argument for the default. Luxe is a coding agent someone sits
/// `interactive` carries the whole argument for the default. luxe is a coding agent someone sits
/// in front of; the loop is a feature, not the usual way it runs. With a person there, using the
/// server the project pinned is ordinary work — and `node_modules/.bin` exists only because
/// `npm install` already ran that project's install scripts, which is arbitrary code and a wider
@@ -96,7 +96,7 @@ pub(crate) fn project_tooling_allowed() -> bool {
/// The process-wide warm LSP service. A singleton (like [`shared_diagnostics_store`]) so the tools,
/// the diagnostics-on-edit hook, and the workspace collector (which reads running-server state for
/// `/lsp`) all share one instance. Luxe runs one project per process; the collector scopes
/// `/lsp`) all share one instance. luxe runs one project per process; the collector scopes
/// running servers to the active root.
pub(crate) fn build_lsp_service() -> SharedLspService {
static SERVICE: std::sync::OnceLock<SharedLspService> = std::sync::OnceLock::new();
@@ -112,7 +112,7 @@ pub(crate) fn build_lsp_service() -> SharedLspService {
/// The process-wide warm code graph service, shared by the `impact` / `module_report` tools and the
/// graph-invalidation hook. Singleton like the LSP/DAP services; rooted at the workspace on first
/// use (Luxe runs one project per process).
/// use (luxe runs one project per process).
pub(crate) fn build_graph_service(
repo_root: &std::path::Path,
) -> Arc<luxe_lsp::graph::GraphService> {
@@ -223,7 +223,7 @@ fn read_config_text(path: &std::path::Path) -> Result<Option<String>, String> {
///
/// A shared handle at the binary boundary: it lets the (core) hook publish per-file diagnostics
/// that the (core-agnostic) collector reads for the Context Dock without threading a channel
/// through every front-end. Luxe runs one project per process; the collector filters entries to the
/// through every front-end. luxe runs one project per process; the collector filters entries to the
/// active root, so a project switch never leaks another tree's diagnostics.
pub(crate) fn shared_diagnostics_store() -> luxe_core::lsp::DiagnosticsStore {
static STORE: std::sync::OnceLock<luxe_core::lsp::DiagnosticsStore> =