- Remote projects are remembered as ssh:// locations: the last session and
a reloaded window reconnect, and the welcome screen lists them
- A rustup proxy without its component counts as a missing tool (install
hint) rather than a server crashing on every restart
- Typo checks the names declarations introduce, not path segments and
calls; more programming terms in the word list
- The query history records statements in execution order
Jump to Source moves next to Next Difference in the diff toolbar, with the edit (pencil) icon. Diffs where neither side is a file on disk (Git Log commit files, Compare Branches, multi-commit diffs) now offer it when the path still exists in the working copy; files deleted by the commit don't.
Jumping from a pane that isn't the opened file used the other pane's caret. The caret line is now mapped through the diff to the target (the other pane's document, or the working copy read at click time including unsaved edits); the column is kept only when the line is unchanged.
Co-authored-by: itning <[email protected]>
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
Adds gutter markers for type hierarchies: ↓ "Is implemented / Has subclasses / Is overridden" on types and methods with implementations, ↑ "Implements / Overrides method in X" on methods that implement or override a supertype method. One target navigates, several open the Choose Implementation popup. Settings › Editor › General › Gutter Icons gets two toggles.
Markers come from the language server: a new batched `lsp_implementations` command (up to 4 concurrent, cancellable, sharing the goto query path) for ↓, and the Super Method lookup for ↑, which now prefers the overload with matching parameters (also fixes ⌘U on overloads). Only declarations near the viewport are queried, debounced, cached per document version and cancelled when superseded. Works for any server with implementation / type hierarchy support; servers without type hierarchy show ↓ only.
Fixes#99
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
In Git Log's Branches pane a single click filtered the log and a double click checked the branch out. Now a click only selects; double click or Enter shows the log of that ref (also the current branch and HEAD), and Enter on a multi-selection filters by all selected branches in tree order. Checkout stays in the context menu. The pane keeps its own selection and follows filter changes made elsewhere. Cmd/Ctrl+C copies the selected branches as full refs, one per line (HEAD as `HEAD`), leaving copy in the search field and elsewhere untouched.
Co-authored-by: itning <[email protected]>
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
The new-task composer only offered the slash commands the agent sent in its last session, stored once globally, so a project that hadn't run a session (or whose last session ran elsewhere) showed none of its skills; matching was substring-only, so `/trig` didn't find `/triage` either.
Offered commands are now remembered per project root. For Claude Code, skills (`.claude/skills/*/SKILL.md`) and commands (`.claude/commands/**/*.md`) are also discovered on disk, from the working directory up to home plus `~/.claude`, and merged with the agent's list (agent entries win). New `agent_commands` IPC serves the home composer (project root) and session composer (the task's cwd). Matching ranks prefix (also after `plugin:`), then substring, then subsequence. Picking a completion inserts `/name ` and the prompt is sent unchanged.
Not covered: a new worktree of a project that gitignores `.claude/` doesn't contain the project's skills.
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
Spring MVC/WebFlux mapping methods (Java and Kotlin) in @RestController / @Controller classes get a gutter icon with "Generate Request in HTTP Client", and the class line gets "Generate Requests for Controller"; both are also actions (`http.generateSpringRequest`, `http.generateSpringControllerRequests`) in Find Action, the editor menu and ⌥↵. Requests are appended to the `generated-requests.http` scratch and opened at the new request.
The URL combines class and method paths ({id} and {id:regex} → {{id}}), with host, port and context path from application.properties / .yml (default http://localhost:8080). @RequestParam, @RequestHeader, params/headers, consumes/produces and @RequestBody (JSON skeleton from the body type, nested one level) are mapped; same-file constants are resolved. Adds an `intentionProviders` editor facet so features can contribute ⌥↵ items.
Fixes#100
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
On Windows the auto-downloaded jdtls ran through `bin\jdtls.bat`, which starts the Python wrapper and ends with `pause`. When Python failed, `pause` wrote its prompt to stdout in the console code page; the transport treated that as a broken LSP header ("non-UTF-8 header"), dropped it and the server was killed, so the status only said "exited with code 1".
- jdtls is now launched with `java` directly on every platform (Equinox launcher, platform config dir, JDK 24+ flags, forwarded JVM args, `-data`), so Python is no longer needed; user wrappers and unusual layouts keep their command.
- Output outside the protocol on stdout (and non-UTF-8 stderr) is decoded (UTF-8, then the Windows OEM/ANSI code page, then lossy), logged and included in the crash message, for stdio and TCP language servers, DAP adapters and ACP agents.
- Crash messages carry the real exit code and the last lines of output.
未在 Windows 实机验证 (Windows paths covered by unit tests and a Windows-target clippy run).
Fixes#90
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
Search Everywhere's query input carries role="combobox", so the global
focus-ring rule drew an outline inside its frame. Text inputs are now
excluded from that rule, and the selector list uses :where() so its
specificity matches a bare :focus-visible and component focus styles
always take precedence.
- Deployment: reject `..` and `/` in remote names and sync paths, keep
excluded paths root-relative, never delete a mapping's remote root,
apply excludes to Sync steps
- SSH: connect per configuration so an unreachable host doesn't block
others; `--` before the host in ssh arguments; read timeout on askpass
- Run targets: export the environment for shell command lines; map the
project folder only as a whole path
- Remote development: use the server's resolved project root; give
streaming channels connection-wide ids; per-install `.part` files;
probe output between markers; require the server's SHA-256; unique
download staging folders; stop runs, terminals and language servers
when the server exits
- JSON-RPC: deliver a response after the notifications sent before it
- Release: build the aarch64 musl server with musl-gcc
About 270 UI strings were JS template literals with interpolation passed to exact catalog lookup, so they could never be translated; about 25 more spliced lowercase English fragments into sentences.
`tx` now also works as a template tag (`` tx`Remove data source '${name}'?` ``): it looks up the entry by the literal parts (`{name}` placeholders), translates interpolated values that have entries, and remembers generated text so it re-translates after a language switch or when the catalog loads later. Call sites use the tag; spliced fragments are written as whole sentences per variant. Adds 280 Chinese translations and removes stale ones.
The catalog guard now also fails on untagged template literals reaching exact lookup (unless allowlisted with a reason), tagged text missing from the catalog, case-changed or ternary-suffix interpolations, and duplicate template patterns.
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
In long files several distant edits could collapse into one large replacement that highlighted unchanged lines, and multi-line changes showed per-line stripes.
Lines are now matched before character-level changes. The line diff is anchored on lines that occur once on both sides (patience-style), with a bounded diff only between anchors, so 50k-line files with distant edits keep separate chunks. Whole inserted or deleted lines render as continuous blocks; paired modified lines keep the line background plus word-level highlights, also inside multi-line chunks. Line decorations are built for the visible range only.
Co-authored-by: zzxzz12345 <[email protected]>
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
Adds Solarized Light, Solarized Dark, One Light and One Dark as built-in themes across the workbench, editor, terminal and Mermaid, keeping the original palettes (MIT notices in THIRD_PARTY_LICENSES.md, mapping in docs/themes.md).
- Native window colours are read from tokens.css at build time (`theme_frame.rs`) instead of a second colour table.
- Startup restores the built-in palette before first paint.
- Editor schemes, the terminal and Mermaid refresh on switches between themes of the same kind.
- The plugin packages of the same themes are removed.
- A test requires every built-in palette theme to define each token (only listed ones may inherit from the base theme), so new tokens can't silently miss them.
Co-authored-by: Claude <[email protected]>
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
Maven sync ran a plain `dependency:tree`, which drops the nodes Maven omits (conflict losers, duplicates), so there was no data to show where a conflict comes from.
Sync now runs `maven-dependency-plugin:3.8.1:tree -Dverbose` (the project's own plugin version when offline) and records omitted / conflictWith / managedFrom per node; omitted nodes stay out of the Maven tool window and External Libraries, and older caches are marked stale. The new Dependency Analyzer (Maven toolbar, module / Dependencies node menus, pom.xml menus, Find Action) lists a module's resolved artifacts or tree with conflict markers, search and scope filters, and every path to the selected artifact with its status. Exclude writes an `<exclusion>` into the declaring pom (module or in-project parent), keeping indentation and line endings, then reloads; Jump to Source and Copy Coordinates are on each path. Gradle is not covered yet.
Fixes#87
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
Adds "Expand Directories with Single Click" to the Project view options (Behavior group, off by default). When on, a plain left click anywhere on a directory row expands or collapses it (a compacted path opens its whole chain); a double-click toggles once, and modifier clicks, drag-and-drop, inline rename and the context menu are unaffected. Project view option toggles now show their checked state in Find Action, and the options menu is translated.
Fixes#94
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
The Solarized dark variant's description no longer appears in the source, so the
catalog check failed on its leftover translation. LaunchError and Launcher are
only used by the unix-only tcp_server tests and were unused imports on Windows.
The source-text catalog (locales/text/zh.json) is keyed by the English text, so reworded or new UI strings silently fell back to English (e.g. the language server JDK hint on the Java settings page). Adds the 321 missing Chinese translations, updates the 3 reworded keys, removes 45 stale keys, and makes the file-refactoring failure titles translatable.
New guard test (src/platform/locales/textCatalog.test.ts) parses src/ for literals that reach the translation layer and fails on missing or stale catalog entries, with an explicit allowlist for untranslated text.
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
Solarized Light/Dark and One Light/Dark ship as theme plugins (plugins/theme-solarized, plugins/theme-one) with complete UI tokens, editor schemes and terminal colours, and the original palettes' MIT notices. Installing a plugin now also copies its LICENSE/NOTICE/COPYING.
Theme handling that benefits every plugin theme:
- The last plugin theme is restored before first paint (splash, workbench); the native window background follows the theme.
- HSL / color-mix() / var() colours are resolved to RGB for native windows, the Settings window and Mermaid.
- Mermaid, the terminal and editor schemes follow switches between two themes of the same kind.
- An explicitly chosen editor scheme wins over a plugin theme's editor colours.
Supersedes #85.
Co-authored-by: bghtyu <[email protected]>
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
@zed-industries/codex-acp is archived; its last release (0.16.0) bundles an
older Codex core that doesn't know newer models, so a config.toml model such as
gpt-6.1-sol fell back to generic metadata with a warning. The maintained adapter,
@agentclientprotocol/codex-acp, runs the Codex App Server from a current
@openai/codex and lists the account's models.
SSH (new crate ide-ssh)
- Shared SSH configurations (Settings › Tools › SSH Configurations) used by
data source tunnels, deployment servers, run targets and terminals.
- Sessions with password, key pair or OpenSSH config + agent authentication,
ProxyJump, known-hosts checks with trust prompts, saved secrets.
- Askpass: credential prompts of `ssh` / `git` child processes shown in the IDE.
- Tools › Start SSH Session opens a terminal tab for a configuration.
- The database SSH tunnel moves here and can use a shared configuration.
Deployment (new crate ide-remote)
- Settings › Build › Deployment: SFTP servers, path mappings, excludes,
automatic upload on save.
- Upload / download / delete / sync with the server, Tools › Deployment
menu, Remote Host tool window for browsing and editing remote files.
Run targets
- Settings › Build › Run Targets: run configurations can run on an SSH host
("Run on"), with project sync before the run and remote interpreters.
Remote development
- New crate ide-host: the backend (projects, documents, LSP, VCS, search,
terminals, runs, debugger, tools, settings) shared by the desktop shell and
`openidea-server`, a headless server speaking JSON-RPC over stdio.
- File › Remote Development… connects to an SSH host, installs the server
(per-build folder, streamed gzip upload; the release build is downloaded
when there is none locally) and opens a remote folder in the window.
- The window routes the commands the server answers to it; streaming
channels and events are forwarded. The server takes the login shell's PATH.
Reference-based features only asked the language server, which analyses its configured scope (e.g. Pyright's `include`) plus open files, so usages in unopened files outside that scope were silently missing.
Find Usages, Show Usages, usage counts, Safe Delete of a symbol, Rename and incoming calls now also search project files that contain the symbol's name (whole word, same language, .gitignore/excludes respected, unsaved edits included, up to 300 files, cancellable). Candidates the first answer didn't cover are opened in the server temporarily, queried and closed again. Files the user has open are never touched, and diagnostics for the temporary files are dropped. An empty result while the server is still indexing is retried once indexing ends. When results may still be incomplete (indexing, file cap), the usages UI says so. `lsp_references` takes a RequestId and is cancelled by newer searches. Not covered yet: the Safe Delete scan when deleting a file or directory.
Fixes#83
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
The splash set its wordmark in Inter, a web font with font-display: swap. When
the font arrived late (on a cold dev start the dev server took up to seconds to
serve it), the splash showed the system font first and switched to Inter just
before fading, so the word grew and shifted. It is now an inline SVG of the same
text in Inter 500, so it renders identically from the first frame. This replaces
the previous attempt, which only moved where the font faces load.
An app started from Finder or the Dock took only PATH from the login shell, so
variables exported in rc files never reached agents, language servers, run
configurations or debug targets. Claude Code configured for Bedrock through
CLAUDE_CODE_USE_BEDROCK / AWS_REGION in ~/.zshrc fell back to the Anthropic API
and failed with 403.
The shell's environment is read once at startup (in the background when PATH
comes from the cache) and kept in memory only, since it may hold secrets. Spawns
add the variables this process lacks before their own.
- Enter that commits an input-method candidate no longer sends the message.
- The composer turns off autocorrect, autocapitalisation and spellcheck, which
popped up correction bubbles over prompts.
- The provider icon in the task header sits inline with the details instead of
on its own line above them.
WebKit sends the Enter (or Space) that commits a candidate after compositionend,
with isComposing already false and keyCode 229. Handlers that checked isComposing
alone ran it: dialogs confirmed and shortcuts fired while typing Chinese or
Japanese. isImeKey() checks both, and every keydown handler that skipped
composition uses it.
The splash named Inter, but the font faces and antialiasing came with the app CSS,
which the dev server injects only once the bundle runs: the wordmark switched from
the system font to Inter, moving up and changing weight, just before the fade. The
font faces are now linked from index.html and the splash sets its own smoothing.
The splash also stays on one compositing layer, so the fade-out doesn't re-snap
the centred text.
The global focus ring listed input and textarea. :is() takes its most specific
argument, so the rule tied with the components' own :focus-visible resets and won
whenever it loaded later: the Settings search field and the agent composer showed
an inner square box inside their focus frame. Text fields show focus with their
frame; the ring stays on checkboxes, radios and other input controls.
Adds Increase Font Size / Decrease Font Size (⌥⇧. / ⌥⇧, in the IntelliJ keymaps) for the active editor, sharing the ±1px step and limits with ⌘/Ctrl + mouse wheel. The Markdown preview, including code blocks, follows its editor's zoom, and Reset Font Size in All Editors resets it too.
Editor zoom had no effect once a font size was configured in settings: the configured size was applied with higher precedence than the zoom. The zoom now wins.
Co-authored-by: zzxzz12345 <[email protected]>
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
In projects with several repositories the Branches popup began with a wrapping row of repository chips and listed only the selected repository's branches, and the widget always read "repo / branch".
The popup now lists one row per repository (colour, name, current branch) whose submenu holds that repository's branches. When all repositories are on the same branch the widget shows just the branch and the popup offers Common Local / Remote Branches whose actions apply to every repository; otherwise it shows the current repository's Recent / Local / Remote Branches and Tags. The widget and the popup use the same current repository: the one owning the active file, else the selected one. Search covers actions, repositories and branches across repositories. New Branch… creates the branch in all repositories. Single-repository projects are unchanged. `vcs_roots` now returns each repository's branch and HEAD.
Fixes#78
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
Fenced ```mermaid blocks in the Markdown preview render as diagrams. mermaid (12.1.0) is loaded lazily, only when a preview contains a diagram, with securityLevel "strict"; colours come from the theme tokens and diagrams re-render on theme changes. Syntax errors show inline under the source. Rendered diagrams are cached by theme + source, so editing elsewhere in the file doesn't redraw them.
Fixes#79
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
Some language servers (basedpyright) send inferred type hints as plain text without a label-part location, so ⌘/Ctrl-clicking `: Series` did nothing. For a type hint right after an identifier, ⌘/Ctrl-click now goes to that identifier's type declaration; hints with locations keep direct navigation. Navigable hints show the link underline and pointer only while ⌘/Ctrl is held, including when it is pressed before hovering.
Co-authored-by: zzxzz12345 <[email protected]>
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
Keymaps only understood keystrokes, so the mouse back / forward side buttons did nothing. Keymaps now accept mouse shortcuts (Button4, Shift+Button5, …); every preset binds Button4 / Button5 to Navigate Back / Forward. The dispatcher handles side-button presses in the capture phase, so they work over the editor, tool windows and empty space, never change the editor selection and don't trigger the webview's own history navigation. Menus and tooltips keep showing the keyboard shortcut.
Fixes#75
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
Adds GDScript, Godot resource and shader file types with tree-sitter highlighting, the GDScript language server over TCP (a headless Godot editor started on demand), "Godot Project" / current-scene run configurations, and debugging through Godot's debug adapter.
Breakpoints set through symlinked paths (/tmp, /var on macOS) were reported verified but never hit; the adapter now sends resolved paths and maps stack frames back to the paths the IDE opened. The duplicate exit event Godot sends after disconnect is ignored.
Fixes#68
Claude-Session: https://claude.ai/code/session_01VUYsdXA2nkF5v6zsC93dpn
## What
An Agents feature: hand a task to a coding agent and follow it from the IDE. Agents speak the [Agent Client Protocol](https://agentclientprotocol.com). Built in: Claude Code, Codex, Gemini CLI, GitHub Copilot, Cursor, OpenCode, Qwen Code, Kimi CLI, Goose, Auggie, Kiro, Mistral Vibe. You can add your own in `<config>/agents/<id>.toml` (`name`, `command`, `args`, `[env]`). An agent whose command isn't on `PATH` is listed under "Not installed".
- **New Task page.** An editor page with a prompt box. You pick the agent, model, mode and thinking effort, and whether it runs in a new worktree (from a chosen branch) or in the checkout.
- **Agents tool window.** Tasks are grouped by what they need: Needs Input, Failed, Running, Ready to Review, Archived. It also shows a stripe badge, a status bar item and notifications.
- **Session tab.**
- The conversation: Markdown, thinking, tool calls with diff stats, the plan.
- Permission requests appear as an editor banner.
- The changed files are compared against the task's base commit and open in the diff viewer.
- A context-usage meter.
- **Composer.**
- `@` references files from the project index. A referenced file's text is sent when the agent accepts embedded context, otherwise a `resource_link`.
- `/` lists the agent's commands; ↑/↓ recall earlier messages.
- Images can be pasted, dropped or picked; Shift+Tab cycles the mode.
- While the agent works, Enter queues the message and ⌘↵ interrupts the turn and sends it now. Queued messages can be sent now, edited or removed.
- **Send to Agent.** In the editor and Project view context menus. It adds `@file:10-20` references to the message you typed last.
- Landing the work is left to the existing Git features: the worktree is a normal branch.
## How
- **`crates/ide-agent`**: one ACP agent process per task, running in the task's directory.
- It folds `session/update` into a timeline and answers `session/request_permission`.
- It tracks the session's settings: `configOptions`, falling back to the older `modes` / `models` fields. The user's choices are applied again on every new or reloaded session.
- Tasks are saved as JSON per task. Sessions resume with `session/load` without duplicating history, and old task files are upgraded when read.
- **`ide-jsonrpc`**: adds newline-delimited framing (ACP) next to Content-Length (LSP).
- **`ide-vcs`**: adds `numstat_with_worktree` for the changes list.
- **Editor**:
- The Switcher (⌃⇥) lists every open editor, as IDEA does, not just files.
- `EditorKindContribution.restorable` lets a kind's tabs be saved with the editor session, so agent tabs survive reloads.
- **UI**: `Select` popups now size to fit short option details (they used to cut them off).
## Testing
- `cargo test -p ide-agent` (14 tests). They use an in-memory ACP agent and cover permissions, restart with `session/load`, settings, the queue with send-now and stop, and prompt blocks.
- There is also an ignored end-to-end test against the real Claude Code adapter: `OPENIDEA_AGENT_E2E=1 cargo test -p ide-agent real_claude -- --ignored`.
- `pnpm vitest run`: 2052 tests.
- `cargo clippy -p ide-agent -p openidea --all-targets -- -D warnings` and typecheck are clean.
- Driven in the running app with Claude Code:
- A task in a worktree, with permission answers, changes listed and the diff opened.
- A follow-up after restarting the IDE.
- `@` reference with embedded content, `/` commands, Shift+Tab mode changes.
- A queued message and send-now, an image prompt, Send to Agent from a selection.
- The Switcher and restored tabs after a reload.
Not covered yet: only Claude Code was driven end to end. The other agents go through the same ACP path but weren't run here.
Merged on the maintainer's go-ahead. Merge with main: kept both sides in Cargo.toml (ide-agent + ide-clipboard), the feature registry (agents + vim) and zh.json (no duplicate keys). After the merge: typecheck clean, full vitest 268 files / 2247 tests passed, ide-agent/ide-jsonrpc/ide-vcs 144 passed, bindings unchanged after regeneration.
l0ng-ai
itninganditning
胡翀andzzxzz12345
Walter WangandClaude
