ide-jsonrpc handled incoming notifications in order on a dedicated thread
but started each incoming request on its own blocking task right away, so
a request could be handled before a notification that arrived ahead of
it. The mock language server in the ide-lsp tests answered
`textDocument/references` before applying the `didOpen` of a file opened
for a project-wide search (missing results, a search finishing before the
test could cancel it or open the file), and on the client side an ACP
`session/request_permission` could overtake the `session/update` that
announced its tool call (ide-agent polled for up to 200 ms to cover it).
A request now starts once every notification received before it has been
handled; requests still run concurrently. The ide-agent polling is gone.
Tests no longer depend on timing:
- the references tests hold the mock's answer with a gate and cancel or
open the file while it is held, instead of sleeping and measuring;
- the indexing retry ends indexing once the first answer is out;
- the crash/restart test counts Ready/Crashed statuses instead of reading
the last one, which a restart can already have replaced;
- the mock records a notification under the same lock as its effect, and
`eventually` waits up to 30 s;
- the TCP tests hold their port bound (not listening) instead of freeing
it for another test to take, and listen only after the command started.
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
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.
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.
Release builds use the GUI subsystem, so a console child spawned without
CREATE_NO_WINDOW gets its own console window. git, gh, docker, builds and
run configurations already set the flag; language servers, SDK and tool
probes, debug adapters, tool downloads, coverage export and perf did not.
Add ide_core::process::no_window and use it at every remaining spawn.
Fixes#16
l0ng-ai