- Fixed indentation and formatting across multiple files for consistency.
- Reformatted code blocks to use tabs instead of spaces and improved line breaks for readability.
- Added connection timeout configuration to benchmark runner for early abort on no events.
- Implemented two-phase timeout strategy in benchmark runner with connection and activity phases.
- Migrated react-edit-benchmark package to typescript-edit-benchmark with pi-mono source repository.
- Added InProcessClient implementation to eliminate subprocess spawning overhead in benchmark runs.
- Extended benchmark configuration with chunk edit variant, retry limits, and conversation dump support.
- Refactored runner.ts to support both RPC and in-process client modes with improved error telemetry.
- Cleaned up 129 benchmark report files from react-edit-benchmark/runs directory.
- Added LSP diagnostic versioning support to track document versions and suppress stale diagnostics.
- Added options parameter to `waitForDiagnostics` and `getDiagnosticsForFile` with version filtering capabilities.
- Enabled `versionSupport` in LSP client capabilities to receive diagnostic version information from servers.
- Added test coverage for stale diagnostic suppression scenarios in LSP diagnostic freshness.
The agent had no way to infer the mcp:// prefix needed to read MCP
server resources. Attempting <server>://<resource> directly fails.
All other internal URL schemes (skill, rule, memory, agent, etc.) are
already listed in the Internal URLs section of the system prompt.
Added mcp://<resource-uri>. URI resolution: exact resource listings
first, then RFC 6570 URI templates advertised by connected servers.
- Fixed memory leak by cancelling idle compaction timer on event controller disposal.
- Fixed session resumption to preserve last non-empty session when starting fresh.
- Fixed stash detection to use git ref resolution instead of output parsing for reliability.
- Fixed secret obfuscation to deobfuscate restored session messages locally while keeping LLM messages obfuscated.
- Fixed stash pop operation to preserve staged changes with --index flag after task branch merges.
- Changed idle compaction settings from enum to numeric type for flexible configuration.
Ensures exit_plan_mode waits for all previous tools to complete
before checking if the plan file exists. Prevents race condition
when LLM calls exit_plan_mode before write completes.
- Fixed PR checkout tool to resolve worktree paths to canonical form using fs.realpath().
- Refactored test mocks to use vi.spyOn for git module functions instead of inline implementations.
- Updated test setup to pass Settings.isolated() with edit.manageImports enabled to EditTool.
- Added afterEach hooks across test suites to restore mocks after each test execution.
Compaction and tool-output pruning can rewrite the session JSONL in
place, shrinking the file. When stat.size < bytesRead, the cache is
now invalidated and the file is re-read from scratch, preventing the
observer from getting stuck on a stale transcript.
Only advance bytesRead to the last complete newline-terminated line,
not to EOF. A partial trailing record (read during mid-write) stays
unconsumed and gets re-read on the next refresh, preventing permanent
entry loss from the transcript.
- Move lifecycle end event after finalizeSubprocessOutput() so
submit_result({status:'aborted'}) is correctly reflected in the
observer instead of showing completed/failed (P1)
- Reset observer registry on session resume and /new command so
stale subagent sessions from prior conversations are cleared (P2)
- Cache parsed JSONL transcript incrementally: track byte offset,
only read and parse new bytes on each refresh instead of reparsing
the entire file, avoiding render loop blocking on large sessions (P2)
Wire EventBus through task runtime so subagent lifecycle and progress
events propagate to the TUI. Add SessionObserverRegistry to track
active sessions, and a session observer overlay accessible via Ctrl+S
that shows a picker of running subagents and a read-only transcript
viewer that reads the subagent's session JSONL file to display
thinking, text, tool calls, and results.
- Add TASK_SUBAGENT_LIFECYCLE_CHANNEL for start/end events
- Add SubagentProgressPayload.sessionFile for session file tracking
- Pass eventBus through sdk.ts -> main.ts -> InteractiveMode
- SessionObserverRegistry with multi-listener onChange pattern
- Overlay picker preserves selection position across live refreshes
- Viewer renders full transcript from session JSONL (thinking, text,
tool calls with smart arg summaries, inline tool results)
When selecting 'Refine plan' during plan review, the refinement text was
placed into the editor via setText() but not submitted. The user had to
press Enter a second time to actually send the message.
After session.abort() at the start of handleExitPlanModeTool, the main
loop cycles back to getUserInput() which sets onInputCallback before the
hook selector and input can resolve. By the time the user finishes typing
the refinement, onInputCallback is available. Submit the refinement
directly through onInputCallback + startPendingSubmission, matching the
pattern already used by handlePlanModeCommand for initial prompts.
Falls back to editor.setText() if onInputCallback is unexpectedly absent.
Fixes#597
When entering plan mode, the thinking level configured on the plan role
(e.g., 'anthropic/claude-sonnet-4-5:xhigh') was discarded because
#resolveRoleModel() called resolveModelRoleValue() but only returned
.model, dropping the thinking level.
This is the same class of bug that was fixed in cycleRoleModels (which
correctly preserves and applies thinking level from role configs).
#applyPlanModeModel was the remaining unfixed call site.
Changes:
- Refactor #resolveRoleModel to #resolveRoleModelFull returning the
full ResolvedModelRoleValue instead of just Model
- Add resolveRoleModelWithThinking() public method exposing the full
resolution result
- Add optional thinkingLevel param to setModelTemporary() for atomic
model+thinking switches (eliminates two-step set-then-apply pattern)
- Bundle model+thinking into paired state objects in InteractiveMode
(#planModePreviousModelState, #pendingModelSwitch) — eliminates
4 parallel fields that could desync
- Simplify #applyPlanModeModel, flushPendingModelSwitch, #exitPlanMode
to single setModelTemporary calls
- Collapse cycleRoleModels two-step into single setModelTemporary call
- Add 6 tests for resolveRoleModelWithThinking
- Added language-specific import management for TypeScript, JavaScript, Python, Rust, Go, and C/C++ with automatic deduplication and grouping.
- Added `imports` parameter to patch, hashline, and replace edit operations for structured import/include requests with named, default, namespace, and alias support.
- Added `edit.manageImports` setting to control post-edit import management behavior across all edit operations.
- Implemented language-specific import handlers with parsing, merging, and rendering logic for each supported language.
- Integrated import application into EditTool with feature flag support and comprehensive test coverage across all languages.
- Added external editor callback support to extension UI dialogs and hook selector for opening plans in external editor.
- Added `matchesAppExternalEditor()` keybinding matcher to detect external editor shortcuts (ctrl+g) across components.
- Added `trimTrailingNewline` option to `openInEditor()` function to preserve trailing newlines when editing externally.
- Modified plan review flow to re-read plan content from disk before approval to reflect external editor changes.
- Refactored plan preview rendering to cache and reuse UI container instead of recreating children on each render.
- Refactored Promise creation in external editor utility to use `Promise.withResolvers()` for cleaner async handling.
- Migrated GitHub CLI utilities from gh-cli.ts module to utils/git.ts github namespace.
- Consolidated GitHub CLI command execution (available, run, json, text) into git.github namespace with unified error handling.
- Removed gh-cli.ts module and ControlledGit class infrastructure; updated all tool imports and call sites.
- Exported git utilities from main package entry point for extension use.
can1357