releases/ClaudeDo
2
0
Fork 0
Code Issues 1 Pull Requests 2 Actions Packages Projects Releases 24 Wiki Activity

Merge branch 'claudedo/df0e925e1ff4419d8b7a45f8a184f6d0'

Browse Source
This commit is contained in:
mika kuns
2026-08-05 09:26:46 +02:00
parent 524836ffe8 87de53e052
commit 7d29ec0725
21 changed files with 27 additions and 278 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/ClaudeDo.Worker/CLAUDE.md
+2 -2
View File
@@ -12,7 +12,7 @@ Worker/
Worktrees/ — WorktreeMaintenanceService
Agents/ — AgentFileService, DefaultAgentSeeder
Runner/ — TaskRunner + Claude CLI integration; TaskRunMcpService/TaskRunMcpContext/TaskRunTokenRegistry (in-task MCP wired during execution)
Planning/ — PlanningSessionManager, PlanningChainCoordinator, PlanningMcpService, PlanningMergeOrchestrator, PlanningAggregator, PlanningSessionContext/PlanningTokenAuth/PlanningMcpContextAccessor, WindowsTerminalLauncher (ITerminalLauncher) — wt launcher for planning sessions + pick-up-in-terminal
Planning/ — PlanningSessionManager, PlanningChainCoordinator, PlanningMcpService, PlanningMergeOrchestrator, PlanningAggregator, PlanningSessionContext/PlanningTokenAuth/PlanningMcpContextAccessor, WindowsTerminalLauncher (ITerminalLauncher) — wt launcher for planning sessions
Refine/ — RefineRunner + RefinePrompt (hub `RefineTask`; broadcasts RefineStarted/RefineFinished)
External/ — ExternalMcpService + sibling tool classes
Config/ — WorkerConfig
@@ -156,7 +156,7 @@ Each CLI invocation is recorded in the `task_runs` table via `TaskRunRepository`
- Review/merge: `ApproveReview(taskId, targetBranch) -> MergeResultDto` (childless task: merges its worktree then Done, conflict stays WaitingForReview; task with children: drives `PlanningMergeOrchestrator` to merge the whole unit), `ContinuePlanningMerge` / `AbortPlanningMerge` (resolve a unit-merge conflict), `PreviewMerge(taskId, targetBranch) -> MergePreviewDto` (non-destructive mergeability check), `RejectReviewToQueue`, `RejectReviewToIdle`, `CancelReview`, `MergeTask`, `GetMergeTargets`
- Single-task conflict resolver (Layer C): `StartConflictMerge`, `GetMergeConflictDocuments` (segments), `WriteConflictResolution`, `ContinueConflictMerge`, `AbortConflictMerge` (service-level `TaskMergeService.ContinueMergeAsync`/`AbortMergeAsync` keep their names)
- Planning sessions: `StartPlanningSession`, `ResumePlanningSession`, `DiscardPlanningSession`, `FinalizePlanningSession`, `QueuePlanningSubtasks`, `GetPendingDraftCount`, `GetPlanningAggregate` (per-subtask diffs), `BuildPlanningIntegrationBranch` (combined diff)
- Interactive sessions (embedded ConPTY, UI process): `ResumeTaskInTerminal` (pick-up-in-terminal), `GetInteractiveLaunchSpec`, `GetAdHocLaunchSpec`, `GetMergeHelperLaunchSpec`, `CreateMergeHelperTask` (creates the ClaudeDo task that owns a list-handler run — `Idle`/`IsManual=true`, `HandlerBaseCommit` stamped to the list repo's current HEAD via `InteractiveLaunchSpecService.CreateMergeHelperTaskAsync` — called by the UI before it opens the task-based ConPTY tile), `SubmitTaskForReview` (branches on whether the task has a `WorktreeEntity`: with one, commits it and moves on; without one, it's a worktree-less list-handler host task and it just stamps `HandlerHeadCommit` to the list repo's current HEAD — both paths then flip the task Idle/Failed → WaitingForReview). **Every** ConPTY spec that `InteractiveLaunchSpecService` builds leads with `--effort <level>` from the relevant model's preset (task/list model for a task session, `PlanningAlias` for planning, list config for the list handler, global default for ad-hoc). `--model` is deliberately NOT forced on an interactive session — the user can still switch models in the TUI. The list-handler spec (`BuildForMergeHelperAsync`) uses `--permission-mode auto` so it runs unattended; the `--allowedTools` allowlist (`mcp__claudedo__*,Read,Grep,Glob,Edit,Bash,WebFetch,WebSearch,Skill`) remains the security boundary.
- Interactive sessions (embedded ConPTY, UI process): `GetInteractiveLaunchSpec`, `GetAdHocLaunchSpec`, `GetMergeHelperLaunchSpec`, `CreateMergeHelperTask` (creates the ClaudeDo task that owns a list-handler run — `Idle`/`IsManual=true`, `HandlerBaseCommit` stamped to the list repo's current HEAD via `InteractiveLaunchSpecService.CreateMergeHelperTaskAsync` — called by the UI before it opens the task-based ConPTY tile), `SubmitTaskForReview` (branches on whether the task has a `WorktreeEntity`: with one, commits it and moves on; without one, it's a worktree-less list-handler host task and it just stamps `HandlerHeadCommit` to the list repo's current HEAD — both paths then flip the task Idle/Failed → WaitingForReview). **Every** ConPTY spec that `InteractiveLaunchSpecService` builds leads with `--effort <level>` from the relevant model's preset (task/list model for a task session, `PlanningAlias` for planning, list config for the list handler, global default for ad-hoc). `--model` is deliberately NOT forced on an interactive session — the user can still switch models in the TUI. The list-handler spec (`BuildForMergeHelperAsync`) uses `--permission-mode auto` so it runs unattended; the `--allowedTools` allowlist (`mcp__claudedo__*,Read,Grep,Glob,Edit,Bash,WebFetch,WebSearch,Skill`) remains the security boundary.
- Worktrees: `CleanupFinishedWorktrees`, `ResetAllWorktrees`, `GetWorktreesOverview`, `SetWorktreeState`, `ForceRemoveWorktree`
- Agents/settings/lists: `GetAgents`, `RefreshAgents`, `RestoreDefaultAgents`, `GetAppSettings`, `UpdateAppSettings`, `UpdateList`, `UpdateListConfig`, `GetListConfig`, `UpdateTaskAgentSettings`
- Reports/notes/prep: `GetWeekReport`, `GenerateWeekReport`, `GetDailyNotes`, `AddDailyNote`, `UpdateDailyNote`, `DeleteDailyNote`, `RunDailyPrepNow`, `ClearMyDay`, `GetLastPrepLog`, `ListPrimeSchedules`, `UpsertPrimeSchedule`, `DeletePrimeSchedule`
src/ClaudeDo.Worker/Hub/WorkerHub.cs
+2 -33
View File
@@ -639,42 +639,11 @@ public sealed class WorkerHub : Microsoft.AspNetCore.SignalR.Hub
return ctx;
}
// Picks up a task's Claude session in a real terminal window (--resume) so the user can
// drive it by hand. Only for tasks the worker isn't actively running, with a persisted
// session id and a live worktree.
public Task ResumeTaskInTerminal(string taskId) => HubGuard(async () =>
{
await using var ctx = await _dbFactory.CreateDbContextAsync();
var task = await new TaskRepository(ctx).GetByIdAsync(taskId, Context.ConnectionAborted)
?? throw new KeyNotFoundException();
if (task.Status is TaskStatus.Running or TaskStatus.Queued)
throw new InvalidOperationException("Can't pick up a running or queued task — interrupt it first.");
var run = await new TaskRunRepository(ctx).GetLatestByTaskIdAsync(taskId, Context.ConnectionAborted);
if (run?.SessionId is not { Length: > 0 } sessionId)
throw new InvalidOperationException("This task has no resumable Claude session yet.");
var worktree = await new WorktreeRepository(ctx).GetByTaskIdAsync(taskId, Context.ConnectionAborted);
if (worktree is null || worktree.State is not (WorktreeState.Active or WorktreeState.Kept))
throw new InvalidOperationException("This task has no active worktree to resume in.");
if (!Directory.Exists(worktree.Path))
throw new InvalidOperationException("The task's worktree directory no longer exists.");
try
{
await _launcher.LaunchResumeAsync(worktree.Path, sessionId, Context.ConnectionAborted);
}
catch (TerminalLaunchException ex)
{
throw new InvalidOperationException(ex.Message);
}
});
// Builds the launch spec an embedded ConPTY terminal (UI process) needs to open an
// interactive Claude session in a task's worktree -- same worktree prep as an
// autonomous run (session-skills seeding, run env vars), --resume if the task has a
// persisted session or a fresh-start spec otherwise. Guards mirror ResumeTaskInTerminal.
// persisted session or a fresh-start spec otherwise. Guards: no running/queued task,
// and (once a worktree exists) it must be live on disk.
public Task<LaunchSpec> GetInteractiveLaunchSpec(string taskId) => HubGuard(() =>
{
if (_interactiveLaunchSpec is null)
src/ClaudeDo.Worker/Planning/Interfaces/ITerminalLauncher.cs
-4
View File
@@ -7,10 +7,6 @@ public interface ITerminalLauncher
{
Task LaunchPlanningStartAsync(PlanningSessionStartContext ctx, CancellationToken cancellationToken);
Task LaunchPlanningResumeAsync(PlanningSessionResumeContext ctx, CancellationToken cancellationToken);
// Resumes an arbitrary task's Claude session (--resume <id>) in a visible terminal so
// the user can pick up the conversation by hand — the "pick up in terminal" action.
Task LaunchResumeAsync(string workingDir, string claudeSessionId, CancellationToken cancellationToken);
}
public sealed class TerminalLaunchException : Exception
src/ClaudeDo.Worker/Planning/WindowsTerminalLauncher.cs
+6 -24
View File
@@ -24,10 +24,12 @@ using ClaudeDo.Data.Models;
namespace ClaudeDo.Worker.Planning;
// Spawns the Claude CLI inside a visible Windows Terminal window. Used for every
// human-driven session: an interactive planning session (start/resume) and the ad-hoc
// "Run interactively" action. Headless task execution does NOT come through here — that
// path is ClaudeProcess (prompt over stdin, no terminal).
// Spawns the Claude CLI inside a visible Windows Terminal window for an interactive
// planning session (start/resume). Headless task execution does NOT come through here —
// that path is ClaudeProcess (prompt over stdin, no terminal) — nor does an embedded
// ConPTY interactive session (IInteractiveLaunchSpecService), which reuses this class's
// arg-building helpers (BuildResumeArgs, BuildPlanningStart/ResumeArgs) for a bare
// Exe/Args pair instead of a wrapped pwsh command line.
public sealed class WindowsTerminalLauncher : ITerminalLauncher
{
private const string AllowedTools = "mcp__claudedo__*,Read,Grep,Glob,WebFetch,WebSearch,Skill";
@@ -84,26 +86,6 @@ public sealed class WindowsTerminalLauncher : ITerminalLauncher
return Task.CompletedTask;
}
public Task LaunchResumeAsync(string workingDir, string claudeSessionId, CancellationToken cancellationToken)
{
if (!Directory.Exists(workingDir))
throw new TerminalLaunchException($"Working directory does not exist: {workingDir}");
var resolvedWt = ResolveWtOrThrow();
var resolvedClaude = ResolveClaudeOrThrow();
var command = BuildResumeCommand(resolvedClaude, claudeSessionId);
StartInWindowsTerminal(resolvedWt, workingDir, command, static _ => { });
return Task.CompletedTask;
}
// Resumes a session by id with no extra flags: the user drives every tool approval in the
// terminal, unlike planning which additionally allowlists its MCP planning tools.
internal static string BuildResumeCommand(string claudePath, string claudeSessionId) =>
BuildPwshCommand(claudePath, BuildResumeArgs(claudeSessionId));
// The raw claude CLI args for a --resume launch, shared with InteractiveLaunchSpecService
// (which needs the bare Exe/Args pair for a ConPTY host, not a wrapped pwsh command line).
internal static IReadOnlyList<string> BuildResumeArgs(string claudeSessionId) =>
src/ClaudeDo.Worker/Runner/InteractiveLaunchSpecService.cs
+3 -2
View File
@@ -15,8 +15,9 @@ namespace ClaudeDo.Worker.Runner;
// session-skills seeded onto disk (reuses ISessionSkillSeeder + TaskRunner.UnionSkillNames,
// exactly like TaskRunner.RunAsync/ContinueAsync) and the same run environment variables
// (reuses ClaudeProcess's MCP_TOOL_TIMEOUT). Exe/Args reuse WindowsTerminalLauncher's
// --resume argument construction. Guards mirror WorkerHub.ResumeTaskInTerminal, except a
// never-run task (no persisted SessionId) is not an error here -- it's a fresh-start spec.
// --resume argument construction. Guards: no running/queued task, and (once a worktree
// exists) it must be live on disk -- but a never-run task (no persisted SessionId) is not
// an error here, it's a fresh-start spec.
//
// The run-scoped "claudedo_run" MCP server (AskUser/SuggestImprovement) that TaskRunner
// wires per headless run is intentionally NOT reused: it exists so an unattended run can