QueuePicker's raw-SQL Queued->Running claim bypasses TaskStateService.StartRunningAsync,
the only other place that broadcasts TaskUpdated on this transition, so the task-list
badge stayed on "Queued" until the run finished. Send TaskUpdated for alreadyClaimed
dispatches in TaskRunner.RunAsync/ContinueAsync to close that gap.
Adds a read-only get_effective_run_config(taskId) tool that reports the
model/max-turns/effort/permission-mode/agent-path/system-prompt/skills a
task will actually run with, each tagged with its source (task/list/
preset/global), plus max-turns' raw requested value and clamp status.
Extracted the model/max-turns/agent-path resolution out of
TaskRunner.ResolveConfigAsync into EffectiveRunConfigResolver so the run
path and the new reporting tool share one codepath instead of risking
drift, per docs/explore-notes/worker-task-pipeline.md's max-turns trap.
The cache-token task added an ITranscriptUsageReader ctor param to TaskRunner
while the failure-diagnosis task added this test file. Both branches were green
in isolation; the merged result did not compile.
StreamAnalyzer now reads subtype/terminal_reason/errors from the CLI's result
event, and TaskRunner builds a specific ErrorMarkdown from them: max_turns names
the turn budget and points at set_task_config + requeue, api_error passes
through the provider's own message (which carries the reset time), and any
other terminal_reason is appended to the previous generic text instead of
staying invisible. Falls through to the old "exited with code N and no result"
text when there's no terminal_reason at all (a real crash).
task_runs.tokens_in only ever held the API's uncached "input_tokens" field
(off by a factor of ~400,000 on a resumed session), and tokens_out summed
only the last result event instead of the whole session. TaskRunner now
reads each run's cache-read/cache-write/input/output totals from the
session transcript via a new ITranscriptUsageReader.ReadSessionTotalsAsync,
storing the delta against prior runs on the same session so a --resume
doesn't double-count. New task_runs.cache_read_tokens/cache_write_tokens
columns; the Session tab now shows the raw total (what actually counts
against the 5h/7d usage limit) with a breakdown tooltip.
Runaway sessions were the single biggest cost driver: model_presets was
never persisted (stayed code-only), default_max_turns shipped at 100, and
ResolveMaxTurns had no upper bound, so a task/list override could run
hundreds of turns unchecked.
- TaskRunner.ResolveMaxTurns now clamps the resolved value to
AppSettings.MaxTurnsCeiling (new column, default 80) and logs a warning
with task id / requested / effective value when it clamps.
- default_max_turns default lowered from 100 to 40 (entity, EF config,
and the seeded row via the new AddMaxTurnsCeiling migration).
- AppSettingsRepository.GetAsync backfills model_presets with the
shipping defaults on first read instead of leaving the column null.
- Settings > General's per-model preset table and the task/list agent
editor now show a hint when a set max-turns value exceeds the ceiling.
Adds GetUsageSnapshot/GetModelUsage/GetTaskUsage to WorkerHub (backed by a
shared UsageSnapshotBuilder), a UsageUpdated broadcast fired after every
UsageMonitorService poll cycle, and records the resolved model on each
task_runs row so per-model/per-task usage can be reported from history.
Persists the merge commit SHA on WorktreeEntity for every successful
single-task and unit merge, and adds a TaskMergeService.RevertMergeAsync
+ revert_merge MCP tool that undoes a merged task's merge via
`git revert -m 1` (never reset/rewrite, since the target checkout is
shared). Rejects cleanly when there's no recorded SHA, the repo is
mid-merge/mid-revert, or the target has foreign uncommitted changes;
a conflicting revert aborts immediately. Also exposes the new
mergeCommit field via get_task_worktree.
BuildForTaskAsync's fresh-session path flattened the task title+description
into one positional CLI argument, which the ConPTY host joins into a single
command line and claude re-splits on whitespace -- any dash-leading token in
the description (e.g. "->", "--abort") was misread as an unknown option, and
a raw multi-line prompt truncated at its first newline regardless. Now the
brief is written to ~/.todo-app/task-sessions/<taskId>/brief.md and exposed
via --add-dir, with a single-line kickoff pointing claude at it -- the same
pattern BuildForMergeHelperAsync and the planning launcher already use.
Phase 0 forced a batch_get_tasks full-fetch across every task just to see
descriptions, which blew past the client's token limit on larger lists.
brief.md lives on disk and has no such limit, so descriptions now render
there directly (fenced with an extended backtick run, indented under the
list bullet, so embedded headings/lists/code fences can't break the task
list's own structure). Phase 0 now treats the brief as the primary source
and only falls back to batch_get_tasks for fields it doesn't carry.
"Let Claude handle it" now creates one ClaudeDo task per run to host the
ConPTY session (Idle/IsManual, never queued) instead of an untracked
ad-hoc tile, so the run has a real title, diff, and review outcome.
Since the handler merges its own changes straight into the list's
working dir, the task never gets a WorktreeEntity; its review range
lives as new HandlerBaseCommit/HandlerHeadCommit columns on TaskEntity
instead, reusing the existing commit-range diff machinery and keeping
it out of the worktrees overview entirely.
Two context-menu entries opened a Claude session for the same task via different
mechanisms (embedded ConPTY vs. an external wt terminal). Drop the external-terminal
path entirely, including its worker hub method, launcher plumbing, and localization
keys, since the embedded ConPTY session already covers every case it did.
A full model id like claude-sonnet-4-6 in list_config.model never matched any
ModelPresets row (only bare aliases did), so every run under it landed on the
hardcoded 30-turn/DefaultLevel fallback instead of the sonnet preset row -
this is what killed two autonomous tasks at the 30-turn limit.
ModelRegistry.TryNormalizeAlias (non-throwing: exact match, then substring
match against a full model id) lets ModelPresets.For resolve a full model id
to its alias's preset row. For a model that still doesn't resolve, the
hardcoded 30 is replaced by a caller-supplied fallbackMaxTurns, and TaskRunner
now passes AppSettings.DefaultMaxTurns there - so that setting has a real
effect instead of being dead, matching the direction already noted in
docs/open.md.
## Problem
Die "Let Claude handle it"-Session (List-Handler) startet mit `--permission-mode default` und fragt
dadurch bei jedem Tool-Aufruf nach Bestätigung. Sie soll autonom durchlaufen können — der User
überwacht die ConPTY-Kachel, statt jede Aktion einzeln freizugeben.
## Ist-Zustand
`src/ClaudeDo.Worker/Runner/InteractiveLaunchSpecService.cs`, `BuildForMergeHelperAsync`
(Zeile ~183-249). Die
ClaudeDo-Task: 50a6027eba294dde8b18e4082dfc1e9b
ClaudeDo never passed --effort, so every session inherited whatever effortLevel
the user's Claude Code config happened to carry. Settings -> General now holds one
row per model alias (haiku medium/20, sonnet high/30, opus high/40, fable high/25)
supplying the global effort and turn defaults; list- and task-level max-turns
overrides still win, and the agent editor's inherited badge follows the model.
--effort is applied to autonomous runs and to every ConPTY spec (task session,
planning start/resume, ad-hoc, list handler). The model itself is deliberately not
forced on interactive sessions. The single global 'Max turns' field is replaced by
the table, and 'fable' joins ModelRegistry.Aliases.
The migration also adds the is_manual columns used by the next commit.
Interactive planning sessions launched with --permission-mode plan, which
gates EVERY MCP tool call regardless of --allowedTools (verified: even a
read-only mcp__claudedo__list_task_lists is denied under plan mode). So the
session prompted the user on the first CreateChildTask -- the whole point of
a planning session.
Switch BuildPlanningStartArgs/BuildPlanningResumeArgs to --permission-mode
default, which honours the allowlist. File edits stay blocked via the planning
system prompt + AllowedTools omitting Write/Edit/Bash. Resume also re-passes
--allowedTools, since the CLI does not restore it across --resume.
The earlier 'glob does not match' hypothesis was empirically falsified:
mcp__claudedo__*, the bare server name, and the explicit tool name all allow
the tool with zero permission_denials in default mode.
Planning start/resume opened an external Windows Terminal (wt) window.
Route them through the embedded ConPTY Command Center pane instead, matching
the existing interactive-session UX (no external window).
- Extract bare planning arg builders (BuildPlanningStartArgs/ResumeArgs) from
WindowsTerminalLauncher; the wt path still uses them (kept, not removed).
- InteractiveLaunchSpecService.BuildPlanningStart/Resume map a planning
context into a LaunchSpec (planning args + env: MAX_THINKING_TOKENS,
CLAUDEDO_PLANNING_TOKEN). Hub GetPlanningStart/ResumeLaunchSpec run
StartAsync/ResumeAsync then return the spec.
- UI: OpenPlanningSession + the resume branch raise OpenPlanningConPtyRequested;
the shell opens Mission Control and hosts a planning ConPTY pane.
Env is process-global by design (sequential human-paced sessions). wt planning
code retained. Tests added for the arg/env mapping.
HasChangesAsync counted untracked files, so a stray file in the shared
target working dir (e.g. left by a concurrent session) blocked merge
preflights even though nothing tracked changed. Add an includeUntracked
overload defaulting to true, and pass includeUntracked: false only from
the two target-working-tree merge preflights (TaskMergeService.MergeAsync,
PlanningMergeOrchestrator.StartAsync). Auto-commit and the worktree
cleanup data-loss guard keep counting untracked files, since those
callers need to know about them.
The embedded ConPTY terminal replaced the in-app streaming interactive session, so
delete the dead stack: StreamingClaudeSession, InteractiveSessionService,
ProcessClaudeStreamTransport, IClaudeStreamTransport, ILiveSession, LiveSessionRegistry,
IdleSessionReaper (+ WorkerConfig.InteractiveIdleTimeoutMinutes), the WorkerHub
interactive methods + HubBroadcaster events, IWorkerClient interactive members, the
TaskMonitorViewModel composer + SessionTerminalView composer markup, and the old
'Run interactively' entry. AskUser/PendingQuestionRegistry, the autonomous path,
planning, ResumeTaskInTerminal, and all ConPTY code are kept. Localization pruned.
A fresh (non-resume) task-based ConPTY session now opens claude on the task's
prompt (title + description) as the positional argument, so the session starts on
the task instead of an empty prompt. Resume sessions and ad-hoc sessions are
unchanged.
BuildForTaskAsync now creates a worktree on demand (via WorktreeManager.CreateAsync,
the same path TaskRunner uses) when a task has a configured working dir but no
Active/Kept worktree, returning a fresh-start spec -- so never-run tasks can be
opened interactively. Adds BuildForDirectoryAsync + GetAdHocLaunchSpec hub/client
for ad-hoc sessions in an arbitrary directory (no task, no worktree, no skill seeding).
Adds InteractiveLaunchSpecService + GetInteractiveLaunchSpec hub method that
prepares a task worktree (session-skills seeding, run env) and returns a
LaunchSpec {cwd,exe,args,env} for a UI-hosted ConPTY terminal. Reuses
ISessionSkillSeeder, TaskRunner.UnionSkillNames, and WindowsTerminalLauncher
resume-arg/resolve logic. Guards mirror ResumeTaskInTerminal; a never-run task
yields a fresh-start spec instead of an error.
Interactive/streaming sessions are persistent claude.exe processes that
wait on stdin and never exit on their own. The only teardown was an
explicit StopInteractiveSession from the UI — there is no client-disconnect
or shutdown sweep — so an abandoned chat (UI closed, navigated away,
crashed) kept its claude.exe (+ conhost) alive for the worker's whole
lifetime. Under a long-running autostart worker these accumulate to dozens
of orphaned child processes.
LiveSessionRegistry now tracks per-session activity (Touch on every output
line and user action) and exposes ReapIdleAsync, which stops sessions idle
past a timeout while skipping any with a turn in flight. IdleSessionReaper
(BackgroundService) sweeps every 5 min; idle timeout defaults to 30 min,
configurable via interactive_idle_timeout_minutes (0 disables).
StreamingClaudeSession.RemoveQueuedAsync drops the first occurrence of a queued
message from _pending and re-broadcasts the updated queue. Wired through
InteractiveSessionService + WorkerHub.RemoveQueuedInteractiveMessage +
IWorkerClient.RemoveQueuedInteractiveMessageAsync. Removal by text (first match)
is robust to a turn flushing mid-click. Fakes + ILiveSession impls updated.
StreamingClaudeSession raises onQueueChanged (pending snapshot) and onUserMessageSent
(on delivery, incl. the seeded first prompt); InteractiveSessionService forwards these
as InteractiveQueueChanged/InteractiveMessageSent broadcasts. Lets the UI show queued
messages above the input and move a message into the transcript only when actually
delivered to Claude. Client events + fakes updated.
StreamingClaudeSession now buffers a mid-turn user message in a FIFO queue and
flushes one when the turn's result arrives (no implicit interrupt). InterruptAsync
only writes the control_request (no-op when idle); the resulting turn-end then
flushes any queued message. New InteractiveSessionService.InterruptAsync +
WorkerHub.InterruptInteractiveSession + IWorkerClient.InterruptInteractiveSessionAsync.
StreamingClaudeSession drives claude --input-format stream-json over a kept-
open stdin: sends user messages, interrupts the in-flight turn via the verified
control_request protocol, and tracks turn state from result events (treating an
interrupt-aborted error_during_execution result as turn-ended). IClaudeStreamTransport
abstracts the process I/O so it is unit-tested with a fake (no real claude).
LiveSessionRegistry maps taskId -> live session for the hub to route into.
Backs the upcoming in-app interactive sessions; autonomous task execution untouched.
A running task can call mcp__claudedo_run__AskUser(question) to block (up to 3
min) on a human answer. PendingQuestionRegistry holds the pending question +
TaskCompletionSource; the tool broadcasts TaskQuestionAsked, awaits the answer
(WorkerHub.AnswerTaskQuestion resolves it), and returns it as the tool result —
or a 'proceed on your judgment' fallback on timeout. The run stays Running
throughout (no status/schema change). ClaudeProcess raises MCP_TOOL_TIMEOUT so
the 60s HTTP-MCP cap doesn't kill the wait; the run MCP is now wired for every
task, not just standalone ones. System prompt updated to reconcile 'unattended'.
TaskRunner appends attached files (absolute paths) to the run prompt as the
read-only Reference files section. Task and list deletes now remove the
on-disk attachment dir eagerly, and a startup AttachmentOrphanRecovery sweep
drops any attachments/<taskId>/ whose task no longer exists (covers list
cascade and planning-discard paths).
The picker claims Queued->Running atomically before dispatch; the new
StartRunningAsync guard then rejected every queue-dispatched run. Add
alreadyClaimed to RunAsync/ContinueAsync (queue passes true, override
slot keeps the guard) and align the routing tests.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Befund (bestätigt): src/ClaudeDo.Worker/Runner/TaskRunner.cs:101 (RunAsync) und :211 (ContinueAsync) ignorieren das TransitionResult von _state.StartRunningAsync. Race-Szenario: Der QueuePicker claimt Queued→Running atomar; ruft der Override-Pfad (RunNow) kurz danach RunAsync für denselben Task auf, schlägt StartRunningAsync fehl (0 rows affected), der Runner startet Claude aber trotzdem → derselb
ClaudeDo-Task: 44f86be2-7f3d-462e-98b3-eb94c0174eea
- Extract FakeClaudeProcess to Infrastructure/FakeClaudeProcess.cs (was
defined inline in QueueServiceTests #region); all consumers updated
- Replace duplicate FakeHubContext/FakeHubClients/FakeClientProxy
(QueueServiceTests) with existing CapturingHubContext from Infrastructure
across all 7 affected files; Planning's file-local FakeHubContext kept
- Rename SeedListWithAgentTag → SeedListAsync (return Task<string>, drop
unused agentTagId tuple element) and SeedListWithAgentTagAsync → SeedListAsync
- PrimeRunnerTests keeps its private nested FakeClaudeProcess: constructor
API (delay/exitCode/lines/result params) differs from the shared one and
replacement would require rewriting every test in that file
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Collapse TryCompleteParentAsync (planning -> Done) and
TryAdvanceImprovementParentAsync (improvement -> WaitingForReview) into one
TryAdvanceParentAsync that surfaces any WaitingForChildren parent for review
once all children are terminal. Planning parents no longer auto-complete.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
1. ArgumentList (fix injection): ClaudeArgsBuilder.Build() now returns
IReadOnlyList<string>; ClaudeProcess populates ProcessStartInfo.ArgumentList
instead of Arguments, so values like system prompts are never shell-split.
DailyPrepPrompt, RefinePrompt, and WeekReportService migrated similarly.
All IClaudeProcess fakes updated.
2. ContinueAsync exception guard: wrap RunOnceAsync in try/catch matching
the RunAsync pattern so an unexpected exception never leaves the task
stuck in Running status.
3. Planning chain cascade: OnChildFinishedAsync now calls CancelAsync on
the immediate blocked successor when a child fails or is cancelled,
triggering a recursive cascade that clears the entire remaining chain
instead of leaving it wedged.
4. FailAsync guard: restrict valid source states to Running and Queued;
WaitingForReview -> Failed is now rejected, preventing an invalid
transition that could corrupt the review workflow.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
mika kuns