The soft request budget resolved to `SOFT_REQUEST_BUDGET[agent.name] ??
configured`, so the bundled entries for scout and sonic replaced the
configured value outright. Lowering `task.softRequestBudget` to tighten
the guard therefore did nothing for exactly the two agents that spawn
most often: a scout kept its 100-request budget no matter how small the
user set the knob. Only 0 (disable) and raising the value for
non-bundled agents had any effect.
Treat both numbers as upper bounds and take the smaller one. The bundled
entries stay ceilings, so a runaway scout is still stopped at 100 by
default and existing behavior is unchanged for anyone who has not
lowered the setting; a configured 0 still disables the guard entirely.
Resolution moves into `resolveSoftRequestBudget`, which also normalizes
negative and fractional inputs, so the rule is testable without standing
up a subprocess run.
This composes with `task.maxEffort` on a separate axis: effort caps how
hard each request thinks, this caps how many requests a run may spend.
(cherry picked from commit f0db29f8f725f11390b64ca9342300c482ff5c5d)
- task.maxEffort only clamped the initial thinking level; a retry
fallback candidate could clamp back up to its model floor and run a
low-capped spawn at high.
- The ceiling now rides the session as thinkingLevelCeiling: clamped in
ModelControls (constructor, setThinkingLevel, auto classifier,
restore) and in applyRetryFallbackCandidate; fallback candidates whose
floor exceeds the ceiling are skipped.
- Effort value import moved to @oh-my-pi/pi-catalog/effort; changelog
attribution added.
- Review follow-up for PR #6794.
- Introduce task.enableEffort setting defaulting to false to hide per-spawn effort parameters.
- Conditionally include effort in single and batch task schemas and descriptions based on the new setting.
A crashed owner's pid can be recycled by an unrelated long-lived
process, so kill(pid, 0) succeeds and the leftover sandbox was pinned
live forever, unreachable by a non-`--all` clear.
The ownership marker now records a process-instance start-time token
alongside the pid (Linux /proc/<pid>/stat field 22, other Unix via
`ps -o lstart`). A live pid whose current token no longer matches the
recorded one is a recycled pid and counts as dead; platforms that can't
report a token degrade to the prior pid-only check.
Fixes#6761
`omp worktree clear` (without `--all`) removed every task-isolation dir
under the worktree base, including sandboxes owned by subagents running
right now, and the "no live task owns it" reason was asserted from the
mere presence of the `m` mount dir with no ownership check.
`ensureIsolation` now stamps each sandbox base dir with a pid-bearing
ownership marker before the backend materialises `m`, and the worktree
scanner classifies a sandbox as live while its owning process is alive,
so `clear` reclaims only crashed leftovers.
Fixes#6761
The prewalk arm/switch guard compared model identity only (modelsAreEqual /
provider+id), discarding the resolved thinkingLevel. A legal same-model target
at a cheaper effort (e.g. prewalk: "@task" resolving to the active model at a
lower level) was dropped as a no-op, so the session ran the expensive effort for
the whole run while still paying the plan/continue nudges — silently on the
session path, logger.debug only on the subagent path.
Compare (provider, id, effective thinking level) via a shared prewalkWouldBeNoop
helper. Effort-only deltas on the same model now switch; a genuine no-op emits a
user-visible notice on the session path and never arms on the subagent path.
Fixes#6659
- Removes `model` field from task item/schema, TaskParams, and TaskItem types.
- Removes model selector validation, formatting, and approval display logic.
- Updates task tool priority docs to reflect that model is no longer per-call overridable.
- Updates eval agent() helper docs and prompt templates to remove model parameter.
- Updates tests to reflect removal of model override capability.
Reverted branch-side edits to spawn-policy prompts/tests, settings tab
groups, mermaid cache typing, prewalk todo gating, and packages/ai test
churn back to merge-base content; trimmed their changelog entries. These
repaired stale CI against an older main and are stale or conflicting
against current main.
The aborted-task follow-up hint always references history://<agentId>,
including when no transcript can actually be served. Following that link
then fails.
Add hasResolvableTranscript beside sessionFilesFromDisk, mirroring the
availability half of HistoryProtocolHandler's resolution semantics: a
registered ref's live session, a retained session file verified on disk,
or a disk-scanned .jsonl under a known artifacts dir (which still serves
hard-aborted children whose refs were unregistered). Probing never
throws; a stale path or unreadable artifacts subtree reads as unavailable
instead of failing delivery of the settled result. Render the transcript
clause from that check, independently of the resume affordance, so a
still-resumable idle/parked agent keeps its hub resume hint and a
disk-backed transcript keeps its link. Idle-completion hints are
unchanged.
Signed-off-by: Christian Stewart <christian@aperture.us>
- Threaded a resolvedModelIsFallback flag through AgentProgress and SingleResult.
- Set the flag from the executor retry-fallback handlers and settled results.
- Rendered the observer/no-session hub path as fallback -> provider/model.
- Added an observer-only fallback-badge regression test.
Fixes#6316
Shared the bundled task prewalk default between runtime execution and the Agent Control Center. Added dashboard regression coverage for task.prewalk.
Fixes#6306
Setting options.additionalDirectories to undefined wasn't enough:
createAgentSession also merges settings.get('workspace.additionalDirectories').
Now createSubagentSettings gets an override to clear the setting
when worktree is set, ensuring isolated runs can't edit outside the
worktree.
Co-authored-by: oh-my-pi <https://omp.sh>
Isolated tasks clone only cwd into the worktree. Forwarding the
parent's additional directories would let the subagent edit absolute
paths under the original extra roots, bypassing isolation. Now
additionalDirectories is undefined when worktree is set.
Co-authored-by: oh-my-pi <https://omp.sh>
Subagents (task tool) now inherit the parent session's
additionalDirectories via ToolSession → ExecutorOptions →
CreateAgentSessionOptions, so delegated agents see the same
<workspace-roots> block and can read/grep/glob added roots.
Co-authored-by: oh-my-pi <https://omp.sh>
MCP-backed tools never declared an explicit strict value, so OpenAI-family
serializers (post-#4336/#4340) had no false to preserve and models over-filled
mutually exclusive optional fields. Task/subagent proxies also rebuilt a raw
tools/call instead of executing through the source MCPTool, bypassing intent
stripping, placeholder pruning, local-URL resolution, reconnect, abort, and
result metadata; strict servers rejected proxied calls with
unrecognized_keys ["i"].
- MCPTool/DeferredMCPTool now declare `readonly strict = false as const`.
- createMCPProxyTools delegates to the current source tool, re-resolved by raw
MCP server/tool metadata so reconnect replacements are honored, and keeps the
Task 60s timeout by combining its abort signal with the caller's.
- Regression coverage: strict flags, proxy parity for i/placeholder shaping,
declared-i passthrough, and reconnect re-resolution.
Fixes#6208
- Seeded dirty-baseline blobs into the parent object database before reconstructing filtered agent commits.
- Used three-way synthetic-tree application for committed and trailing task state while preserving parent WIP.
- Added a focused merge regression covering unrelated edits in the same tracked file.
Fixes#6135
Tracked active subagent session model changes in progress snapshots so prewalk handoffs replace the starting-model badge.
Added regression coverage for a prewalk handoff and documented the fix.
Fixes#6083
The flat single-spawn task wire schema carries arktype `"+": "delete"`, so a
batch `{ context, tasks[] }` payload sent while `task.batch` is disabled has
those keys stripped and is then rejected as `task must be a string (was
missing)` in the agent loop. That preempts the tool's own actionable checks
(validateShapeParams / validateSpawnParams), so the model only ever saw the
misleading arktype error instead of "task.batch is disabled…".
Mark TaskTool with lenientArgValidation so the agent loop forwards the raw
args to execute() on any arktype failure, letting the tool's shape checks
surface the real reason. Valid calls still normalize through arktype; the
success path is unchanged. Mirrors the existing yield-tool pattern.
Fixes#6039
The barrier in driveSessionToYield was unreachable for terminal yields:
the yield tool's shouldTerminate fired requestAbort("terminate"), so
abortSignal was always aborted before the barrier's loop condition ran,
and a run with pending owner jobs completed immediately with whatever
the pre-async yield said (Codex review on #6119).
- Split "stop the free-running turn after yield" from "terminate the
run": a terminal yield with pending owner async work now parks the
run with a recoverable session abort (budget-stop precedent) via
requestYieldTurnStop; only a quiescent yield terminates.
- An async-result follow-up injected after a recorded yield un-latches
it (transcript-ordered, in the run monitor) and re-runs the reminder
ladder, so the run only completes on a yield that postdates every
delivered result — including results injected during the notice turn.
- A run that never refreshes a superseded yield fails (exit 1) with an
explicit reason; the stale payload ships only as failed-run salvage
through the existing failed-after-yield finalize path.
- Rewrote subagent-async-pending.md: the "your current yield stands"
option contradicted the enforced contract.
- Regression tests: parked yield -> injected result -> fresh yield wins;
refusal -> stale payload fails; no-async fast path unchanged.
Narrow the invalid-yield guard to !abortSent so array-typed incremental
yield sections no longer suppress the infinite-submit-loop abort; add
regression coverage for incremental yield followed by repeated malformed
terminal yields.
Add an optional `apply` parameter to the `task` tool so
`isolated: true, apply: false` captures patch/branch artifacts without
applying changes to the parent checkout. Available as a flat top-level
control and per `tasks[]` item. Shares the task/eval isolation-to-executor
translation via a single `toStructuredSubagentIsolationControls` adapter.
Three-piece architecture so subagents inherit async.enabled and
bash.autoBackground.enabled instead of having both force-disabled:
- Owner-routed delivery: AsyncJobManager gains registerDeliverySink /
waitForOwnerJobs; every AgentSession registers a sink for its own agent
id, so background job results inject into the owning agent's run.
Owned deliveries with no live sink dead-letter (result retained on the
job row) instead of misrouting into the first top-level session.
- Quiescence barrier: a subagent's final yield with owner jobs still
running/undelivered is a scheduling pause, not completion. The run
driver notifies the model once (hub wait/cancel), settles owner work,
and folds results in as async-result follow-ups; teardown cancels and
awaits surviving jobs before isolation worktree capture/cleanup.
- Steering soft channel: queued steering no longer hard-aborts
non-interruptible tools; it aborts interruptible waits and raises a
cooperative ToolCallContext.steeringSignal. The mid-batch watch runs
for every batch, and auto-backgroundable bash backgrounds itself on
steer so incoming messages inject promptly with no work lost.
Nik Divjak