- Removed deprecated eval prelude helpers `append`, `tree`, `diff`, `sort`, `uniq`, and `counter` from all supported runtimes.
- Cleaned up runtime implementations, protocol definitions, and UI rendering logic associated with the removed helpers.
- Updated project documentation, prompts, and test suites to reflect the reduced helper API surface.
- Recorded functional changes in the package changelog.
The eval agent() helper used `agent_type`/`return_handle` (snake_case) in
Python/Ruby/Julia and `agentType`/`returnHandle` (camelCase) in JS, forcing
the prelude docs to repeat every option twice ("JS same but camelcased").
Both are now single lowercase words identical across all four runtimes, and
`agent` matches the `task` tool's existing agent-selection parameter.
- Renamed across py/js/rb/jl preludes (signatures, forwarding, docstrings).
- Renamed the `__agent__` bridge wire protocol + `EvalAgentArgs` (`agentType`
→ `agent`, `returnHandle` → `handle`) so no prelude-side remap is needed.
- Updated prompt docs (workflow-notice.md, tools/eval.md), repo docs
(docs/tools/eval.md, docs/python-repl.md), and all bridge/prelude tests.
- CHANGELOG: Breaking Changes entry under [Unreleased].
Resolves conflict in test/task/worktree.test.ts by keeping both the
getRepoRoot (main) and applyNestedPatches (PR) describe blocks.
Extends the PR's Python/JS work to the remaining workflow runtimes:
- eval/rb/prelude.rb, eval/jl/prelude.jl: agent() now accepts and
forwards isolated/apply/merge (as booleans) plus returnHandle, and the
return_handle node carries isolated/patch_path/branch_name/
nested_patches/changes_applied/isolation_summary.
Post-merge fixups:
- task/index.ts: drop dead commitStyle var (the dedup refactor reads
task.isolation.commits inside makeIsolationCommitMessage).
- CHANGELOG: move the misplaced Added entry under [Unreleased], correct
the stale "defaults track task.isolation.mode" wording to the final
strict opt-in behavior, and note all four runtimes.
Fixes#3196
Per maintainer ruling on #3196, eval agent() now defaults to non-isolated regardless of task.isolation.mode, mirroring the task tool. isolated=true is the only way to turn it on; isolated=true while task.isolation.mode === "none" still throws the same clear error.
Updated tests, workflow-notice.md, and Python agent() docstring to reflect the strict opt-in contract. Existing isolation tests now pass isolated:true explicitly; the inherit-from-settings assertion is replaced with a default-off + isolated=true opt-in regression.
Fixes#3196
- Extracted common kernel and executor logic into `BaseKernel` and `executor-base` to eliminate duplicated implementations for Julia, Python, and Ruby.
- Migrated shared operational workflows--including session namespacing, environment filtering, and result mapping--to centralized backend helpers.
- Consolidated runtime discovery and resolution logic into a unified `runtime-env` utility module.
- Simplified language-specific modules by delegating subprocess lifecycle, IPC, and configuration management to the newly established base classes.
Branch-mode isolation can capture nested repository changes without creating a root branch. Eval agent() with apply=false previously treated that shape as no captured changes and returned no recoverable nested patch payload after the isolation worktree was removed.
Expose captured nested patches in EvalAgentResult details and copy them onto JS/Python returnHandle nodes (nestedPatches / nested_patches). Document the return_handle escape hatch and add regression coverage for branch-mode nested-only apply=false runs.
Fixes#3196
Eval preludes now forward returnHandle to the bridge so no-session eval runs can preserve the temp artifacts backing returned agent:// handles. The bridge keeps those temporary artifact directories whenever returnHandle is requested, including non-isolated runs and successful isolated applies.
Branch-mode isolation now treats nested-only changes as merge-eligible even when no root branch was produced, letting callers apply nested patches instead of dropping them when the root repo had no diff.
Added regression coverage for returnHandle artifact preservation and nested-only branch isolation.
Fixes#3196
When agent() ran with schema and apply=false, the bridge correctly returned the captured patch/branch in details, but the preludes only forwarded id/agent/handle/data on the returnHandle node. Structured workflows had no way to recover the artifact for a manual apply.
Both runtimes now copy isolated, patchPath/branchName, changesApplied, and isolationSummary onto the returnHandle node (snake_case in Python, camelCase in JS), keeping null changesApplied so apply=false stays distinguishable from a successful apply. Updated the workflow notice and the Python agent() docstring to point callers at return_handle as the artifact escape hatch for isolated+apply=false runs. Added prelude tests locking the new node shape in both runtimes.
Fixes#3196
The workflowz eval path bypasses the task tool's isolation wrapper and
calls runSubprocess() directly, so parallel agent() fan-outs that edit
overlapping files all land in the parent worktree.
Extends the eval agent bridge schema with isolated/apply/merge, forwards
them through the Python and JS preludes, and adds a shared
task/isolation-runner.ts so the lifecycle (prepare context → run in
worktree → capture patch/branch → merge → cleanup) is implemented once
for both TaskTool and the bridge.
Default mirrors task.isolation.mode: isolated by default when settings
allow it, off when mode === 'none'. isolated=False explicitly disables;
isolated=True with mode === 'none' errors out to match the task tool.
apply=false keeps captured changes inside the worktree and surfaces the
patch path / branch name in details. merge=false forces patch mode even
when task.isolation.merge === 'branch'.
Fixes#3196
- Renamed the global `INTENT_FIELD` constant from `_i` to `i`.
- Updated documentation strings, type annotations, and test expectations across packages to reflect the new field name.
- Ensured consistent usage of the constant in tool schema construction and intent serialization.
- Moved the `INTENT_FIELD` constant from `@oh-my-pi/pi-agent-core` to the specialized `@oh-my-pi/pi-wire` package to permit broader usage across the monorepo.
- Updated all references across `agent`, `ai`, `coding-agent`, `collab-web`, and `snapcompact` packages to import the constant from the new location.
- Added `@oh-my-pi/pi-wire` as a dependency to all affected packages.
- Added `sessionId` and `cwd` tracking to JS evaluator session instances.
- Updated `acquireSession` in the JS context manager to update existing session identity and directory metadata on reuse.
- Forwarded the current working directory from `executePerCall` and `executeOnSession` callers to the underlying Python kernel tasks.
- Added Matplotlib figure PNG rendering and display tracking in Python runner to emit PNG output immediately when figures are displayed via display(fig).
- Extended session persistence to externalize oversized image payloads in both content and details.images, enabling tool result images to survive session reload.
- Enhanced session loader to resolve image data payloads and blob references across content and details.images during session reconstruction.
- Added image cache invalidation in TUI image component when image protocol, cell dimensions, or Kitty Unicode placeholder mode changes.
- Added comprehensive test coverage for Matplotlib display, image persistence across reload, and TUI image rendering with protocol and dimension changes.
- Added a return_handle (Python) / returnHandle (JS) option to the eval agent() helper that returns a DAG node dict { text, output, handle, id, agent } instead of bare text, where handle is the spawned agent's recoverable agent:// URI.
- Enabled downstream pipeline/parallel stages to reference a large transcript by handle/output instead of re-inlining it; the default path stays backward compatible (bare text, or the parsed object under schema).
- Documented return_handle and the acyclic DAG-wiring pattern in the eval tool description and added a VM-level prelude regression test for the node shape and the no-details fallback.
Shared background now flows through a '/Users/can/.omp/agent/sessions/-Projects-.tree-pi-commit/2026-06-10T15-36-32-782Z_019eb22d-970e-7000-8964-72c98becf3e8/local' file referenced in each prompt instead of a context string forwarded into the subagent's system prompt. The JS and Python preludes drop the context kwarg from agent(), the subagent system prompt drops the {{#if context}} block and the conversation-context file pointer, and runEvalAgent no longer writes a per-call conversation context file. AgentSession sheds the now-unused formatCompactContext() helper that supplied the file's body, and ToolSession.getCompactContext is removed alongside it.
Resolve the explicit interpreter from the session's Settings instance
(ToolSession.settings / AgentSession.settings) instead of re-reading the
process-global Settings.init() singleton, so project-scoped and cloned
session settings take effect. The availability cache is now keyed by
cwd + interpreter, and PythonKernel.start/executePython accept the
resolved interpreter as an option. Also expand home-relative paths
(~/...) before resolving against cwd, and document the contract of
resolveExplicitPythonRuntime.
Addresses review feedback on #2204.
single OutputSink owner per cell artifact; JS parallel() honors its documented barrier (allSettled) instead of orphaning in-flight thunks; Python subprocesses no longer inherit the NDJSON frame pipe (stdout captured and forwarded); JS timeouts annotate the VM reset; console bridge implements dir/time/group/assert/trace; python availability probe cached; runner frames coalesce per write.
- Added status.done and tool.* symbols to theme mappings and presets.
- Replaced generic success glyphs with contextual +/-, tool icons, and warnings.
- Mapped tool/task/job completions to status.done or status.enabled with icon overrides.
- Triggered runtime provider refresh after extension registration and warned on failure.
- Substituted injected on-disk roots for `local://` in read/write/append (Python and JS).
- Pinned `local://` to the session's own root so eval writes land where reads resolve.
- Rejected path traversal and unknown `scheme://` paths instead of creating junk `local:` dirs.
- Added unit and integration tests covering resolution, guards, and plain-path passthrough.
- Renamed eval oneshot helper from llm() to completion() across JS/Python APIs.
- Remapped eval bridge internals to completion semantics (__completion__, runEvalCompletion, completion status op).
- Updated docs, prompts, and timeout guidance to describe completion() usage and behavior.
- Adjusted completion defaults for active-session model preference, fallback parsing, and slow-tier effort handling.
- Coalesced concurrent JS and Python reset requests by awaiting in-flight promises instead of throwing.
- Aligned non-reset execution calls to wait for in-progress resets before running on a recreated session.
- Updated `packages/coding-agent/src/eval/py/prelude.py` to accept positional `offset` and `limit` in `read()`.
- Added regression coverage in `packages/coding-agent/src/eval/py/__tests__/prelude.test.ts` for positional read signatures.
- Added timeout pause/resume control ops and helper to suspend idle timers during bridge work.
- Fixed TimeoutError on long delegated agent()/llm() calls by pausing timeout while silent.
- Updated IdleTimeout with reference-counted pauses, ignored checks while paused, and resumed fresh.
- Replaced heartbeat keepalives with timeout-control events in bridge paths and status routing.
The TTY-OR heuristic still mis-classifies the all-stdio-redirected case:
`omp -p "..." < in.txt > out.txt 2> err.log` from a real Windows Terminal
session has `stdin.isTTY === stdout.isTTY === stderr.isTTY === false`,
but the host process can still own a console that the kernel could
inherit. The TTY signals reflect handle redirection, not console
attachment — `GetConsoleWindow()` is the authoritative Win32 signal.
`spawn-options.ts` now:
- Calls `kernel32!GetConsoleWindow()` via `bun:ffi` on Windows. A non-NULL
HWND means the host has a console regardless of how the standard
streams are wired, so `windowsHide` stays `false` and the kernel
inherits the console — which is what fixes the #1960
numpy/pandas `LoadLibraryExW` hang and lets SIGINT recover via
`GenerateConsoleCtrlEvent`.
- Falls back to the TTY-OR heuristic when the FFI probe is unavailable
or off-Windows. That keeps the predicate working on non-Bun-FFI
runtimes and on POSIX, where `windowsHide` is a no-op anyway.
- Caches the probe result; console attachment is stable for the host's
lifetime in practice, and dlopening kernel32 on every kernel spawn
would be wasteful.
The pure helpers (`shouldHideKernelWindow`, `consoleAttachedViaTTY`)
stay separately exported so they remain unit-testable. The integration
boundary `hostHasInheritableConsole()` is what the kernel spawn site
calls, and its return is asserted to be a concrete boolean (kernel spawn
must commit to a `windowsHide` value).
Stand-alone `process.stdout.isTTY` mis-classifies any partial stdio
redirection — `omp -p "..." > out.txt` reports `stdout.isTTY === false`
even though the parent still owns a console via stdin/stderr. With the
previous check the kernel would have been spawned with `CREATE_NO_WINDOW`
in that scenario and re-hit the #1960 numpy/pandas `LoadLibraryExW` hang
and broken SIGINT.
The host owns a console it can share with the child whenever ANY of
stdin / stdout / stderr is still a TTY; only a fully detached launch
(true service / daemon, or `< in > out 2> err`) has nothing to inherit.
Renamed the predicate parameter to `hostHasInheritableConsole` so the
contract is unambiguous, and added five regression tests covering the
realistic shell redirection combinations.
`PythonKernel.start()` spawned the runner with `windowsHide: true`, which
in Bun maps to the Win32 `CREATE_NO_WINDOW` flag — that detaches the
long-lived child from any inherited console. Two consequences for OMP's
Python eval on Windows:
1. Native extensions that probe the console at init (e.g. NumPy's
`_core/_multiarray_umath.pyd` plus its bundled OpenBLAS/SLEEF
thread-pool init) can deadlock inside `LoadLibraryExW`, so the very
first `import pandas` / `import numpy` after a cold kernel never
returned. The reporter's faulthandler stack pinned the hang to
`_bootstrap_external.create_module` → `numpy/_core/multiarray.py:11`.
2. SIGINT cannot be delivered to a console-less process via
`GenerateConsoleCtrlEvent`, so the host's `proc.kill("SIGINT")`
silently no-ops — matching the reporter's "kernel unresponsive to
interrupt" log line. The 5s escalation then hard-kills the kernel.
Plain `python.exe -u -c "import pandas"` from the same venv inherits the
terminal's console and runs to completion, which is the contract this
change restores. The Python kernel now hides its window only when the
host itself has no console to share (service / piped launch); an
interactive TUI launch lets the kernel inherit the parent's console —
analogous to `python.exe` invoked from `cmd.exe`.
The `shouldHideKernelWindow` predicate lives in its own module so it can
be unit-tested without dragging in the kernel's runtime dependencies.
Short-lived helper subprocesses elsewhere (LSP probes, git, plugin
installs) intentionally keep `windowsHide: true` — they don't load
complex native modules and a brief console flash would be user-visible
noise.
Fixes#1960
- Resolved in-flight bridge calls the instant the cell's signal aborts.
- Let the kernel unwind via KeyboardInterrupt instead of being hard-killed.
- Preserved persistent session state across wide subagent fan-out teardown.
- Removed the `concurrency` argument from `parallel()` and `pipeline()` in both JS and Python runtimes.
- Added `__concurrency__` bridge to resolve the pool ceiling live from `task.maxConcurrency` (default 32; 0 = unbounded).
- Eval fan-outs now run as wide as a `task` tool batch instead of being capped at 16.
- Only bridge heartbeats (`agent()`/`llm()`) now re-arm the watchdog; compute, stdout, `log()`/`phase()`, and ordinary tool calls count against the budget.
- Emitted an immediate heartbeat at bridge call start to avoid early abort near budget edge.
- Removed `idle` flag and "of inactivity" suffix from timeout annotation strings.
- Updated docs, prompts, and comments to reflect the new wall-clock semantics.
The per-cell `timeout` is an inactivity budget that only re-arms on status
events, but host-side bridge calls can run long stretches with no
intermediate status (a subagent's time-to-first-token on a reasoning
model, a long quiet nested tool, or an entire oneshot llm() request).
The watchdog mistook that for a stall and aborted working subagents
mid-flight.
Pump a lightweight heartbeat while a bridge call awaits, re-arming the
watchdog through the existing emitStatus -> onStatus channel. The
heartbeat is a pure keepalive: forwarded to bump the timer but never
stored or rendered, so a genuinely stalled cell is still interrupted
once the call settles.
- eval/heartbeat.ts: withBridgeHeartbeat() + EVAL_HEARTBEAT_OP
- agent-bridge/llm-bridge: wrap runSubprocess / completeSimple
- js+py executors: forward heartbeat to onStatus, drop from displayOutputs
- tools/eval.ts: bump on heartbeat, skip persist/render
- Changed eval timeout behavior from hard wall-clock deadlines to per-cell inactivity budgets in all executors.
- Added IdleTimeout watchdog support, including bumps on status/tool activity and timer cleanup after execution.
- Updated executor option plumbing to replace deadlineMs with idleTimeoutMs and emit inactivity timeout annotations.
- Added IdleTimeout and shared-executor tests and updated prompt/repl docs for the new timeout contract.
- Added optional `onStatus` callback wiring across eval backends and JS/Python executors for live status streams.
- Added collectDisplay-based forwarding so `emitStatus` and `onDisplay` route status outputs consistently.
- Expanded agent status payloads with preview/model/token-cost context and kept completion updates single-pass.
- Added status upsert and render adjustments in `tools/eval.ts` to coalesce agent events with progress stats.
- Added status/progress test coverage for running/completed agent events, final metric retention, and parallel placement.
- Updated CHANGELOG Unreleased notes to record live progress updates and completion-status metric fixes.
- Added +Nk/+Nm turn-budget parsing with whitespace-boundary matching, multipliers, and hard `!` indicator.
- Added per-turn budget lifecycle plus APIs (`getTurnBudget`, `recordEvalSubagentUsage`) and hard-cap checks in eval runs.
- Added hard budget observability in eval preludes and docs by exposing `budget.hard` and documenting ceiling modes.
- Fixed streaming preview stutter with max-row tracking and padding, with tests for preview height and budget parsing.
- Dropped `args` input from eval tool schema, JS/Python executors, and worker protocol.
- Removed per-call `args` injection from JS runtime and Python kernel/runner.
- Deleted related tests and updated docs to reflect removal.
- Added `agent()` in JS/Python preludes to call host bridge and parse returned text when schema is set.
- Added JS `parallel()` and `pipeline()` with bounded `__pool()` pools and concurrency normalization.
- Added `runEvalAgent` bridge logic with argument parsing plus plan-mode, allowlist, depth, and artifacts checks.
- Added tool routing and tests documenting new `agent/parallel/pipeline` behavior, defaults, and validation failures.
- Replaced single-candidate resolution with `enumeratePythonRuntimes`, returning venv, managed env, and system interpreter in priority order.
- Availability check now probes each candidate and falls through to the first that executes, so a stale `uv`-managed Python no longer fails the whole session.
- Kernel spawn reuses the probed runtime from the availability result instead of re-resolving independently.
- Expanded test coverage for enumeration, fallback, and env isolation between candidates.
- Namespaced `AgentSession.executePython()` session IDs before invoking the Python executor.
- Added a regression test proving eval state is visible to the user shortcut path.
- Replaced external watchdog timers with per-request SDK timeouts for first-event budget across OpenAI, Anthropic, and Azure providers.
- Keyed Python shared kernels by (sessionId, cwd) to prevent cross-directory state bleed.
- Deduplicated concurrent cold-start session acquisition for JS and Python executors.
- Moved `isOpenAIResponsesProgressEvent` to shared module and scoped display output routing per run for interleaved async cells.
- Added `cwd` and `env` optional parameters to kernel execution API for runtime working directory and environment variable control.
- Implemented runtime environment setup in Python runner with `_apply_request_runtime()` to apply cwd and env from request before code execution.
- Enhanced SIGINT handler management with `active_executions` counter and `_begin_exec_sigint()` / `_end_exec_sigint()` functions to prevent state mutation during concurrent execution.
- Changed `SearchRenderArgs.paths` parameter type from `string[]` to `string | string[]` to accept single string paths.
- Added comprehensive test coverage for kernel cwd updates, timeout interruption safety, and SystemExit handling in shared executor sessions.
- Removed per-session run queues from JS and Python backends, allowing async cells on the same session id to interleave.
- Introduced `getEvalSessionId` on ToolSession so subagents spawned via `task` inherit the parent's executor id and share JS VM and Python kernel state.
- Switched JS runtime state from module-level fields to AsyncLocalStorage so concurrent runs route output and tool calls to their own context.
- Changed Python runner to an asyncio event loop with per-request tasks and ContextVar-based run id tracking for concurrent execution.
- Added mtime-based module cache eviction to preserve singleton state across re-imports of unchanged local files.
A per-cell timeout used to kill the persistent Python kernel, losing
all session state. The kernel.ts timeout path now sends SIGINT first
(letting the cell raise KeyboardInterrupt) and only escalates to
shutdown after a 5s grace window if the interrupt is ignored.
KernelExecuteResult gains an optional kernelKilled flag (defaulting to
false, propagated by the escalation timer and the unexpected-exit
handler). executor.ts formats two distinct timeout annotations: one
that says the kernel is still alive and reset:true would clear state,
one that says the kernel was killed and will be recreated.
- Added generation-aware snapshot contracts with generation, serverNowMs, refresher, and rotatesInMs fields.
- Reworked /v1/snapshot serving and client fetching for If-None-Match long-poll with 304/200 status handling.
- Added status checks in remote-store and SDK/CLI snapshot paths, applying updates only when fetch returns 200.
- Added StreamOptions.onAuthError and stream one-shot 401 retry dispatch using refreshed credentials.
- disposeAllKernelSessions removes sessions from the registry before awaiting shutdown, preventing queued work from seeing a registered-but-disposing session and spawning a replacement kernel.
- replaceSessionKernel re-checks registry membership after old-kernel shutdown and after starting a replacement, shutting the replacement down if the session was disposed mid-replace.
- owner-scoped disposal now preserves ownerIds when shutdown is unconfirmed so a later disposeKernelSessionsByOwner(ownerId) can retry.
- executeOnSession's runQueued callback now re-checks sessions.get(sessionId) === session at slot entry and before the dead-kernel replace retry; if the session was removed by disposeAllKernelSessions / disposeKernelSessionsByOwner / resetSession during the queue wait, the queued caller now rejects with PythonExecutionCancelledError instead of spawning an untracked replacement kernel on a stale session object.
- Both dispose paths now inspect KernelShutdownResult.confirmed and retain (or re-insert) sessions whose shutdown was unconfirmed or rejected, logging a warn so a later dispose can retry the kernel instead of orphaning the subprocess.
can1357