Skip to content

Latest commit

 

History

History
579 lines (440 loc) · 21.4 KB

File metadata and controls

579 lines (440 loc) · 21.4 KB

Background Orchestration

Background orchestration is the default orchestration model for oh-my-opencode-slim. It assumes native OpenCode background subagents are available and changes the orchestrator from a primary worker into a scheduler.

The old model was:

orchestrator works directly → delegates when useful → waits for result

The default background-orchestration model is:

orchestrator plans → dispatches background specialists → monitors → reconciles → verifies

This is a clean rebuild, not a compatibility layer over the old blocking model.


Runtime Requirement

Background orchestration requires an OpenCode release that includes native background subagents, launched with background subagents enabled:

OPENCODE_EXPERIMENTAL_BACKGROUND_SUBAGENTS=true opencode

The task API and background-control tools are:

Tool Purpose
task(..., background: true) Start a specialist in the background and immediately return a task ID
hook-driven completion OpenCode injects terminal background task results automatically
task_status Check the status of a tracked task
task_result Retrieve a tracked task's result
task_message Queue a non-interrupting message and return queued
task_cancel Stop a generation while retaining its session
task_revive Resume a retained session with a new instruction
wait_for_user Plugin-provided orchestrator tool that pauses automatic orchestrator wakes while the user performs external manual work

If these are not available, the scheduler cannot use the default background workflow. Configure the environment variable through the installer or use the one-shot export above before starting OpenCode.

Use an OpenCode release that includes native background subagents and hook-driven completion; run opencode --version and update if background tasks are missing.


Core Principle

The orchestrator is not the default implementation worker.

Its job is to:

  • understand the user request,
  • break work into dependent and independent units,
  • choose the right specialist for each unit,
  • schedule background work,
  • track task IDs and states,
  • avoid conflicting writes,
  • integrate specialist results,
  • run or route final verification,
  • communicate concise progress and outcomes to the user.

Specialists do the work. The orchestrator manages the work.


Execution Loop

Every non-trivial request follows this loop:

Understand
  ↓
Plan dependency graph
  ↓
Dispatch independent specialists in background
  ↓
Track task IDs and ownership
  ↓
Continue only independent coordination work
  ↓
Wait for hook-driven completion
  ↓
Reconcile results and resolve conflicts
  ↓
Dispatch follow-up work if needed
  ↓
Verify
  ↓
Final response

The orchestrator should not act on assumptions from a still-running task. It can continue scheduling independent work, but dependent work waits for terminal task results.


Scheduler Responsibilities

1. Build a dependency graph

Before dispatching agents, the orchestrator identifies:

  • which questions must be answered before implementation,
  • which tasks can run in parallel,
  • which tasks must be sequential,
  • which files or subsystems each writer owns,
  • which outputs are needed for final verification.

This does not need to be a long plan. It should be just enough structure to avoid wasted work and conflicting edits.

2. Dispatch background specialists

Independent work should be launched with background tasks:

task(
  description="Search auth flow",
  subagent_type="explorer",
  background=true,
  prompt="Find the auth entry points, session storage, and login callback paths. Return file paths and a concise map. Do not edit files."
)

The orchestrator records the returned task ID and keeps working only on safe, independent coordination.

3. Track ownership

The scheduler must prevent write conflicts.

Rules:

  • Only one write-capable specialist owns a file at a time.
  • Do not run two fixer tasks against overlapping folders unless ownership is explicit.
  • UI work that touches shared components should not run beside implementation work that edits the same components.
  • Review tasks can run in parallel with read-only discovery, but not with edits they are supposed to review.

4. Wait, message, cancel, and revive

Background tasks are not complete until OpenCode injects their terminal result or hook-driven completion marks them terminal.

The orchestrator should use background completion events to:

  • wait for dependent results,
  • check long-running tasks,
  • collect outputs before final response,
  • surface failures or blocked tasks clearly.

Use task_status to inspect a task and task_result to collect its result. task_message queues a non-interrupting message and returns queued; it does not stop the current generation. Use task_cancel to stop a generation while retaining its session, then inspect and reconcile any partial file changes before launching replacement work. Use task_revive to resume a retained session with a new instruction.

A cancelled or errored retained session may be revived immediately once its retained state has been verified safe. Acknowledgement controls parent and job-board consumption and reusable-pool display, not same-session revival.

Terminal jobs are reconciled automatically after their result is injected into the orchestrator session. That lifecycle state is not proof the output was used; the orchestrator must still verify it consumed the relevant result before finalizing. Separately, the default-on orchestrator wake scheduler may prompt an idle parent with incomplete todos after continuous idle time; it does not depend on the local job board.

After a full OpenCode or plugin restart, persisted running background-task history is rehydrated into the local job board and immediately reconciled against live host session status. A missing or idle child is a stop candidate: after a 5s confirmation grace it is surfaced as stopped, unreconciled, while a busy child remains running; status lookup failures remain uncertain rather than being treated as completion.

Specialist outputs are inputs, not final truth. The orchestrator reconciles them against each other and the original user goal.

5. Verify

Verification remains orchestrator-owned and should be proportionate to the change. Use focused checks against the final state, broadening them only when risk or uncertainty warrants it. Oracle review is conditional: dispatch it for material semantic or architectural risk, unresolved uncertainty, or another high-cost decision—not automatically.

The final response should only happen after relevant background work is terminal, reconciled, and supported by final-state evidence.


Specialist Roles

Explorer

Read-only reconnaissance and codebase mapping. Usually the first background task for unfamiliar work.

Librarian

External docs, version-specific API behavior, and real-world examples. Runs in parallel with Explorer when implementation depends on current library behavior.

Fixer

Bounded implementation worker. Receives a clear objective, file ownership, constraints, and validation expectations.

Designer

User-facing UI/UX implementation and review. Owns visual polish, responsive layout, interaction quality, and design consistency.

Oracle

Architecture, code review, simplification, risk analysis, and high-stakes debugging. Often used after implementation or before risky refactors.

Council

Multi-model decision support for critical trade-offs. It is not a worker pool; it is for judgment where disagreement is useful.

Observer

Visual/media analysis isolated from the orchestrator context.


Direct Work Boundary

Background orchestration removes the orchestrator-as-worker default.

The orchestrator may directly:

  • ask clarifying questions,
  • read minimal context needed to route work,
  • create and update todos,
  • launch and monitor tasks,
  • synthesize results,
  • run final checks when that is cheaper than delegating.

The orchestrator should delegate:

  • broad code search,
  • unfamiliar library research,
  • implementation,
  • test creation or updates,
  • UI polish,
  • architecture review,
  • visual/media analysis.

This keeps the main context focused on coordination instead of filling it with worker detail.


Task Prompt Contract

Every delegated task should be self-contained.

Include:

  • objective,
  • constraints,
  • relevant files or search scope,
  • ownership boundaries,
  • expected output format,
  • whether edits are allowed,
  • validation to run or report,
  • what not to do.

Task-fit rejections

If a task is outside a specialist's role, it must not attempt partial work. It returns a brief reason to the orchestrator. The orchestrator treats that reason as routing input to reroute or clarify the task and must not retry the unchanged task with the same specialist.

Good background task prompt:

Investigate src/hooks/task-session-manager for assumptions that a task tool
result means the child task has finished. Do not edit files. Return:
1. exact files/functions involved,
2. which assumptions break with background tasks,
3. recommended code changes,
4. tests that should be added.

Bad background task prompt:

Look into background tasks.

State The Orchestrator Must Track

The prompt/runtime treats background tasks as a small job board:

Field Meaning
task ID Native OpenCode background task/session ID
specialist Agent type assigned
objective What the task is responsible for
state running; stopped (runtime ended without terminal task output); completed, error, or cancelled (explicit terminal task output); reconciled (terminal result consumed)
ownership Files/folders/subsystems the task may edit
dependencies Tasks that must complete first
result Final task output once terminal
status certainty status uncertain when the live status map is malformed or unavailable; it never implies completion

Cancelled and errored sessions can remain retained for a later task_revive. They may be revived immediately once their retained state has been verified safe. Acknowledgement controls parent and job-board consumption and reusable-pool display, not same-session revival.

The current todo list can represent user-visible work, but task IDs and file ownership need to be explicit in the orchestrator's working context.


Runtime Integration

The plugin is aware that a task return can mean "background job launched" rather than "work complete". It tracks running task IDs, exposes recent work in the background job board, updates aliases from task results, and keeps multiplexer panes attached while the parent orchestrator continues scheduling.

Orchestrator wake scheduler

When an orchestrator parent stays continuously idle, the plugin may send a periodic internal wake prompt so incomplete TODOs are not abandoned. This is enabled by default with a 5-minute interval:

{
  "backgroundJobs": {
    "orchestratorWake": {
      "enabled": true,
      "intervalMs": 300000
    }
  }
}

intervalMs must be an integer from 60000 to 2147483647. 0 is invalid. Set enabled: false to disable wakes while keeping idle reconciliation and background-job orchestration.

Behavior:

  • Per-session recursive setTimeout(...).unref() after continuous parent-idle time (never a global interval).
  • Only sessions known as the parent orchestrator via session metadata.
  • Host client APIs are authoritative (session.get, todo, children, status, promptAsync with the nested directory request shape). The local Background Job Board is never read or used as a gate.
  • Wake requires valid host response shapes, parent currently idle, and at least one TODO with status pending or in_progress. Unknown/malformed status fails closed. Active children do not suppress a wake.
  • Suppress/clear on question/permission input waits, wait_for_user, foreground fallback, session busy, session deletion, external user messages, and server disposal.
  • One in-flight evaluation/wake per session. Status/waits/generation are rechecked immediately before promptAsync. Cooldown/reservation is recorded before the call so a failed promptAsync cannot storm retries.
  • Default-on safety: the scheduler evaluates a bounded host-progress fingerprint (TODO statuses plus child status/update evidence) to decide whether to keep waking. After two successful wakes with an unchanged fingerprint, further wakes stop for that continuous idle spell. A real external user message or host-observed progress re-arms the cap. Busy caused by the wake itself does not rearm the cap; unrelated busy/error lifecycle events do. The wake prompt text is static and does not include a fingerprint or snapshot.
  • Static wake text (internal initiator part via promptAsync only — no message transform injection or history rewrite):
<system-reminder>
Finish any incomplete TODOs. Await running agents; if one appears stuck, assess it and cancel/respawn only when justified. Do not respond to this reminder.
</system-reminder>

The scheduler does not perform automatic cancellation and does not rely on the local job board. When no incomplete TODOs remain, it ends the current idle spell and stops polling until new activity.

v2 availability: the v2 shim lacks the required session APIs, so this capability-gated feature remains inactive there.

For external manual work, the orchestrator first gives the user concrete steps, then calls wait_for_user as its final tool action. This explicit signal covers text-only HITL turns without attempting to infer intent from assistant prose. The wait remains armed across hook/plugin recreation in the same process and is cleared only by a distinct real external user message or genuine session deletion. Re-observing the user message that preceded the wait, synthetic/internal messages (including foreground-fallback replays), fallback teardown, session errors, and idle/busy events do not clear it. Immediate choices, clarifications, and pasted command output continue to use the question tool. If wait_for_user is intentionally listed in disabled_tools, the orchestrator uses the question tool as the blocking boundary instead.

Background Job Board Injection

By default, each prompt uses the latest board strategy. The hook removes prior metadata-tagged board messages and injects the current board snapshot, preserving the existing strip-and-replace behavior.

For checkpoint-oriented workflows, opt in to the append-only strategy:

{
  "backgroundJobs": {
    "strategy": "checkpoint-compatible",
    "maxRetainedSnapshots": 20
  }
}

checkpoint-compatible preserves prior board snapshots and appends a snapshot anchored after the current real-message tail only when the formatted board changes. Once created, snapshots are replayed on every managed turn, including turns with an internal initiator or an empty board; those turns do not create a new snapshot. Re-running injection with an unchanged board does not create a duplicate. This changes board message history only; task coordination, storage, terminal reconciliation, and reusable-session behavior remain unchanged. The retained snapshot cache is in memory, is limited to 20 snapshots per cache epoch, and is reset when OpenCode reports a session boundary or a compacted/rebased message history. maxRetainedSnapshots controls the epoch size and accepts integers from 1 to 100 (default 20). When a changed snapshot would exceed the configured limit, all retained snapshots are discarded and only the new current snapshot is kept. This intentionally creates one cache miss at the epoch boundary, after which a fresh run of up to the configured limit can accumulate. The cache is lost on plugin restart, so snapshots are not restored beyond those present in the current OpenCode message history.

Runtime Liveness Reconciliation

The job board is a local projection; OpenCode's live session-status map is the liveness authority. After a tracked task launches, the plugin periodically checks that single map for every board job still marked running, while normal session events remain the fast path.

busy and retry confirm that a job is live and reset any pending stop confirmation. An explicit idle state or an absent session in an otherwise valid map is not immediately terminal: the first observation starts a 5s confirmation grace and keeps the job running, status uncertain. Repeat non-busy evidence after that grace records stopped, unreconciled rather than completed: it means execution ended before a native terminal task result was delivered, not that the task succeeded. Stopped sessions are never reusable and stay visible to the parent for recovery. A later live busy observation can revive an unreconciled stopped job. After the parent has been woken and the stop acknowledged, stale busy cannot flip the job back to running. Only explicit terminal task output proves completion, error, or cancellation.

Malformed status entries and failed status requests are surfaced as status uncertain; they never prove that a job stopped or completed and do not confirm a pending stop. Each observation is generation-aware, so a delayed response cannot modify a relaunched task.

Opt-in Wall-clock Supervisor

The plugin can apply a one-shot wall-clock deadline to native background task child sessions. It is disabled by default:

{
  "backgroundJobs": {
    "wallClockTimeoutMs": 900000,
    "abortGraceMs": 10000
  }
}

This supervisor recognizes only an explicit task(..., background: true) call. Foreground tasks and calls where background is omitted or false are not supervised. The deadline begins at the first launch observation for the current run. Duplicate session.created/tool-hook observations, busy activity, tool activity, and liveness timestamps do not renew it. An explicit relaunch or reuse starts a new run generation.

When the deadline wins a race with a real terminal transition, the board records a persistent hard-deadline marker, marks cancellation as requested, starts the bounded abort grace period, and issues exactly one native session abort. The grace timer is independent of whether the SDK abort resolves, rejects, or hangs. An error, cancellation, or child deletion during grace publishes one stable timed-out terminal outcome. If no terminal confirmation arrives before grace expires, the outcome is error, timedOut: true, and statusUncertain: true, with a summary stating that abort was not confirmed.

Late completion, busy, retry, or error events cannot replace a published hard timeout, and a hard wall-clock timeout is not recoverable through the existing external task-wait timeout path. The timeout outcome remains visible to the parent through the normal terminal-unreconciled Background Job Board flow; no prompt or raw task-result rewrite is used. Timeout terminals also issue a permanent logical pane-close intent so generic and cmux multiplexer paths do not respawn a pane on late busy events.

wallClockTimeoutMs accepts 0 or integers from 60000 through 2147483647; abortGraceMs accepts integers from 1000 through 60000. This feature is wall-clock-only: no no-progress/plateau policy, foreground fallback, model swap, session deletion retry, or worker-death guarantee is implied.


Startup Behavior

The installer and docs configure background subagents as a requirement for the default scheduler workflow. If background subagents are unavailable, treat it as an environment or OpenCode-version issue rather than an intentional V1 fallback:

Background orchestration requires OpenCode background subagents.
Start OpenCode with:

OPENCODE_EXPERIMENTAL_BACKGROUND_SUBAGENTS=true opencode

No automatic legacy fallback keeps the mental model clean.


Example Flow

User asks:

Make background subagents first-class in this plugin.

The orchestrator should do something like:

  1. Create todos for discovery, design, implementation, docs, tests, and verification.
  2. Launch Explorer in background to map task-session hooks and task lifecycle.
  3. Launch Oracle in background only if the change has material semantic or architectural risk or unresolved uncertainty.
  4. Continue by preparing the dependency graph and file ownership plan.
  5. Wait for the launched specialists via hook-driven completion.
  6. Dispatch Fixer to implement prompt/config/hook changes with clear ownership.
  7. Dispatch a second Fixer for tests if file ownership is separate.
  8. Wait for implementation results.
  9. Reconcile the implementation and dispatch Oracle only if remaining risk or uncertainty makes independent review worthwhile.
  10. Run proportionate checks against the final state.
  11. Report final state.

At no point does the orchestrator become the main implementer.


Success Criteria

Background orchestration is working when:

  • the orchestrator launches independent specialists in background by default,
  • task IDs are tracked until terminal state,
  • dependent work waits for real task results,
  • file ownership prevents concurrent write conflicts,
  • final responses only happen after reconciliation and verification,
  • users see faster progress on multi-step work,
  • the orchestrator context stays focused on decisions instead of worker detail.

Background orchestration is not just "parallel agents." It is a scheduler-centered operating model for OpenCode's native background subagents.