iiRecord
Agentic AI Atlas · GAP-PAR-003: Multi-Harness Parallel Dispatch
page:docs-harness-features-backlog-gaps-parallelization-gap-par-003a5c.ai
II.
Page reference

page:docs-harness-features-backlog-gaps-parallelization-gap-par-003

Reading · 1 min

GAP-PAR-003: Multi-Harness Parallel Dispatch reference

Orchestrate multiple harness invocations working in parallel on related tasks, with persistent group identity across iterations and coordination between dispatched work. Reframed from CC agent swarms to babysitter multi-harness parallel dispatch.

Pagewiki/docs/harness-features-backlog/gaps/parallelization/GAP-PAR-003.mdOutgoing · 0Incoming · 1

GAP-PAR-003: Multi-Harness Parallel Dispatch

FieldValue
Categoryparallelization
PriorityHigh
EffortXL
StatusPartial

Description

Orchestrate multiple harness invocations working in parallel on related tasks, with persistent group identity across iterations and coordination between dispatched work. Reframed from CC agent swarms to babysitter multi-harness parallel dispatch.

Current State

ctx.parallel.all() and ctx.parallel.map() provide one-shot parallel effect dispatch with deduplication. No persistent team identity across iterations, no inter-effect communication during execution, no fan-out/fan-in patterns with intermediate checkpoints.

Target State

Named effect groups with persistent identity across iterations. Multiple harnesses working concurrently on related tasks. Coordinator process pattern for ongoing management. Fan-out/fan-in delegation with intermediate checkpoints.

Dependencies

Key Files

ComponentPath
Task batchingpackages/sdk/src/tasks/batching.ts
Harness adapterspackages/sdk/src/harness/
Process contextpackages/sdk/src/runtime/processContext.ts

Recommendation

Phase 3-4 implementation. Phase 3: named effect groups with persistent identity. Phase 4: inter-effect messaging via journal events and coordinator process pattern.