fix(parallel): remove broken node-counting completion + resolver claim cross-block

[waleedlatif1]

Summary

• Parallel blocks used totalExpectedNodes = branchCount * allNodesInParallel to detect completion, which is wrong when branches have conditional paths (error edges, conditions)
• Success branches complete fewer nodes than error branches, so the count never matches and the counting-based aggregation misfires
• Removed the counting mechanism entirely — sentinel-end (triggered by edge processing) is the correct and sole mechanism for detecting parallel completion
• Wrong resolver claiming known properties to fix issue Parallel block fails #4043

Type of Change

• Bug fix

Testing

All 670 executor tests pass. Tested with parallel blocks containing error edges. Added tests for cross-block case.

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs	Skipped		Apr 8, 2026 5:42pm

[cursor]

PR Summary

Medium Risk
Changes parallel completion/aggregation behavior by removing node-count-based completion detection and relying solely on sentinel-end aggregation, which could affect workflows with complex branching if any code still expected early aggregation. Also extends variable resolution to read parallel outputs across blocks, touching execution state access patterns.

Overview
Fixes parallel execution completion by removing the per-node counting mechanism (dropping completedCount/totalExpectedNodes from ParallelScope/ExecutionContext) and making aggregation happen only when the parallel sentinel-end runs, avoiding mis-detection with conditional/error paths.

Updates node orchestration and parallel scope initialization to no longer compute/pass an expected terminal-node count, and adjusts ParallelResolver to support resolving a parallel’s stored results output via <parallel.result(s)> (including named parallel references) with nested path access; adds/updates tests to cover the cross-block named-reference cases.

^{Reviewed by Cursor Bugbot for commit c4ee839. Configure here.}

[greptile-apps]

Greptile Summary

This PR removes the broken node-counting completion detection for parallel blocks (where completedCount/totalExpectedNodes misfired on conditional paths) and makes the sentinel-end the sole aggregation trigger. It also fixes cross-block variable resolution: named parallel references like <parallelName.result> now correctly route to resolveOutput instead of falling through to the iteration-context path that requires the caller to be inside the parallel.

Confidence Score: 5/5

Safe to merge — the counting-based completion removal is architecturally correct and all 670 executor tests pass.

Both fixes are well-scoped: removing the broken node-count aggregation path eliminates a class of misfires on conditional branches, and the OUTPUT_PROPERTIES early-return in the resolver correctly separates result-access from iteration-context resolution. No P0/P1 findings. New tests cover named-reference resolution from outside a parallel, cross-block paths, nested paths, and InvalidFieldError cases.

No files require special attention.

Vulnerabilities

No security concerns identified. Changes are confined to executor internals (parallel completion detection and variable resolution) with no auth, user-input handling, or data persistence surface.

Important Files Changed

Filename	Overview
apps/sim/executor/orchestrators/parallel.ts	Removes completedCount/totalExpectedNodes from ParallelScope initialization; handleParallelBranchCompletion now returns void and only stores output — aggregation exclusively driven by sentinel-end.
apps/sim/executor/orchestrators/node.ts	Drops the allComplete guard around aggregateParallelResults in handleParallelNodeCompletion; initializeParallelScope calls no longer pass nodesInParallel.
apps/sim/executor/variables/resolvers/parallel.ts	Adds OUTPUT_PROPERTIES set {'result','results'} checked before KNOWN_PROPERTIES; resolveOutput reads block output stored by aggregateParallelResults; availableFields in InvalidFieldError updated to include 'result'.
apps/sim/executor/variables/resolvers/parallel.test.ts	Adds a full describe block for named parallel references covering result/results resolution from outside, nested paths, cross-block access, and InvalidFieldError for unknown properties.
apps/sim/executor/execution/state.ts	Removes completedCount and totalExpectedNodes from ParallelScope interface, consistent with the orchestrator changes.
apps/sim/executor/types.ts	Minor type cleanup related to the parallel scope changes.
apps/sim/executor/orchestrators/parallel.test.ts	Updates initializeParallelScope call to drop the nodesInParallel argument, keeping test consistent with the new signature.
apps/sim/executor/utils/iteration-context.test.ts	Adds cross-block test cases for the iteration context utilities.

Sequence Diagram

sequenceDiagram
    participant NE as NodeExecutionOrchestrator
    participant PO as ParallelOrchestrator
    participant PR as ParallelResolver
    participant State as BlockState

    Note over NE,State: Branch node completes
    NE->>PO: handleParallelBranchCompletion(ctx, parallelId, nodeId, output)
    PO->>PO: extractBranchIndex(nodeId)
    PO->>PO: scope.branchOutputs.set(branchIndex, output)
    Note right of PO: Returns void — no counting logic

    Note over NE,State: Sentinel-end fires (sole aggregation trigger)
    NE->>NE: handleParallelSentinel(ctx, node, 'end', parallelId)
    NE->>PO: aggregateParallelResults(ctx, parallelId)
    PO->>State: setBlockOutput(parallelId, { results })
    PO-->>NE: { allBranchesComplete: true, results, totalBranches }

    Note over PR,State: Cross-block resolution (new fix)
    PR->>PR: OUTPUT_PROPERTIES.has('result'|'results') = true
    PR->>PR: resolveOutput(targetParallelId, pathParts, ctx)
    PR->>State: getBlockOutput(targetParallelId)
    State-->>PR: { results: [...] }
    PR-->>PR: return output.results

Loading

_{Reviews (2): Last reviewed commit: "fix resolver claim" | Re-trigger Greptile}

[icecrasher321]

bugbot run

[icecrasher321]

@greptile

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit c4ee839. Configure here.}