improvement(execution): memory usage for aggregated results

[icecrasher321]

Summary

Large array manifests to prevent excessive memory usage of data produced in loops.

Type of Change

• New feature

Testing

Tested manually

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		May 19, 2026 6:34am

[cursor]

PR Summary

Medium Risk
Touches core execution/runtime plumbing (function execution, workflow API responses, loop/parallel aggregation, and snapshot serialization), so regressions could impact workflow outputs and run-from-block behavior. Changes are well-covered by new tests but introduce new manifest/materialization paths and access-control tracking.

Overview
Large array results are now compacted into manifest references to reduce memory/response payload size, including loop/parallel aggregated outputs, function results, and Variables block assignments; manifests can be materialized back when required.

Execution/runtime support is extended with sim.values.readArray(...) (isolated-vm broker + docs), stricter rejection of large refs in runtimes without ref-native helpers (shell/E2B/imports), and end-to-end propagation of largeValueKeys/fileKeys to precisely authorize/materialize only the large values/files actually touched.

API response handling is hardened: Response block HTTP output now materializes inline refs/manifests under size limits, runFromBlock snapshots preserve source execution lineage for large-value access, and pause snapshot serialization adds bounded-size checks to avoid full JSON.stringify on oversized state.

^{Reviewed by Cursor Bugbot for commit ec1e107. Configure here.}

[greptile-apps]

Greptile Summary

This PR introduces a LargeArrayManifest system that stores large arrays produced in loops as chunked references in durable storage, preventing excessive in-memory accumulation. It adds fine-grained key-based access control alongside the existing execution-ID-scoped approach and refactors the streaming response to materialize large values before sending them to clients.

• New chunked manifest type: LargeArrayManifest splits large arrays into separately-stored chunks, offloading loop iteration results to durable storage.
• Access control overhaul: The DAG executor collects keys from the filtered source snapshot for runFromBlock so exact cross-execution refs are authorized without broadening workflow-scope access.
• Streaming/response materialization: createHttpResponseFromBlock is now async and calls materializeInlineExecutionValue; the streaming onBlockComplete handler materializes each selected output inline.

Confidence Score: 4/5

The critical execution paths have all been correctly wired and the refactoring is internally consistent.

The three previously-flagged regressions are all addressed in the current diff. The remaining findings are quality/performance concerns. The score reflects the breadth of interconnected changes touching core execution, access control, streaming, variables, and loop state, which warrants thorough end-to-end testing.

streaming.ts and inline-materialization.server.ts deserve a second look, but neither finding is blocking.

Important Files Changed

Filename	Overview
apps/sim/lib/execution/payloads/large-array-manifest.ts	New core chunked-manifest implementation; chunking arithmetic and slice reads are correctly bounded.
apps/sim/lib/execution/payloads/inline-materialization.server.ts	New recursive materializer with budget tracking. Memoized/shared object references contribute byteLength 0, which can undercount the true JSON size for shared object graphs.
apps/sim/lib/workflows/streaming/streaming.ts	onBlockComplete now materializes selected outputs inline with budget tracking; buildMinimalResult re-serializes the entire growing output object after each selected output is appended (O(n²) JSON.stringify calls).
apps/sim/lib/execution/payloads/materialization.server.ts	assertLargeValueRefAccess and assertExecutionFileScope now support an allowedKeys fast-path before the execution-ID checks.
apps/sim/executor/execution/executor.ts	Collects largeValueKeys/fileKeys from the filtered source snapshot for runFromBlock, enabling precise cross-execution key access.
apps/sim/lib/workflows/executor/execution-core.ts	Removed warmLargeValueRefs call for source snapshot; parseVariableValueByType gains a parseLargeExecutionValue guard for JSON-serialized refs.
apps/sim/executor/orchestrators/loop.ts	Iteration results compacted before accumulation; new hasReachedConfiguredIterationLimit short-circuits doWhile loops at maxIterations.
apps/sim/lib/workflows/utils.ts	createHttpResponseFromBlock is now async and calls materializeInlineExecutionValue(data, context), properly materializing LargeArrayManifest before sending to external callers.
apps/sim/executor/handlers/variables/variables-handler.ts	Variables block now calls compactWorkflowVariableValue on assignment values before storing, converting large arrays to manifests.
apps/sim/app/api/workflows/[id]/execute/route.ts	largeValueExecutionIds/Keys/fileKeys/allowLargeValueWorkflowScope threaded through all execution paths correctly.

_{Reviews (13): Last reviewed commit: "fix tests" | Re-trigger Greptile}

[icecrasher321]

@greptile

[icecrasher321]

bugbot run

[icecrasher321]

@greptile

[icecrasher321]

bugbot run

[icecrasher321]

@greptile

[icecrasher321]

bugbot run

[icecrasher321]

bugbot run

[icecrasher321]

@greptile

[icecrasher321]

bugbot run

[icecrasher321]

@greptile

[icecrasher321]

bugbot run

[icecrasher321]

bugbot run

[icecrasher321]

@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 e87c280. Configure here.}

[icecrasher321]

@greptile

[icecrasher321]

@greptile

[icecrasher321]

bugbot run

[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 ec1e107. Configure here.}

[icecrasher321]

@greptile