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

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs	Skipped		May 6, 2026 4:47pm

PR Summary

Medium Risk
Touches workflow execution boundaries (snapshots, executor context, tool/provider request building) by coercing/stripping malformed persisted values, which could change behavior for edge-case runs. Changes are mostly defensive normalization with added unit tests, so overall risk is moderate.

Overview
Hardens workflow execution against legacy/corrupted persisted state by normalizing environmentVariables, workflowVariables, selectedOutputs, and other context maps before they cross tool/provider boundaries.

Introduces shared utilities (normalizeStringArray, normalizeStringRecord, normalizeRecord, normalizeRecordMap, normalizeWorkflowVariables) and applies them across the executor (DAGExecutor, BlockExecutor streaming), snapshot creation/serialization (ExecutionSnapshot now normalizes at construction), block handlers (FunctionBlockHandler, condition/agent), provider tool execution (prepareToolExecution), and function_execute request body construction.

Adds targeted tests covering normalization behavior for arrays/records, ExecutionSnapshot, and function handler execution with malformed context inputs.

^{Reviewed by Cursor Bugbot for commit da50222. Configure here.}

Greptile Summary

This PR adds a normalization layer at all execution-context boundaries to handle legacy or corrupted workflow state where workflowVariables might arrive as an array instead of a record, environment variables might contain non-string values, or selectedOutputs might contain non-string elements.

• Introduces normalizeStringArray, normalizeRecord, normalizeStringRecord, normalizeRecordMap, and normalizeWorkflowVariables utilities that sanitize unknown inputs into the shapes expected by internal APIs, replacing || {} / ?? [] fallbacks throughout handlers, providers, and tools.
• Refactors ExecutionSnapshot to accept unknown for workflowVariables and selectedOutputs, normalizing at construction time, and moves the parseVariableValueByType in-place mutation loop to before Executor construction so typed values are visible at initialization.

Confidence Score: 4/5

Safe to merge — the change is additive normalization that degrades gracefully and is well-covered by new unit tests

The normalization utilities are correct and test coverage is solid. Two low-risk design concerns exist: normalizeWorkflowVariables silently lets a later legacy-array entry overwrite an earlier one when keys collide, and the function returns the original object reference for already-valid records, creating an implicit coupling with the in-place mutation in execution-core.ts. Neither issue manifests in the normal (non-legacy-array) path that most workflows use today.

apps/sim/lib/core/utils/records.ts — the normalizeWorkflowVariables key-collision and shared-reference behaviour are the two points most worth a second look

Important Files Changed

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["ExecutionSnapshot / DAGExecutor constructor (unknown inputs)"]
    B["normalizeWorkflowVariables"]
    C{"isPlainRecord?"}
    D["Return same ref"]
    E{"Array?"}
    F["Map id/name to record"]
    G["Return {}"]
    H["normalizeStringRecord"]
    I{"isPlainRecord?"}
    J["Drop null, coerce to String"]
    K["Return {}"]
    L["normalizeStringArray"]
    M{"Array?"}
    N["Filter string elements"]
    O["Return []"]
    A --> B & H & L
    B --> C
    C -- yes --> D
    C -- no --> E
    E -- yes --> F
    E -- no --> G
    H --> I
    I -- yes --> J
    I -- no --> K
    L --> M
    M -- yes --> N
    M -- no --> O
    D & F & G & J & K & N & O --> P["Executor / Handler typed inputs"]

_{Reviews (1): Last reviewed commit: "improvement(func-exec): normalize inputs..." | Re-trigger Greptile}