fix(core): Make n8n-workflow a peer dependency in ai-utilities

[DawidMyslak]

Summary

When published to npm, @n8n/ai-utilities had n8n-workflow as a direct dependency, which allowed npm/pnpm to install a separate nested copy. Community nodes using @n8n/ai-node-sdk would then end up with two instances of n8n-workflow at runtime — one from the host (n8n) and one nested inside @n8n/ai-utilities — causing type mismatches and instanceof failures.

This PR moves n8n-workflow from dependencies to peerDependencies in @n8n/ai-utilities, so a single host-provided instance is used throughout.

Related Linear tickets, Github issues, and Community forum posts

https://linear.app/n8n/issue/NODE-4532

Review / Merge checklist

• PR title and summary are descriptive. (conventions)
• Docs updated or follow-up ticket created.
• Tests included.
• PR Labeled with release/backport (if the PR is an urgent fix that needs to be backported)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[cubic-dev-ai]

cubic analysis

No issues found across 3 files

Linked issue analysis

Linked issue: NODE-4532: SDK n8n-workflow types is incompatible with peer n8n-workflow

Status	Acceptance criteria	Notes
✅	Move n8n-workflow from dependencies to peerDependencies in @n8n/ai-utilities	Moved n8n-workflow out of dependencies into peerDependencies
✅	Add n8n-workflow to peerDependencies in @n8n/ai-node-sdk	Added peerDependencies entry for n8n-workflow
✅	Update pnpm-lock.yaml to reflect peer dependency changes (avoid nested copy)	pnpm-lock updated specifier/version to reflect change
⚠️	Ensure a single host-provided n8n-workflow instance is used at runtime (prevent type mismatches / instanceof failures)	Peer deps added but no runtime/tests to verify single instance

Architecture diagram

sequenceDiagram
    participant Host as n8n (Host Process)
    participant Node as Community Node
    participant SDK as @n8n/ai-node-sdk
    participant Utils as @n8n/ai-utilities
    participant Workflow as n8n-workflow (Shared Instance)

    Note over Host,Workflow: Runtime Dependency Resolution

    Host->>Workflow: Load singleton instance
    Host->>Node: Load & Execute Node

    Node->>SDK: Call AI helper (e.g. supplyModel)
    
    SDK->>Workflow: CHANGED: Resolve peer dependency
    Note right of SDK: Uses Host's version of n8n-workflow
    
    SDK->>Utils: Forward request
    
    Utils->>Workflow: CHANGED: Resolve peer dependency
    Note right of Utils: No longer uses internal/nested copy
    
    alt Validation Path
        Utils->>Workflow: Perform instanceof check (ISupplyDataFunctions)
        Workflow-->>Utils: Match Found (Single Instance)
        Utils-->>SDK: Return initialized Model/Memory
        SDK-->>Node: Return data
    else Type Mismatch Path (Fixed)
        Note over Utils,Workflow: Previously: Instanceof would fail here <br/> due to nested n8n-workflow copies
        Utils-->>SDK: Error: Invalid type
    end

    Node-->>Host: Return node execution result

Loading

[yehorkardash]

Do we need it here, if ai-utilities defines n8n-workflow dependency?

[DawidMyslak]

Ah, since community node templates already declare their own "n8n-workflow": "*" peer dependency independently, you're right - the SDK doesn't need to duplicate that.

[n8n-assistant]

Got released with [email protected]