invoke-ai/InvokeAI · error · ValueError

Workflow return key '{key}' was not found.

Error message

Workflow return key '{key}' was not found.

What it means

Workflow Return Get looks up self.key in the workflow's named return values dict. If the stripped key is non-empty but not present among the values the workflow returned, invoke() raises ValueError naming the missing key. This guards against typos and keys removed from the return node.

Source

Thrown at invokeai/app/invocations/workflow_return.py:138

)
class WorkflowReturnGetInvocation(BaseInvocation):
    """Extracts one named value from a callable workflow return."""

    values: dict[str, Any] = InputField(
        default={},
        description="The named workflow return values.",
        title="Values",
        ui_type=UIType.Any,
        input=Input.Connection,
    )
    key: str = InputField(default="", description="The return key to extract.", title="Key")

    def invoke(self, context: InvocationContext) -> WorkflowReturnGetOutput:
        key = self.key.strip()
        if not key:
            raise ValueError("Workflow return key must not be empty.")
        if key not in self.values:
            raise ValueError(f"Workflow return key '{key}' was not found.")
        return WorkflowReturnGetOutput(value=self.values[key])

View on GitHub (pinned to 0b6a024f2f)

Solutions

1. Align the get node's 'key' with an existing key on the workflow-return node's values list
2. Check exact spelling, case, and surrounding whitespace of both key fields
3. Re-run the workflow and inspect which keys the return node actually emits
4. Update stale workflow JSON so get nodes reference current return keys

Example fix

// before
get = WorkflowReturnGetInvocation(values=return_values, key='outpt')
// after
get = WorkflowReturnGetInvocation(values=return_values, key='output')

def key_exists(invocation, return_values):
    k = invocation.key.strip()
    return k in return_values

if not key_exists(invocation, return_values):
    raise KeyError(f"'{invocation.key}' not in workflow return keys: {list(return_values)}")

try:
    output = invocation.invoke(context)
except ValueError as e:
    if "was not found" in str(e):
        log_missing_return_key(extract_key_from_message(str(e)))
    else:
        raise

When it happens

Trigger: invoke() on a WorkflowReturnGet invocation where self.key (after .strip()) is not a key in the values dict produced by the workflow-return invocation, e.g. the return node no longer emits 'result' but the get node still requests it.

Common situations: Renaming or deleting a key on the workflow-return node while downstream get nodes still reference the old name; importing a workflow whose return keys changed between versions; a case/whitespace mismatch between the two key fields.

Related errors

• A saved workflow must be selected before executing call_save
• The selected saved workflow '${self.workflow_id}' could not
• Workflow return key must not be empty.
• Duplicate workflow return key '{key}'.
• Unsupported batch group id '{batch_group_id}' in called work

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/236ed48fee732a1a. Report an issue: GitHub.