What causes "Dataset item has no input." in langfuse/langfuse?

Clean the dataset: remove or fix items with null input before running the experiment.

langfuse/langfuse · error · Error

Dataset item has no input.

Error message

Dataset item has no input.

What it means

replaceVariablesInPrompt in the experiments worker requires the dataset item to have an input object; when itemInput is null it throws 'Dataset item has no input.' before building the chat messages.

Source

Thrown at worker/src/features/experiments/utils.ts:80

const fetchPrompt = async (promptId: string, projectId: string) => {
  const promptService = new PromptService(prisma, redis);

  const rawPrompt = await prisma.prompt.findUnique({
    where: { id: promptId, projectId },
  });

  return promptService.resolvePrompt(rawPrompt);
};

export const replaceVariablesInPrompt = (
  prompt: PromptContent,
  itemInput: Record<string, any> | null,
  variables: string[],
  placeholderNames: string[] = [],
): ChatMessage[] => {
  if (!itemInput) {
    throw Error("Dataset item has no input.");
  }

  const processContent = (content: string) => {
    // Extract only template variables from itemInput (exclude message placeholders)
    const filteredContext = Object.fromEntries(
      Object.entries(itemInput).filter(
        ([key]) => variables.includes(key) && !placeholderNames.includes(key),
      ),
    );

    // Apply template ONLY if the content contains `{{variable}}` pattern
    if (content.includes("{{")) {
      return compileTemplateString(content, filteredContext);
    }
    return content; // Return original content if no placeholders are found
  };
  if (typeof prompt === "string") {
    return [

View on GitHub (pinned to 59d92c7cf3)

Solutions

Clean the dataset: remove or fix items with null input before running the experiment.
Ensure ingested dataset items actually capture the input payload you map variables from.
Adjust the prompt to only use variables present in every item.

Defensive patterns

Strategy: validation

Validate before calling

const usable = items.filter(i => i.input != null && typeof i.input === "object");
if (usable.length < items.length) warnAndSkipNullInputItems();

Type guard

const hasInput = (item: DatasetItem): item is DatasetItem & { input: Record<string, any> } => item.input != null && typeof item.input === "object";

Try / catch

catch (e) { if (e.message === "Dataset item has no input.") reportSkippedItemAndContinue(); }

Prevention

Filter null-input dataset items before creating experiments
Ensure ingestion captures input payloads on dataset items
Keep prompt variables limited to fields present on every item

When it happens

Trigger: Running an experiment whose dataset contains an item with no input (null/empty), while the prompt template declares variables that must be filled from item input.

Common situations: Datasets with mixed items (some input-less), items created from spans without input payloads, or prompts relying on variables the dataset doesn't consistently provide.

Related errors

AI-assisted analysis of langfuse/langfuse@59d92c7cf3 (2026-08-27). Data as JSON: /api/errors/7b1f6117a51ea8e2. Report an issue: GitHub.