Entity Enricher icon

Entity Enricher

Enrich entities with multi-model LLM fusion, multilingual output, and expertise-driven strategies

Actions18

Schema → Generate Sample

AI-generated

Summary

Generate one or more realistic JSON samples for an entity type from a free-text request, optionally grounded in uploaded source documents. The operation streams a generation job and returns one output item per sample.

Inputs

  • Authentication (required) — Choose API Key or OAuth2 credentials for Entity Enricher. API Key is the default.
  • Connected To Name or ID — Displays the organization and credential associated with the connection.
  • Request — Describe the entity and the properties, size, depth, or structure wanted. Required unless Attachment IDs are provided; with attachments, the document is the request and this text can narrow it.
  • Sample Count — Number of samples to generate (default 1; UI range 1–20). Attachment IDs require this to be 1. The first sample establishes the fields; follow-up samples use the same fields.
  • Typical Instances — Comma-separated concrete entities to use as examples, in sample order, up to Sample Count. Remaining sample names are chosen by the model.
  • Naming Convention — Field naming style: Auto (default), snake_case, or camelCase.
  • Attachment IDs — Comma-separated UUIDs from Add Attachment. Uses source mode to transcribe document content or describe visible photo attributes, and forces one sample.
  • Enable Web Search — When on, requests built-in web search to ground knowledge-mode generation. It is ignored in source mode and for models without search support.
  • Language — Output language code for field names and values. Auto is the default; the server infers it from the request and typical instances, then the attachment, otherwise English.
  • Model — Model to use, or auto (default) to let the server select the organization's default sample-generation model.
  • Timeout (Ms) — Maximum wait for generation to complete; defaults to 300000 ms.

Output shape

A list of items, normally one per generated sample.

Successful items include success, sample, sample_index, samples_generated, samples_requested, and object_type. The first item may also include ambiguity_report, attachment_coherence, and generation cost, token, and processing-time metadata. If generation returns no samples or fails, the operation returns one item with success: false and error_message. If multiple input items reach the node, it runs once per input item and pairs each output with its originating item.

Links

Discussion