Skip to content

Information Extractor

Extract structured information from a text using a LLM

Sends the input text to a language model with instructions describing which fields to pull out, and returns the extracted information. The fields to extract can be described in plain language, or shaped by an optional connected Output Parser dependency for structured output.

Use this node when you need to turn unstructured text (emails, documents, form responses, transcripts) into structured data — for example, pulling a name, date, and amount out of an invoice, or extracting entities from a support ticket for downstream processing.

SettingNotes
TextThe text to extract information from. Rendered as a text area.
What to extractDescribes the fields to extract, or connect an Output Parser. Rendered as a text area.
Options → System messageOptional additional system message appended to the extraction instructions. Rendered as a text area.

Returns { extracted }, containing the model’s structured extraction result (shaped by the connected Output Parser when one is provided).

  • Model (required) — connect a language model dependency (e.g. Chat OpenAI). Used to perform the extraction.
  • Output Parser (optional) — connect an Output Parser dependency to constrain the response to a specific structured shape (e.g. JSON schema) instead of relying solely on the “What to extract” description.

Connect a Chat OpenAI node to the Model dependency, feed text from an upstream node (e.g. an HTTP Request or file read) into Text, describe the fields you want in “What to extract” (or connect an Output Parser for a strict schema), and consume extracted in the next node.

  • If extracted fields are inconsistent, tighten the “What to extract” description or connect an Output Parser to enforce a schema.
  • Verify the Model dependency is connected — the node throws if no model is found.
  • Use the optional System message to add extraction constraints (e.g. output language, date formats) without changing the field description.