This workflow demonstrates how to reliably generate structured JSON output from a large language model (LLM) and transform it into usable, structured data in KNIME.
The workflow uses a small set of customer feedback examples and shows how to:
instruct an LLM to return valid JSON using an explicit prompt schema
understand that LLM JSON output is text, not a native JSON type
convert JSON-formatted text into a KNIME JSON column
extract fields from JSON into regular table columns
This pattern is useful whenever LLM output needs to be parsed, validated, or integrated into automated workflows such as dashboards, data pipelines, or agent tools.