Process AI Output to Structured JSON with Robust JSON Parser

How it works

This workflow is a robust and forgiving JSON parser designed to handle malformed or "dirty" JSON strings often returned by AI models or scraped from web pages. It takes a text string as input and attempts to extract and parse a valid JSON object from it.

1. Cleans Input: It starts by trimming whitespace and removing common Markdown code fences (like ```json) that might wrap the JSON.
2. Applies Multiple Fixes: It systematically attempts to correct common JSON errors in a specific order:
   • Escapes unescaped control characters (like newlines) within strings.
   • Fixes invalid backslash escape sequences.
   • Removes trailing commas.
   • Intelligently attempts to fix unescaped double quotes inside string values.
3. Parses Strategically: If a direct parse fails, it tries to extract a potential JSON object from the text (e.g., finding a {...} block inside a larger sentence) and then re-applies the cleaning logic to that extracted portion.
4. Outputs Clean Data: If successful, it outputs the parsed JSON fields. By default, it removes the detailed parsing_status object, but you can deactivate the final "Set" node to keep it for debugging.

Set up steps

Setup time: ~1 minute

This workflow is designed to be used as a sub-workflow and requires no internal setup.

1. In your main workflow, add an Execute Sub-Workflow node where you need to parse a messy JSON string.
2. In the Workflow parameter, select this "Robust JSON Parser" workflow.
3. Ensure the data you send to the node is a JSON object containing a text field, where the value of text is the string you want to parse. For example: { "text": "{\\\"key\\\": \\\"some broken json...\\\"}" }.
4. The workflow will return the successfully parsed data. To see a detailed log of the cleaning process, simply deactivate the final Remove parsing_status node inside this workflow.