File to JSON Extraction Workflow with Groq & Gemini

Extract text (and structured JSON) from files or URLs with Groq + Gemini in n8n

This n8n workflow sub-workflow turns incoming attachments (or a provided URL) into readable text—and, when you supply a schema, converts that text into structured JSON using Groq and Google Gemini (vision for images).

What this workflow does

• Accepts inputs from a parent workflow including binary attachments and an optional extraction object (with field_schemas to define required fields).
• Creates one item per file (or fetches web page text when a URL is provided), then builds a JSON Schema from extraction.field_schemas.
• Routes by MIME type and extracts readable content using native n8n extractors for supported formats such as PDF, JSON, CSV/XLSX, and text.
• Processes images with Gemini (vision) by converting images into a compatible format and extracting visible information into JSON, including content_class and class_confidence.
• Structures extracted text with Groq: if a schema is provided, the workflow sends the extracted text to a Groq-hosted chat model to produce structured JSON that matches your requested fields.
• Returns a normalized result per input, including status, and—when applicable—data: { text, content_class, class_confidence }.
• For unsupported file types, it returns a standardized unsupported status with an error payload.

Use cases

• Turn uploaded documents into JSON-ready data for downstream automation.
• Extract structured fields from PDFs, spreadsheets, and text files without manual parsing.
• Use Gemini vision to capture information from images (e.g., screenshots) and classify confidence scores.
• Fetch a web page via URL and convert the extracted text into structured JSON.

Technical details

• Integrations/Credentials: Google Gemini (PaLM) for vision; Groq API for schema-driven JSON output.
• Node behavior: if, set, code, switch, aggregate, and edit image to orchestrate routing, extraction, and image handling.