n8n OCR Invoice Extractor with DGX Spark & Qwen3.6

Extract invoice data in seconds with n8n OCR + DGX Spark (Qwen3.6)

Turn an uploaded invoice image (PNG/JPG) into structured fields—supplier name, invoice number, invoice date, total, and currency—using an OCR pass on a DGX Spark LiteLLM OpenAI-compatible endpoint, all delivered back through an n8n form completion page.

What this workflow does

• Collects an invoice image via an n8n form (single page upload).
• Converts the upload to Base64 so the image can be embedded in an API request.
• Runs OCR using a DGX Spark-hosted LiteLLM endpoint (one HTTP request that sends the image with an OCR model to produce invoice text).
• Extracts structured invoice fields with Qwen3.6 by sending the OCR text to the same /v1/chat/completions endpoint and requesting JSON output.
• Parses the JSON response and displays the extracted values in the n8n completion page.

Use cases

• Accounts payable automation: capture supplier, invoice number, date, total, and currency from scanned invoices without manual typing.
• SaaS operations workflows: route invoice metadata into downstream processing steps after a user uploads a document.
• Automation engineers: validate and prototype an OCR + field extraction pipeline on your own DGX Spark LiteLLM deployment.

Technical details

• n8n workflow nodes: set, form, sticky note, form trigger, http request, extract from file.
• Integration: DGX Spark LiteLLM via an OpenAI-compatible endpoint (e.g., http://<spark-ip>:4000/v1/chat/completions).
• Models: uses an OCR model and Qwen3.6 (verify model identifiers like ocr and qwen3.6 for your deployment).

Best for n8n users who need reliable invoice OCR and JSON field extraction using DGX Spark—without leaving the n8n experience.