Vision & Document Processing

TinyVision nodes process images, PDFs, and documents to extract text, tables, form fields, and structured data. No OCR setup required; everything runs with AI.

All TinyVision nodes: Color = Purple (#8B5CF6), have test modules.

Available nodes

Common input field

All TinyVision nodes require:

The file URL can come from a previous node (e.g., a webhook payload containing an uploaded file URL) or be a static URL.

Extract Text (OCR)

Extracts all text from an image or scanned document.

Additional fields:

• Enhance (Boolean, default true): apply image enhancement before OCR
• Query (FX formula, optional): specific text to look for

Output: { text: string, confidence: number }

Read PDF

Extracts the full text content of a PDF file.

Output: complete text content across all pages.

Extract Table

Extracts structured table data from images or PDFs containing tables.

Output: rows and columns as structured data; use with For Each to process each row.

Extract Fields

Extracts key-value pairs from forms, invoices, and receipts. Best for documents with labeled fields.

Output: key-value pairs from detected form fields (e.g., { "Total": "$150.00", "Date": "2026-01-15" }).

Ask Document

Ask a natural language question about a document and get an AI-generated answer.

Additional fields:

Example: Upload an invoice PDF, ask "What is the total amount and due date?" The AI reads the document and answers.

Classify

Classifies a document into a type or category.

Output: document classification (e.g., "invoice", "receipt", "contract", "letter").

Common patterns

Invoice processing

Webhook (receive invoice URL) → Extract Fields → If-Else (check amount) → Create Record (log in database)

Document routing

Webhook → Classify → If-Else (route by type) → Different handlers per document type

PDF data extraction

Read PDF → TinyGPT Analyzer (extract structured data from text) → Create Record