​

Core Concepts

This page explains key concepts and terminology used across the Anyformat API.

​

Workflows

A Workflow defines what information should be extracted from your documents. Think of it as a template or schema for data extraction. Each workflow contains:

• A unique identifier
• A name and description
• A set of fields that define what data to extract
• Optional manual fields for user-provided information

​

Fields

Fields define specific data points to extract from your documents. Each field has:

• A name (e.g., “invoice_number”)
• A description that helps the AI understand what to extract
• A data type (string, integer, float, date, etc.) - see Field Types for details

​

Extractions

An Extraction is the process and result of applying a workflow to a document. When you submit a document to be processed with a workflow, an extraction job is created, which contains:

• A unique job ID
• The workflow ID used for extraction
• Status information (pending, in_progress, processed, error)
• Extracted data points (when processing is complete)

​

Results

Results are the structured data extracted from your documents. Each extracted data point includes:

• The field name
• The extracted value
• A confidence score (0-100)
• Evidence information (location in the document where the data was found)

​

Evidence

Evidence is an array of metadata objects that indicate where in the document pieces of information were found. This is an array because depending on the type of information one is looking for, sometimes it is inferred instead of directly extracted from a concrete place. The evidence array is therefore way more useful when it comes to validating the results! Each evidence object provides:

• The actual snippet of text from which the data was extracted
• The page number in the document

​

Confidence

The Confidence score (0-100) indicates how certain the system is about an extracted value. Higher scores indicate greater confidence in the extraction accuracy.

​

Additional Resources

For detailed information about specific topics, see:

• Response Formats - Different output formats and the as_lists=true parameter
• Error Handling - Complete error handling guide with examples
• Workflow Creation - Creating and managing workflows