Skip to main content

DAN AI — Document Extraction Platform

DAN AI turns every uploaded file into a document — a single record holding the original file, the structured fields the AI extracted from it, a confidence score, the AI provider metadata used, and a full audit trail. Try the live product at dan.sdlccorp.com.

Running the AI pipeline over a file to pull out structured fields is called an extraction. This documentation explains every way to start an extraction in DAN AI, what happens at each step, and how to get the data back out.

What DAN AI does​

  • Accepts PDFs, PNGs and JPGs and converts them into structured JSON.
  • Classifies the document into one of 21 supported types (or general as a fallback) so the right prompt is used.
  • Runs the file through a multi-provider AI pipeline — Ollama (primary) → OpenRouter → Groq — with automatic fallback when a provider is unavailable.
  • Stamps each extraction with a per-field confidence score so reviewers know what to spot-check.
  • Exports completed documents as JSON or Excel and pushes webhook events on completion.

Document lifecycle​

Every document, no matter how it was created, moves through the same states:

draft → processing → done   (or → error)

Once a document reaches done, you can review the extracted fields, edit anything the AI got wrong, mark the document as verified, and export it.

Three ways to start an extraction​

MethodBest for
Dashboard uploadOperations staff processing files one at a time through the browser.
Email inbox sync (Gmail / Outlook)Recurring invoice or receipt mail from known senders, picked up on a schedule.
REST APIDevelopers integrating DAN AI into their own apps.

See Creating an Extraction for full walkthroughs of each path.

Key capabilities​

  • 21 tuned document types across finance, legal, logistics, people & ID, health, and general categories — see the Supported Document Types reference.
  • Custom field selection so you can extract only the fields you care about, per-document or account-wide.
  • Account-wide custom prompts (admin only) for non-standard layouts.
  • AES-256-GCM encryption of OAuth refresh tokens at rest.
  • Two export formats: JSON and Excel (.xlsx) with a line-items sheet for invoices, receipts, bank statements and medical reports.
  • Webhooks with signed payloads for every extraction.completed event.

Quick start​

  1. Sign up at dan.sdlccorp.com and confirm your account.
  2. Open Documents → New Extraction and drop a sample PDF (an invoice works well).
  3. Pick the matching document type, optionally tick a subset of fields, and click Extract.
  4. When the document reaches done, review the extracted fields in the two-column detail view, edit anything that's wrong, and click Export → JSON or Export → Excel.
  5. Once you're comfortable, connect a Gmail/Outlook inbox or generate an API key to automate the same flow.

Full setup steps are on the Getting Started page.

What's in this documentation​