Skip to main content

DAN AI — JSON and Excel Exports

A completed document can leave DAN AI in two file formats: JSON and Excel. Both are available from the dashboard (the Export dropdown on the document detail page) and via the API.

For receiving extractions as event callbacks instead of polling/exporting, see Webhooks.

JSON export​

JSON returns the structured fields exactly as the AI returned them, with human edits layered on top.

Where to get it​

  • Dashboard — open a document, click Export → JSON. The file downloads as <reference>.json.
  • API — GET /api/external/v1/documents/:id returns the same payload shape. There's also GET /api/v1/documents/:id/export?format=json for the dashboard-equivalent file response.

Shape​

{
  "fields": {
    "invoice_number": "INV-0001",
    "vendor_name": "Acme Corp",
    "invoice_date": "2026-04-15",
    "total_amount": 1250,
    "currency": "USD"
  },
  "lines": [
    { "description": "Consulting", "quantity": 10, "price_unit": 125 }
  ],
  "confidence": { "invoice_number": 95, "total_amount": 97 },
  "ai": {
    "provider": "ollama",
    "model": "qwen2.5vl:32b",
    "process_time_ms": 3120,
    "total_tokens": 1842
  },
  "meta": {
    "success": true,
    "document_id": "42",
    "reference": "DAN-00042",
    "doc_type": "invoice",
    "state": "done"
  }
}

Verified vs raw values​

By the time you export, fields contains whichever values are current — the original AI output if nobody touched it, or the human-edited values if someone fixed something and marked the document verified.

The original AI response is not included in the export payload to keep it small. If you need it (for audit or to compare against a custom-prompt change), query the document_extractions.raw_response column directly.

Excel export​

Excel returns a .xlsx workbook with up to two sheets:

SheetContents
FieldsEvery top-level field as a column. One row per document. A _confidence column appears next to each field that had a confidence score.
Line ItemsOne row per line item. Populated for invoice, receipt, bank_statement and medical_report. Omitted otherwise.

Where to get it​

  • Dashboard — open a document, click Export → Excel. The file downloads as <reference>.xlsx.
  • API — GET /api/v1/documents/:id/export?format=xlsx returns the binary .xlsx.

When the line-items sheet is empty​

Resumes, contracts, ID cards, shipping labels and other non-tabular types don't have line items — the Line Items sheet is omitted entirely. The export is still valid; only the Fields sheet is present.

Exporting multiple documents​

The dashboard's Documents list supports bulk export: tick the documents you want, click Export → JSON (zip) or Export → Excel (combined).

  • The JSON zip contains one file per document.
  • The Excel combined export merges all selected documents into a single workbook — each document is one row in the Fields sheet and its line items appended to the Line Items sheet (with a document_id / reference column for grouping).

For programmatic bulk export, iterate the documents list endpoint and pull each one individually.

Choosing a format​

Use caseBest format
Importing into an ERP, accounting system, or your own databaseJSON — preserves types, easy to parse.
Sharing with non-developers (accounting team, auditors)Excel — opens in Excel/Numbers/Sheets without extra tooling.
Long-term archival alongside the original fileJSON — stable schema, smaller files.
Quick spot-check during a manual reviewExcel — humans read tables faster than JSON.
  • Webhooks — push completion events instead of polling.
  • REST API — full reference for the document/export endpoints.
  • Confidence Scores — interpret the _confidence columns in the Excel export.