Learn to extract, review, and export document data with AI

Dynamite Docs turns business documents such as invoices, receipts, contracts, statements, and forms into structured, editable tables. There is no template setup and no training: you upload a document, the AI infers the fields, and you correct anything it missed. It saves those corrections as patterns that can be reused for a similar layout.

A document open beside extracted table rows in Dynamite DocsClick to enlarge

Extraction is where Dynamite Docs earns its keep. It operates on the core loop of infer → correct → remember:

Steps to run extraction:

The Data Studio panel sits at the bottom of the app. It shows the active extraction as an editable table, the same shape as a spreadsheet but built specifically for documents.

Data Studio Inline Table EditorClick to enlarge

Every extraction can be pulled out of the app in the format you actually use. Exporting never changes the table. It only creates a copy of the current state.

Managing Completed ExtractionsClick to enlarge

Every extraction can be pulled into your own stack. Dynamite Docs exposes a REST API for exporting structured data, outbound webhooks that fire when an extraction completes, and a Google Sheets push so non-technical teammates can stay in the spreadsheet. Together they make Dynamite Docs an infrastructure layer with no lock-in. Bring your own AI keys, or use ours.

Public API & API keys through Dynamite Docs

The public API lives under /api/v1 and is authenticated with a scoped API key sent in either the Authorization: Bearer header or the X-API-Key header. Mint and manage keys anytime directly from the Webhooks & API panel in the workspace header. Scoped keys can list and export extractions, check usage, and submit documents for inference. File, settings, account, and billing changes still require your workspace session.

curl -X POST -H "Authorization: Bearer <session-token>" \
  -H "Content-Type: application/json" \
  -d '{"name":"ci-pipeline"}' \
  https://app.dynamitedocs.com/api/tokens

• Keys look like dd_live_<64 hex>; only their SHA-256 hash is stored, so a leaked database never exposes a working key.
• Manage them with GET /api/tokens (list) and DELETE /api/tokens/:id (revoke) using your session token, or visually in the Webhooks & API modal.

1. Export an extraction (GET /api/v1/extractions/:id/export)

# Standard Bearer authorization:
curl -H "Authorization: Bearer dd_live_<your-key>" \
  "https://app.dynamitedocs.com/api/v1/extractions/<id>/export?format=json"

# Or using the custom X-API-Key header:
curl -H "X-API-Key: dd_live_<your-key>" \
  "https://app.dynamitedocs.com/api/v1/extractions/<id>/export?format=csv"

{
  "id": "ext_01j5x9b3",
  "fileId": "file_8829f0",
  "fileName": "vendor_invoice_1092.pdf",
  "docType": "Invoice",
  "label": "August Operations",
  "folderId": "fld_01h8q2",
  "isVirtual": false,
  "columns": ["Invoice Number", "Date", "Vendor", "Line Total", "Tax", "Amount Due"],
  "rows": [
    ["INV-1092", "2026-08-01", "Acme Industrial Ltd", "$1,250.00", "$100.00", "$1,350.00"]
  ],
  "metadata": {
    "vendor": "Acme Industrial Ltd",
    "invoice_number": "INV-1092",
    "total": "1350.00",
    "currency": "USD"
  },
  "confidence": 0.98,
  "fieldConfidence": { "Invoice Number": 0.99, "Amount Due": 0.98 },
  "summary": "Vendor invoice from Acme Industrial Ltd for precision CNC milling totaling $1,350.00.",
  "createdAt": "2026-08-15T09:12:00.000Z",
  "updatedAt": "2026-08-15T09:12:00.000Z"
}

2. Infer directly from a document (POST /api/v1/infer)

curl -X POST \
  -H "Authorization: Bearer dd_live_<your-key>" \
  -F "file=@invoice.pdf" \
  https://app.dynamitedocs.com/api/v1/infer

{
  "id": "ext_99a8b7",
  "fileId": "file_77c2d1",
  "fileName": "invoice.pdf",
  "docType": "Invoice",
  "columns": ["Invoice #", "Vendor", "Due Date", "Total"],
  "rows": [
    ["INV-4410", "Apex Logistics", "2026-08-30", "$3,420.00"]
  ],
  "metadata": {
    "vendor": "Apex Logistics",
    "total": "3420.00",
    "currency": "USD"
  },
  "confidence": 0.97,
  "createdAt": "2026-08-22T14:10:00.000Z"
}

3. List extractions & check monthly usage

curl -H "Authorization: Bearer dd_live_<your-key>" \
  "https://app.dynamitedocs.com/api/v1/extractions?page=1&limit=20&sort=created_at&order=desc"

curl -H "Authorization: Bearer dd_live_<your-key>" \
  https://app.dynamitedocs.com/api/v1/usage

# Response:
{
  "usage": {
    "month": "2026-08",
    "calls": 42,
    "limit": 3500,
    "remaining": 3458,
    "unlimited": false
  }
}

• Query params: page (default 1), limit (1–100, default 20), sort (created_at | updated_at | label | doc_type), order (asc | desc), doc_type (exact filter). Responses carry a pagination object with total, totalPages, and hasMore.
• Formats: csv (default, with BOM for Excel), json, and xlsx.
• Inference: POST /api/v1/infer uploads a document (multipart/file) and runs the same extraction engine as the app, returning the structured table in the response. It counts as one API call plus the document’s pages against your monthly processing budget; extractions persist and trigger webhooks for the key’s owner.
• Status codes: 200 ok, 401 missing/invalid key, 404 not found, 429 rate limit or monthly quota exceeded. Errors are JSON with a stable code.
• Limits: rate-limited per key (default 60/min) and metered against your monthly API-call allowance. Check it with GET /api/v1/usage (usage checks don’t count toward your allowance). Revoking a key takes effect on the very next request.

Webhooks

Outbound webhooks notify your endpoint when an extraction completes. No polling needed. Set them up from the Webhooks button in the app header. Up to 5 endpoints per account; only public HTTPS URLs are accepted.

• Event: extraction.completed fires after every successful extraction, queue or standalone.
• Signature: each POST carries an HMAC-SHA256 signature in the X-Dynamite-Signature header, computed over the raw body with your webhook secret. Verify it before trusting the payload.
• Reliability: deliveries retry with exponential backoff and jitter on network errors, 429s, and 5xx (up to 3 attempts), with a 10-second timeout. Delivery status is visible in the app.

{
  "event": "extraction.completed",
  "timestamp": "2026-08-13T12:00:00.000Z",
  "data": {
    "extractionId": "…",
    "fileId": "…",
    "docType": "Invoice",
    "columns": ["Invoice #", "Date", "Total"],
    "rows": [[…]],
    "metadata": {…},
    "confidence": 0.97
  }
}

Google Sheets push

Connect Google Sheets once from the app header, then push any extraction with one click from the Data Studio export menu. Each push lands on a fresh tab named after the document type, and a separate “Document Fields” tab carries the metadata. Pushes reuse your last-used spreadsheet and never overwrite earlier tabs.

Google Drive import & export

The Google Picker integration lets you select importable files from your Google Drive with multi-file selection, then streams only the files you choose into the normal upload pipeline. Completed extractions can be saved back to Drive as XLSX, CSV, or JSON files from the Data Studio export menu.

Folders group your documents and extractions into a navigable tree, and labels add an extra layer of categorization on top. Use them together to keep even large libraries structured and accessible.

Dynamite Docs ships with seven hosted AI providers: Mistral, Gemini, OpenAI, Anthropic, Groq, OpenRouter, and Cloudflare. Extraction is routed to the best available provider automatically, so there is nothing to configure to get started.

AI Configuration Panel & Models ListClick to enlarge

Local Ollama companion (on-device model inference)

For documents that cannot leave your workstation due to client confidentiality, HIPAA, or strict internal data residency rules, Dynamite Docs supports local model extraction through Ollama. The Dynamite Docs Windows companion connects your web workspace to the local extraction service on loopback port 8756. Local model requests go to Ollama at port 11434 without calling a hosted AI provider. Local model inference runs on your computer; account, token, and workspace features remain hosted. Choose browser-only file storage when the source should stay on your computer.

Keep the companion and Ollama running while you extract documents. Sign in, import files, review results, and export from your browser workspace. The workspace and account services require internet access.

• Available on Starter, Pro, and Scale plans.
• Requires Ollama installed on your machine with a multimodal vision model like llama3.2-vision or minicpm-v.
• Read the comprehensive Local Document AI guide for full hardware recommendations, quantization choices, and runtime benchmarks.
• 

Choose storage and model processing separately. Review the selected provider and your active routing rules before processing sensitive documents.

For stricter needs, AI Settings exposes routing rules that lock extraction to the providers you trust:

• Sensitive Documents mode: only use models that do not train on your data. Gemini's free tier is excluded by default.
• No-training only: block any model that may retain or train on submitted content (Gemini, OpenRouter).
• Free models only: spend only your connected free keys, never host or paid quota.
• Local only: route every extraction through the Windows companion (127.0.0.1:8756) to Ollama instead of an external AI provider. See our Local Document AI guide for setup details.

The table shows each family’s entry package. Starter offers 100 or 500 PE, Pro offers 1,000 or 2,000 PE, and Scale offers 5,000 PE per month. Compare all subscription packages.

• Hosted extraction stops at the monthly PE allowance, so there are no surprise overage bills. If you need more capacity, email us at support@dynamitedocs.com or contact support.
• BYOK provider models are available on every plan; Pro and Scale use them without a PE cap.
• Anonymous (not signed in) usage is limited to a daily anti-abuse cap.

Shortcuts work when the focus is outside an input field. On macOS, use ⌘ instead of Ctrl.

If the guide does not solve the issue, email us with the relevant document details, forward files to email capture, or open the support desk from the app.