n8n-nodes-brainiall-documents

Community node for n8n that exposes Brainiall document and speech operations:

1. PDF to Markdown — Brainiall Document engine
2. Extract Fields — structured fields from a document image
3. Transcribe Audio — word-level transcript with timing

This package is only the document/speech product surface. It does not wrap a language-model gateway.

Install

# n8n desktop / self-hosted
npm install n8n-nodes-brainiall-documents

Or in n8n: Settings → Community nodes → Install → n8n-nodes-brainiall-documents.

Credentials

Create Brainiall API credentials:

Every request sends Authorization: Bearer <key>. Without a key the API returns 401 before routing.

Operations

1. PDF to Markdown

Converts a PDF into Markdown text.

Endpoint (verified production contract)
POST https://api.brainiall.com/v1/document/pdf-to-markdown/base64

Request body

{
  "file": "<base64-encoded PDF bytes>",
  "page_range": "1-5"
}

page_range is optional. Omit it to convert the entire document.

Response — plain Markdown text (not a structured JSON document):

# Quarterly Report

## Summary

Revenue grew 12% year over year.

Node inputs

Node output (example)

{
  "markdown": "# Quarterly Report\n\n## Summary\n\nRevenue grew 12% year over year.\n",
  "page_range": "1-5",
  "char_count": 72,
  "engine": "Brainiall Document engine"
}

2. Extract Fields

Extracts structured fields from a PNG or JPEG document image.
PDFs are rejected with a clear error — use PDF to Markdown for PDFs (or rasterize one page first).

Endpoint
POST https://api.brainiall.com/v1/document/extract

Request body

{
  "image": "<base64-encoded PNG or JPEG>",
  "doc_type": "invoice"
}

Response (example)

{
  "request_id": "req_01HZX…",
  "doc_type": "invoice",
  "fields": {
    "invoice_number": "INV-100",
    "vendor_name": "Acme Supplies",
    "total": 42.0,
    "currency": "USD",
    "issue_date": "2026-03-01"
  },
  "text": "Invoice INV-100 …",
  "char_count": 184,
  "extraction_engine": "Brainiall Document Intelligence engine"
}

Node inputs

If a PDF is supplied, the node fails immediately with:

Extract Fields accepts PNG or JPEG images, not PDFs. Use the "PDF to Markdown" operation…

3. Transcribe Audio

Transcribes audio and returns words[] with timing (not segments[]).

Endpoint
POST https://api.brainiall.com/v1/stt/transcribe/base64

Request body

{
  "audio": "<base64-encoded audio>"
}

Response (example)

{
  "words": [
    { "word": "hello", "start": 0.0, "end": 0.4, "confidence": 0.99 },
    { "word": "world", "start": 0.45, "end": 0.9, "confidence": 0.98 }
  ],
  "text": "hello world"
}

Node inputs

Node output (example)

{
  "words": [
    { "word": "hello", "start": 0.0, "end": 0.4, "confidence": 0.99 },
    { "word": "world", "start": 0.45, "end": 0.9, "confidence": 0.98 }
  ],
  "text": "hello world",
  "word_count": 2,
  "engine": "Brainiall Speech engine"
}

Errors

The node never surfaces raw upstream bodies. Failures become short, actionable messages:

Supports n8n Continue On Fail: failed items emit { "error": "…" } and the workflow continues.

First useful result (shortest path)

1. Create a free API key at app.brainiall.com.
2. In n8n, add Brainiall Document → operation PDF to Markdown.
3. Feed any small PDF (or set Base64 Content).
4. Read markdown on the output item — that is the first useful result.

No promotional credit is required or offered by this node.

Develop

npm install
npm test
npm run build

Tests mock the HTTP client. They never call production with a real key.

License

MIT