olaverse

n8n node for Olaverse: document extraction with calibrated confidence, decisions, language ID in 608 languages, diacritics and question generation in 324 languages

Package Information

Released: this month
Downloads: 0 weekly / 0 monthly
Latest Version: 0.2.2
Author: Olaverse Ltd

Documentation

n8n-nodes-olaverse

An n8n community node for Olaverse Lab: document extraction with
calibrated confidence, decisions, language identification in 608 languages, diacritic restoration and question
generation in 324 languages: multilingual, including the languages other tools skip.

Turn receipts, invoices, mobile-money messages and forms into structured fields in your workflows (from an email
attachment straight into a spreadsheet), triage support messages, and route text by language.

Install

In n8n: Settings → Community Nodes → Install, and enter n8n-nodes-olaverse.
See n8n's community nodes guide.

Credentials

Create an API key in the Olaverse console (new accounts get $1 of free credit),
then add an Olaverse API credential in n8n with that key. Testing the credential makes one tiny language check
(about $0.0003).

Operations

Document

  • Extract: turn a document into JSON shaped like your template, with a calibrated confidence (0–1) for every
    field. The document can be text, or a JPEG/PNG image from a previous node's binary data (an email
    attachment, an uploaded photo or a scan). Text over 120,000 characters runs as a long-document job: split,
    extracted in parallel and merged. For a PDF, extract its text first (for example with n8n's Extract from File
    node) and pass it as text.

    Template leaves name a type: verbatim-string, string, number, integer, date, date-time, boolean,
    ["a", "b"] (pick one), ["string"] (a list) or [{...}] (a list of objects, e.g. line items).

Text

  • Decide: answer typed questions about a message or record (noul yes/no, choice, score) with a calibrated
    probability for every answer. 19 languages, including English, Pidgin, Hausa, Yorùbá, Igbo, Swahili and Arabic.
  • Detect Language: the language of a text among 608, accurate on very short text.
  • Restore Diacritics: tone marks, accents and Arabic vowel marks back in Yorùbá, Igbo, Hausa, Arabic and 7 more.
  • Generate Questions: the questions a passage answers, in 324 languages, in its own language or another (Question
    Language); a set number from 1 to 5, or let the model decide how many.

Example workflows

Receipts from email to a spreadsheet

  1. Gmail Trigger (or Email Trigger (IMAP)): new emails with attachments, with Download Attachments on.
  2. Olaverse → Document → Extract: Input Image (Binary), field attachment_0, template
    {"merchant": "verbatim-string", "date": "date", "total": "number", "currency": "verbatim-string"}.
  3. IF: {{ $json.confidence.total }} is at least 0.95.
  4. Google Sheets → Append row with {{ $json.data.merchant }}, {{ $json.data.date }}, {{ $json.data.total }}
    on the true branch; a Slack message asking someone to check it on the false branch.

Support messages triaged by language and intent

  1. Webhook (or your helpdesk's trigger): a new message.
  2. Olaverse → Text → Detect Language, on the message text.
  3. Olaverse → Text → Decide, with State {{ $('Webhook').item.json.body.message }} and questions
    {"refund": {"type": "noul", "instructions": "The customer is asking for money back."}, "team": {"type": "choice", "instructions": "Which team should handle this?", "criteria": {"billing": "payments and refunds", "delivery": "orders in transit"}}}.
  4. Switch on {{ $json.answers.team.choice }} to route the ticket; flag it as urgent when
    {{ $json.answers.refund.noul }} is above 0.9.

Using the confidence

Every extracted field comes with a confidence. A common pattern: an IF node after Olaverse that sends fields
under 0.95 to a person for review and lets the rest through. Check a threshold on a sample of your own documents
before automating on it.

Pricing

Calls are billed to your Olaverse key at the usual prices. A
one-page receipt costs about $0.0015.

Resources

Licence

MIT

Discussion