Tiki

The Metadata Problem

You have a file. You want to publish it as open data. Between those two things lies a form — and the form wants to know the title, description, theme, licence, publisher, update frequency, and a dozen other things you either don’t know or don’t have time to look up.

So the file sits in a folder. Unpublished. Undiscoverable. Not because the data isn’t valuable, but because the metadata barrier is too high.

Tiki removes the barrier. Drop the file. Walk away with a standards-compliant DCAT-AP record.

Imagine you want to share a book with a library. The library needs a card that says what the book is about, who wrote it, when it was published, and what language it’s in. Filling in that card is annoying.

Tiki reads the book and fills in most of the card for you. You just check it over and add anything it couldn’t figure out.

1. Drop a file

Drag any file onto the Tiki interface — Word document, PDF, spreadsheet, image, CSV.

2. Tika extracts metadata

Apache Tika analyses the file and returns: MIME type, language, author, created and modified dates, and title from document properties.

3. Claude enriches the record

The extracted text is passed to Claude, which suggests a DCAT theme from the EU controlled vocabulary, generates a description if none exists, and proposes keywords.

4. Review and complete

The UI shows the assembled DCAT record. Fields that can’t be inferred — licence, publisher — are flagged for completion. Edit in place.

5. Export

Copy or download the valid DCAT-AP JSON-LD record. Or call the API directly from your own pipeline.

Understanding DCAT

For anyone learning about open data standards, Tiki makes DCAT-AP tangible. Drop a real file, see what a valid metadata record looks like, understand which fields matter and why.

Data portal onboarding

Any organisation onboarding datasets to a CKAN or similar portal can use Tiki to generate the metadata record before submission.

Metadata quality improvement

Existing datasets with poor or missing metadata can be run through Tiki to generate richer, standards-compliant records.

Pipeline integration

The API endpoint accepts any file and returns DCAT JSON-LD. Integrate into any publishing workflow via a single HTTP call.

Stack

Django backend. Apache Tika server for content extraction. Anthropic Claude API for enrichment. Tailwind CSS via CDN. No JavaScript frameworks.

API

POST /api/enrich/ accepts a multipart file upload and returns DCAT-AP JSON-LD. Auth via API key in x-api-key header.

Output format

Valid DCAT-AP JSON-LD with full @context namespace mappings. Compatible with CKAN ckanext-dcat harvester.

Deployment

Docker. No database required — stateless enrichment service.