In development

Many sources in. One record out.

FutureDataHub is built to merge tourism data from feratel, Outdooractive, contwise and your editorial team, and to deliver it openly.

GET /v1/feeds/events.jsonld
{
  "@type": "Event",
  "name": { "de": "Almfest Alpenmitte" },
  "startDate": "2027-07-17T11:00:00+02:00",
  "meta": {
    "source": "redaktion",
    "license": "CC-BY-4.0",
    "open_data": true,
    "approval_status": "approved"
  }
}
Excerpt from the API contract with synthetic test data: provenance and license travel with every record.

Measured, not promised

9connectors in the code
GeoSphere Austria, ODTA feeds, iCal and CSV are tested with test data; feratel DS-API and DSI, contwise, Outdooractive and destination.one await source access
15 msat 300 requests per second
95 % of read requests faster, 40,000 records, one API process
0false merges
120 entity resolution test pairs, 36 of 40 duplicates detected automatically
18.6 kBfor all widgets
gzip, without the map; the map only loads when needed

Measured on our development and test systems with synthetic data, October 2026. Real source access follows with the pilot.

How your data gets into shape

Four steps, each one traceable. Nothing is overwritten without you knowing where it came from.

  1. Sources

    Every source runs in its own connector container following the open Airbyte protocol. The original is stored unchanged.

  2. Entity resolution

    The same mountain hut from three systems? Name, location and address are compared. Uncertain cases become a review task for you.

  3. Golden record

    One record per thing, assembled field by field. Every field knows its source, retrieval time and license. Your editorial team wins.

  4. API, widgets, feeds

    REST, GraphQL, JSON-LD, webhooks, widgets, DCAT and an MCP server for AI agents. All from the same data.

Architecture overviewData sources such as feratel, contwise, Outdooractive, weather services and CSV or iCal are each connected through an isolated connector. Data passes three layers: raw stores it unchanged, staging normalises it to ODTA, core builds the golden record with field provenance and a license per field. Delivery runs via REST, GraphQL, JSON-LD, webhooks, widgets and an MCP server.SourcesConnectorsData layersDeliveryferatelcontwiseOutdooractiveWeather servicesCSV, iCalone isolatedcontainerper sourcerawunchangedstagingnormalised (ODTA)coregolden recordRESTGraphQLJSON-LDWebhooksWidgetsMCP server

What's inside

  • Provenance per field

    The German name from feratel, the English one from Outdooractive, the description from your editors: every field knows where it came from and under which license.

    1. Editorial
    2. feratel DS-API
    3. feratel DSI
    4. contwise
    5. Outdooractive
    6. ODTA feeds
    7. iCal
  • Tenants like real life

    State -> region -> town -> business. Every level sees its own part, enforced in the database with row level security.

  • Editorial included

    Maintain events and POIs directly, with approval and roles. What you maintain yourself always beats imports.

  • Review tasks, not guesswork

    When entity resolution is unsure, a person decides. Only clear cases are merged automatically.

  • Open data in one step

    Approved open data is published as JSON-LD feeds and a DCAT catalogue for portals such as data.gv.at or data.europa.eu.

From silo to hub

feratel, destination.one and Outdooractive are strong systems for their purpose. Each one keeps its data in its own world with its own interface.

Alpine valley at dusk seen from above, lines of light connect villages and huts to one point in the middle of the valley.
Many places, one shared data set. AI-generated image

Today

  • Every website, app and map connects to every system separately.
  • The same hut appears three times online, with three different opening hours.
  • Who may pass on which data under which license is recorded in no record.

With FutureDataHub

  • Your existing systems stay. The hub reads them and merges them.
  • One golden record per thing; duplicates are detected or handed to you for review.
  • Provenance and license sit on every field, in delivery too.

Comparison with sources

For every level of tourism

Alpine valley in summer with a village and church tower, mountain huts on the slopes and limestone peaks behind.
State, region, town: every level maintains its own part. AI-generated image
State
A statewide view across all regions, open data as a DCAT catalogue for open data portals, uniform interfaces for partners and tenders.
Region
Bundle the data of your towns, connect your own sources and deliver them with widgets on the regional website. Every town keeps its own area.
Tourism board
Maintain events and POIs yourself, even without an editorial system of your own. Mixed operation with feratel works, import from destination.one included.

More for regions and boards

AI that cites its sources

AI is part of the core and follows rules: no answer without a source, personal data is removed before every model call, suggestions instead of silent changes.

  • Hybrid search

    built

    Combines keyword and semantic search. Keyword search works today; how well semantic search matches will be measured with the chosen model before the pilot.

  • Answers with citations

    built

    Every answer quotes your records word for word. If no source remains, there is no answer.

  • MCP server for AI agents

    built

    Assistants such as Claude read your data via the Model Context Protocol, with provenance and license.

  • Translation and alt texts

    in progress

    As suggestions, with a glossary per tenant. Applied only after review.

  • Quality score

    planned

    A value from 0 to 100 per record, with hints on what is missing.

  • Capture from documents

    planned

    Programmes and event lists become draft entries.

POST /v1/ai/answer
{
  "answer": "Das Almfest Alpenmitte findet am 17. Juli 2027 auf der Hochalm statt, mit Blasmusik und Kinderprogramm.",
  "sources": [{
    "name": "Almfest Alpenmitte",
    "field": "description@de",
    "quote": "Traditionelles Almfest mit Blasmusik, regionalen Schmankerln und Kinderprogramm auf der Hochalm.",
    "license": "CC-BY-4.0"
  }],
  "disclosure": "Diese Antwort wurde automatisch mit KI erstellt."
}
Sample answer from the API contract (shortened, German test data): quote, license and AI notice always come along.

In the test environment the AI features run against a test model. The model provider will be chosen before the pilot.

Open and under your control

Open source

The complete hub is to be released under AGPL-3.0, including editorial interface, connectors and APIs. The license setup is currently under legal review.

Run in Innsbruck or by you

We offer operation in our own data centre in Innsbruck as a service; pilot operation is being prepared. You can also run the hub yourself; an installation package for Docker Compose is planned.

No third-country services

Database, object storage, map tiles, sign-in and error monitoring are self-hosted. Fonts and scripts come from your own server.

Standards, not lock-in

Data model based on ODTA and schema.org, OpenAPI 3.1, JSON-LD, DCAT and OIDC. Your data stays portable.

Open source and service

Built for developers

Contracts first: data model, OpenAPI and GraphQL schema are fixed before code is written. And webhooks are signed with HMAC-SHA256.

  • REST · OpenAPI 3.1
  • GraphQL
  • JSON-LD · schema.org · ODTA
  • DCAT
  • Webhooks
  • Web Components
  • MCP

Everything for developers

Terminal
# Events as JSON-LD
curl -H "Accept: application/ld+json" "https://<host>/v1/events?limit=2"

# Hybrid search
curl "https://<host>/v1/search?q=Almfest&limit=5"

# Open data as a DCAT catalogue
curl "https://<host>/v1/catalog.dcat"
index.html
<script type="module" src="/assets/hub-widgets.js"></script>
<hub-event-list api="https://<host>" lang="de" limit="6"></hub-event-list>

Where we are right now

  1. Phase 0, Month 1-2

    Foundation

    completed

  2. Phase 1, Month 2-4

    Ingestion MVP

    built, pilot pending

  3. Phase 2, Month 4-6

    Delivery

    built, pilot pending

  4. Phase 3, Month 6-8

    AI layer

    in progress

  5. Phase 4, Month 8-10

    Pilot and release

    planned

As of:

View roadmap

Frequently asked questions

Does FutureDataHub replace feratel or destination.one?

Not necessarily. You can keep using your existing systems; the hub reads them through connectors once source access is in place, and merges the data. For tourism boards without an editorial system the hub can be the leading system itself. For switching from destination.one there is an import.

Where is my data stored?

Either in the data centre of web-crossing GmbH in Innsbruck or on your own infrastructure. Database, storage and sign-in are self-hosted, without third-country services.

Which standards are supported?

The data model follows ODTA and schema.org. Delivery runs via REST with OpenAPI 3.1, GraphQL, JSON-LD, webhooks, DCAT and the Model Context Protocol (MCP).

What if the AI gets it wrong?

Translations, alt texts and captured entries are suggestions that a person reviews and applies. Answers quote the records word for word and carry a notice that they were generated by AI.

What does it cost?

The core is meant to be open source and free to use. Operation, premium connectors and AI modules are offered as a service. We will publish prices once the license and service model is approved.

When can I start?

The platform is in development and a pilot is being prepared. We are looking for regions and tourism boards who want to have a say from the start.

Who is behind it?

FutureDataHub is developed by web-crossing GmbH, a digital agency in Innsbruck for web development, hosting and digital marketing since 1999.

Let's talk about your data.

Want to know whether it fits your region, or join as a pilot? Write to us or give us a call.