---
title: Manage services
url: https://docs.triply.cc/how-to/manage-services/
description: Create a Virtuoso, Jena or Elasticsearch service for a dataset, keep it in step with its
  data, and manage every service on the instance.
---

# Manage services

A **service** runs a dedicated engine over a copy of a dataset: a Virtuoso or Jena SPARQL endpoint,
or an Elasticsearch text index. SPARQL needs none, because **Speedy**, TriplyDB's built-in engine,
queries every dataset without setup. See [Service](https://docs.triply.cc/concepts/what-triplydb-is/#service) for
when a service is worth creating, and why one can fall behind its data.

This guide covers creating and maintaining services for a dataset, and managing every service on
the instance as an administrator. It needs permission to manage the dataset's services; without
it, the Services page does not appear until the dataset has a service.

## Create a service

1. Open the dataset and click **Services**, under **Ecosystem** in the left-hand sidebar. The page
   lists **Speedy** first, then the services the dataset already has.
2. Click **Create new service**. The button is disabled until the dataset holds at least one graph.
3. Enter a **Name** for the service.
4. Choose a **Service type**. An instance offers the types it is configured for:

    | Type | What it gives the dataset |
    | --- | --- |
    | SPARQL – Virtuoso | A SPARQL endpoint run by Virtuoso |
    | SPARQL – Jena | A SPARQL endpoint run by Apache Jena, optionally with reasoning |
    | Elasticsearch | A text index, searched from **Elasticsearch** in the sidebar |

5. For a Jena service, choose a **Reasoner**: **None**, **RDFS** or **OWL**.
6. Click **Create service**.

![The "Create service" dialog, offering three service types](https://docs.triply.cc/assets/create-service.png)

Instance administrators also see a **Memory limit** field, which takes a size in bytes or in `mb`/`gb`
notation; left blank, the limit is set automatically.

## Keep a service in step with its data

Each service on the Services page has its own card, showing its type, status, protocols and the
number of statements loaded.

![A service widget](https://docs.triply.cc/assets/service-widget.png)

- **To synchronise a service** after its dataset changed, follow the link ending in "Click here to
  update." that appears on its card. The service has to be running before it can be updated.
- **To start or stop a service**, click **Start** or **Stop** on its card.
- **To rename a service**, use **Edit service name** on its card.
- **To remove a service**, use **Remove service** on its card, and confirm.

A query on a service that returns nothing, or old results, usually means a service that has
stopped or has not been synchronised; see
[Why a service falls out of step](https://docs.triply.cc/concepts/what-triplydb-is/#why-a-service-falls-out-of-step).

## Set up Elasticsearch index templates

An Elasticsearch service answers two kinds of search. A *simple* search takes a single term and builds a fuzzy query from it; a *custom* search accepts a JSON body in [the Elasticsearch query DSL](https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl.html). The Search endpoints in the [API reference](https://api.triplydb.com/api-docs) describe both.

By default the service indexes every literal with Elasticsearch's own defaults. [Index templates](https://www.elastic.co/guide/en/elasticsearch/reference/current/index-templates.html) and [component templates](https://www.elastic.co/guide/en/elasticsearch/reference/current/indices-component-template.html) let you say more than that — for example that a property holds a date, so that it can be sorted or used in a range query.

Templates are passed in `config` when the service is created through the API, and are read only then: changing one means creating the service again. Three constraints are specific to TriplyDB:

- The service builds a single index named `index`, so every index template must have `"index_patterns": "index"`. A template matching anything else is silently unused.
- A component template only takes effect once an index template names it in `composed_of`, so the two are always created together.
- A property's field name is its IRI with every `.` replaced by a space, because Elasticsearch reads a dot in a field name as a path separator. Write mappings that way, and expect hits that way.

The example below maps `https://schema.org/dateCreated` to a date:

```sh
curl -H "Authorization: Bearer TRIPLYDB_TOKEN" -H "Content-Type: application/json" \
  -X POST "https://api.INSTANCE/datasets/ACCOUNT/DATASET/services" \
  -d '{
    "type": "elasticSearch",
    "name": "SERVICE_NAME",
    "config": {
      "indexTemplates": [
        { "name": "my-template", "index_patterns": "index", "composed_of": ["dates"] }
      ],
      "componentTemplates": [
        { "name": "dates",
          "template": { "mappings": { "properties": { "https://schema org/dateCreated": { "type": "date" } } } } }
      ]
    }
  }'
```

The equivalent in TriplyDB.js is [`Dataset.addService`](https://static.triply.cc/triplydb-js/classes/Dataset.html#addservice).

## Manage every service on the instance

Administrators can update, synchronise or remove any service on the instance, one at a time or in
bulk, from the Services page of the admin settings. See
[Services page](https://docs.triply.cc/reference/admin-settings/#services-page) for its columns and actions.

## Related

- [Publish data](https://docs.triply.cc/how-to/share-and-publish/) — publishing a dataset publishes its services
- [Query data](https://docs.triply.cc/how-to/query-data/) — the SPARQL IDE and text search that services enable
- [SPARQL: the basics](https://docs.triply.cc/concepts/sparql-basics/) — what a SPARQL service is for