# Integrations

## Shyft Time Series

The following is provided as a python package (wheel): `trident-shyft`. It
provides integrations with Shyft's distributed time series system.

The following shows how it connects a Trident engine to a DTS server:

![shyft dtss](shyft-dtss-diagram.svg)

### API Overview

The `trident-shyft` package provides the following components, to be used together
with the Trident engine:

- `ShyftDataService` - factory strategy object
- `ShyftDataProvider` - read-only time series data provider for input data
- `ShyftDataStore` - read-write time series data store for result data

### Basic usage

The following will set up Shyft as the default data provider for time series
data:

```python
import trident.core
import ngltmshyft

engine = trident.core.Engine()
...

shyft_host = "localhost"
shyft_port = 91919
service = ngltmshyft.ShyftDataService()
service.shyft_ref_prefix = "shyft://trident" # will be prepended to each shyft request
service.connect_dts("{shyft_host}:{shyft_port}")
engine.data_service = service
```

### Shyft URL's

When requests are made from Trident through the `ShyftDataService`, Shyft URL's are chosen as follows:

When requesting input data: `<prefix>/datasets/<dataset_id>/<resource_id>`; e.g. `shyft://trident/datasets/my_dataset/my_ts`

When storing result data: `<prefix>/runs/<run_id>/results/<resource_id>`; e.g. `shyft://trident/runs/my_run/results/my_ts`

### More examples

For a more complete set of examples, see the `shyft` folder in the `trident.notebooks` repository.

### Source code

The source code is available in the `trident-shyft` repository.