Skip to content
RDA CLI Tooling
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


A cli for accessing the RDA API.

For more information on the RDA API, see the documentation here and here.


The rda command line tool is a Go based executable for accessing RDA functionality. As such, it is a statically linked executable and should run without hassle on most systems.


To install rda, navigate to releases page here and download the most recent package for your operating system (note that Darwin is Max OSX). Unpack your download and you will find a binary executable named rda. Place this in your path so that you can access it from the command line wherever you're at, or run it directly from where you downloaded it.

If you want to build it yourself, make sure you have Go available on your system (see here) and run go get -u This should clone the repository and install the rda tool in your Go path (by default $HOME/go/bin on linux/osx).

Using rda

In general, rda --help is your guide, and note that --help works for all subcommands as well. You can find what version of rda you're running via rda --version.

You can also use the --debug flag for any of the commands. When provided, the tool will log to stderr information on the http requests and responses being made. If you encounter a bug, you may try this option to try to get a better idea of the HTTP requests being made and their responses and if the issue is in the cli or with RDA.

Some of the commands return JSON responses; formatting JSON is easy by piping the output of the command to jq or python -m json.tool. For instance, rda operator DigitalGlobeStrip1B | jq yields nicely formatted JSON describing that operator.

rda configure

The first time you use rda, you need to configure it to store your GBDX credentials (or set the environment variables GBDX_USERNAME and GBDX_PASSWORD). Once configured, rda will cache your GBDX token and refresh it on demand for you without you needing to intervene. Simply type

rda configure

and provide the requested GBDX credentials. Note that it is possible to provide a --profile if you have more than one set of GBDX credentials (this is similar to how the AWS cli behaves). If --profile is not provided, default is used. All subcommands support --profile.

rda token

This will return to you a valid GBDX token. If the cached one is not set or expired, it will be refreshed before returned to you. This is nice if you want to use curl or postman and need a token ASAP.

rda operator

Returns JSON describing all the RDA operators available. To get information on a single operator, you just specify the name, e.g. rda operator DigitalGlobeStrip. JSON is returned so you may want to pipe the output to a formatting tool.

rda stripinfo

rda stripinfo returns JSON desribing the given catalog. For instance, rda stripinfo 103001000EBC3C00 will give you all the information you might want about 103001000EBC3C00. Remember to pipe this to a JSON formatter if you want a pretty view of it.

In addition, you can provide a --zipfile <location of zip> (IMD files, etc) to download the original metadata that came with the imagery when provided by DG's internal factory.

Currently, the way to check if a strip is available in RDA is via this command and seeing if you recieve a 404 Not Found upon request.

rda dgstrip

rda dgstrip is a subcommand offering the additional commands metadata, realize, and batch, described below.

rda dgstrip metadata

rda dgstrip metadata will return JSON describing what you can download via rda dgstrip realize or rda dgstrip batch. Try it out via

rda dgstrip metadata 103001000EBC3C00 --gsd 0.000146 --dra --bands RGB --bandtype PS --crs EPSG:4326

and you'll get a response desribing how large 103001000EBC3C00 is in a geographic projection (EPSG:4326 is the code for a lat/long projection; you can specify any that you like via the --crs flag; the default projection is UTM in the zone that strip is located).

rda dgstrip realize

This command concurrently downloads and creates a VRT composed of tiles realized from RDA. Note that you can provide either --srcwin or --projwin flags if you don't want to download an entire image, which is usually a good idea, as large images are best aquired using rda dgstrip batch. rda dgstrip realize downloads all tiles that intersect the provided window where that window intersects the global image window.

Try out rda dgstrip realize --help to see the flags you can provide to control how the image is processed.

For example,

rda dgstrip realize 103001000EBC3C00 103001000EBC3C00-ovr.vrt --gsd 0.000146 --dra --bands RGB --bandtype PS --crs EPSG:4326 --projwin -116.79,37.86,-116.70,37.78

Will return a downsampled version of catalog id 103001000EBC3C00 to you as a VRT. Just load it up into QGIS/ArcGIS/your favorite viewer that can read VRTs and profit!

The actual tiles are stored in a directory named 103001000EBC3C00 adjacent to the VRT. The VRT format is an xml based format that describes how to lay out the tiles as if they were a single image. You can create a single geotiff out of the downloaded product via GDAL, e.g. gdal_translate 103001000EBC3C00.vrt 103001000EBC3C00.tif should do it if you have GDAL installed.

rda dgstrip batch

batch takes all the same flags as realize (except the vrt location), but rather than realize the tiles it submits a batch materialization request to RDA. You will get a response that includes a job id, which as you'll see below you can use to status and download the output of the batch materialization job. For example, running

rda dgstrip batch 103001000EBC3C00 --gsd 0.000146 --dra --bands RGB --bandtype PS --crs EPSG:4326 --projwin -116.79,37.86,-116.70,37.78

The output of this is a json message, that includes a field "jobId" whos values you can use as described below.

rda dg1b

rda dg1b is a subcommand allowing access to DigitalGlobe 1B images. 1Bs are unrectified imagery (and hence not georeferenced) often used in algorithms that exploit the camera perspective (e.g. stereo matching) or when one wants to use a custom elevation model during orthorectification.

You can inquire about the image parts, metadata, and realize image parts from RDA with this subcommand. Note that you cannot currently use RDA's batch materialization capabilities for ungeoreferenced imagery, so realize is your only option for downloading 1B imagery.

rda dg1b parts

parts returns a description of the image parts that compose a given catalog ID. For instance, running

rda dg1b parts 5bf6f01d-ef58-450c-8a68-48a03d0cabb6-inv


  "pan": {
    "numParts": 7,
    "imageIDs": [
  "vnir": {
    "numParts": 7,
    "imageIDs": [

which tells us that 5bf6f01d-ef58-450c-8a68-48a03d0cabb6-inv has pan and vnir bands, each with 7 parts. The image IDs are the unique IDs RDA uses internally to track the location of the underlying images.

rda dg1b metadata

Just like rda dgstrip metadata, this subcommand returns the RDA metadata that describes an image. You must provide a catalog id, band (pan, vnir, swir, or cavis), and a part number (starting at 1), in that order. For example,

rda dg1b metadata 5bf6f01d-ef58-450c-8a68-48a03d0cabb6-inv pan 3

will return information describing the size of the image, data type, and so on.

rda dg1b realize

Note that this command may go through more refinement as we test to see if its outputs work well with consuming applications.

realize downloads all the tiles that compose a 1B image part, just like rda dgstrip realize. In addition, it will also download the original factory metadata that is associated with that image part. It is invoked just like metadata above, but you provide an output directory as the last argument. For example,

rda dg1b realize 5bf6f01d-ef58-450c-8a68-48a03d0cabb6-inv pan 3 ~/Downloads/1B/part3

populates ~/Downloads/1B/part3 with


where PAN_P003.vrt is a VRT stitching together all the individual tiles downloaded to the tiles/ directory.

rda template

rda template provides access to a more generic set of capabilities related to RDA templates. In fact, rda dgstrip and rda dg1b are just user friendly entry points to specific RDA templates.

rda template describe

Given an RDA template ID, returns a description of the given RDA template. For example

rda template describe DigitalGlobeStrip


  "defaultNodeId": "SmartBandSelect",
  "edges": [
      "id": "edge-0",
      "index": 1,
      "source": "DigitalGlobeStrip",
      "destination": "UniversalDRA"
      "id": "edge-1",
      "index": 1,
      "source": "UniversalDRA",
      "destination": "SmartBandSelect"
  "nodes": [
      "id": "DigitalGlobeStrip",
      "operator": "DigitalGlobeStrip",
      "parameters": {
        "CRS": "${crs:-UTM}",
        "GSD": "${GSD}",
        "bands": "${bands:-MS}",
        "catId": "${catalogId}",
        "correctionType": "${correctionType:-DN}",
        "fallbackToTOA": "${fallbackToTOA}"
      "id": "UniversalDRA",
      "operator": "UniversalDRA",
      "parameters": {
        "draType": "${draType:-None}"
      "id": "SmartBandSelect",
      "operator": "SmartBandSelect",
      "parameters": {
        "bandSelection": "${bandSelection:-All}"

Note the ${...} fields are RDA template parameters (if followed by :-, the following value indicates the default value); we will manipulate these below.

rda template upload

rda template upload uploads a json file desribing an RDA template. We validate that the template is a DAG, and will auto select a default evaluation node (see the field "defaultNodeId": "SmartBandSelect" in the example template above) for you that is the furthest distance from the source nodes in the graph if not specified. Specify "defaultNodeId" if you'd like control over the default node.

Lets do an example. First download an existing template to start from via

rda template describe DigitalGlobeStrip > dgstrip.json

This is the same template in the example above, and you'll notice it uses the DigitalGlobeStrip operator (not to be confused with the template) as the first node in the graph. rda operator DigitalGlobeStrip informs us that a parameter that isn't exposed in the template is Resampling Kernel. Editing dgstrip.json such that the first node's parameter section reads as

"parameters": {
  "CRS": "${crs:-UTM}",
  "GSD": "${GSD}",
  "bands": "${bands:-MS}",
  "catId": "${catalogId}",
  "correctionType": "${correctionType:-DN}",
  "fallbackToTOA": "${fallbackToTOA}",
  "Resampling Kernel": "INTERP_BICUBIC"

and then uploading our modified template via rda template upload dgstrip.json, we get back


We can see our modified template via rda template describe c21bf003b5803f03b0f0358c607ab1ffa76b89a88a64d0f4a54ab9cb73470bae, which now has a fancier resampling kernel.

rda template metadata

Just like the other metadata subcommands, this returns RDA metadata describing the evaluated template. One caveat is you can change which node in the graph you want metadata from via the --node option. If not provided, the default node is assumed. You also need to populate any template parameters that are not defaulted; this is done via key/value pairs specified like --kv "key,value". For example,

rda template metadata c21bf003b5803f03b0f0358c607ab1ffa76b89a88a64d0f4a54ab9cb73470bae --kv "catalogId,1040010038952900" --kv "GSD,5.0" --kv "bandSelection,RGB"


  "ImageMetadata": {
    "ImageWidth": 3363,
    "ImageHeight": 25406,
    "NumBands": 3,
    "MinX": 0,
    "MinY": 0,
    "DataType": "UNSIGNED_SHORT",
    "TileXSize": 256,
    "TileYSize": 256,
    "NumXTiles": 14,
    "NumYTiles": 100,
    "MinTileX": 0,
    "MinTileY": 0,
    "MaxTileX": 13,
    "MaxTileY": 99,
    "AcquisitionDate": "2018-02-28T13:35:08.858Z",
    "ImageID": "b3bca8fd-cf46-4b3c-a6e5-8801ec13de3f",
    "TileBucketName": "rda-images-1"
  "ImageGeoreferencing": {
    "SpatialReferenceSystemCode": "EPSG:32723",
    "TranslateX": 333540.0423765521,
    "ScaleX": 5,
    "ShearX": 0,
    "TranslateY": 7458901.487530498,
    "ShearY": 0,
    "ScaleY": -5

rda template realize

rda template realize is very similar to rda dgstrip realize, but you provide a template id and template parameters to populate via --kv as in rda template metadata. --node is also supported to evaluate a node other than the default in the graph. --srcwin and --projwin are also supported with the caveat that --projwin only makes sense to use with nodes that provided georeferenced outputs. You can use --maxconcurrency to increase the number of concurrent tile downloads.

Let's see what the output is like between the DigitalGlobeStrip template, which uses bilinear resampling, compared with our new template that used a cubic kernel.

Bilinear output can be generated like so:

rda template realize DigitalGlobeStrip --kv "catalogId,1040010038952900" --kv "bandSelection,RGB" --kv "bands,PanSharp" --kv "draType,HistogramDRA" --kv "correctionType,ACOMP" --projwin 339830.435,7391058.064,341126.283,7389710.445 bilinear.vrt

Cubic output is realized via:

rda template realize c21bf003b5803f03b0f0358c607ab1ffa76b89a88a64d0f4a54ab9cb73470bae --kv "catalogId,1040010038952900" --kv "bandSelection,RGB" --kv "bands,PanSharp" --kv "draType,HistogramDRA" --kv "correctionType,ACOMP" --projwin 339830.435,7391058.064,341126.283,7389710.445 ~/Downloads/cubic.vrt

You should see that the cubic output is sharper.

Note the only change is the template ID in these calls. Also note that rda dgstrip realize is just a nicer way of expressing the same API calls to RDA for the first example.

rda template batch

rda template batch uses RDA's batch materialization to generate tiles for you, just like rda dgstrip batch. The arguments are the same as for realize above, except that RDA batch mode only supports georeferenced outputs, meaning you cannot batch materialize something like a DG 1B. Here's an example of how to use it:

rda template batch c21bf003b5803f03b0f0358c607ab1ffa76b89a88a64d0f4a54ab9cb73470bae --kv "catalogId,1040010038952900" --kv "bandSelection,RGB" --kv "bands,PanSharp" --kv "draType,HistogramDRA" --kv "correctionType,ACOMP" --projwin 339830.435,7391058.064,341126.283,7389710.445

Use rda job and its subcommands to check on the job id and download its outputs.

rda job

rda job hosts subcommands lets you status and download the outputs from RDA's batch materialization endpoint. The subcommands of interest are download, downloadable, status, and watch.

rda job downloadable

This command returns all the job ids that are listed in your GBDX customer data bucket under the rda prefix. You should be able to status, download, or watch these jobs.

You can optionally provide a job id as an argument; if you do so, you will be returned a list of the objects that can be downloaded for that job id.

rda job download

download will download all the outputs for the given job id that are in S3. Use the watch subcommand to watch a job and greedily download outputs as they arrive.

Note that you can also provide the path to an individual object (e.g. a path returned from rda job downloadable where you provide a job id as an argument) to pull down just that object. This is implemented via prefix matching, so in reality you can provide a prefix to match as the job id and all matching objects will be returned. This is similar to how the aws cli command functions.

rda job status

This returns the status of the given job id associated with an RDA materialization request.

rda job watch

This is a combination of the functionality of download and status; essentially it polls RDA for job status, greedily downloading any of its produced outputs as they are created, and continues to poll until the job is complete and all outputs are downloaded to the given output directory.

For example, running

rda job watch ~/Downloads/rdaout 21a12531-2bfe-4e29-84b0-52b9433f7a61

downloads the output of job id 21a12531-2bfe-4e29-84b0-52b9433f7a61 to ~/Downloads/rdaout on my machine.

rda job rm

This removes all artifacts in S3 associated with the a given RDA batch job id. For instance,

rda job rm 21a12531-2bfe-4e29-84b0-52b9433f7a61

would remove all S3 objects in your GBDX customer data bucket associated with the batch job 21a12531-2bfe-4e29-84b0-52b9433f7a61.

You can’t perform that action at this time.