Skip to main content

The Dojo API

This document outlines how to interact with the Dojo API to fetch models, execute models, and fetch model runs.

Model Discovery

You can search Dojo for models using the GET /models. For example, to find models related to crops you would send:

curl -X 'GET' \
'' \
-H 'accept: application/json'

This will return an array of model objects defined by the Dojo model schema.

A key thing to note is that each model has as a set of parameters. Parameters are model specific and the model metadata will include default values for each parameter. Understanding a model's individual parameters is crucial for executing a model.

Additionally, you may retrieve models directly based on their id. For example, here we retrieve DSSAT based on its id:

curl -X 'GET' \
'' \
-H 'accept: application/json'

Model Execution

To execute a model, We use runmodel from the dojo-cli.

First, you must obtain the model's metadata to appropriately set the model's parameters. Any parameter not explicitly set will fall back to its default setting which was specified at the time of model registration.

Let's say we wish to execute DSSAT with the following parameters:

  • Fertilizer amount addition: 25 kg[N]/ha
  • Rainfall multiplier: 1.25
  • Offset to planting date window: 30 days

dojo runmodel --model="DSSAT" --params='{"fertilizer_amount_addition": 25, "rainfall_multiplier": 1.25, "offset_to_planting_date_window": 30}'

Your model will now run in "attached mode". The normal output generated by the model will be displayed on the screen and the command will run until the model completes. Details will be provided at the end of the run which will indicate where all of the generated files have been located on your local system.

You can also run in "detached mode" by adding --attached=False to the arguments above. In this case, the model will run in the background and you must retreive your run results when they complete.

Retrieving Detached Model Runs

To retrieve results. We use results from the dojo-cli

--id : id of the docker container
--name : name of the docker container (as returned by the runmodel command)
--config : name of configuation file; defaults to .config

dojo results --name=dojo-DSSAT20211227140758

Run completed.
Model output, run-parameters, and log files are located in "/mydojodata/runs/DSSAT/33cf1a60-2544-420f-ae08-b453a9751cfc/20211227140758".

Debugging Model Runs

After using runmodel from the dojo-cli, you can debug model runs by fetching the model execution logs. go to the log locations /runs/{model-name}/{uuid}/logs.txt

Searching for Model Runs

To print a description of the model, we use describe from the dojo-cli

For example, we can query for:

  • model_name: Population Model

dojo describe --model="Population Model"

Population Model


The population model serves as an ancillary tool to distribute, disaggregate yearly population projections onto a geospatial representation. Occasionally, the output of this model is required as an independent variable for downstream models.y

Working with Model Results

Model results are stored in the Geotemporal format. They can be interacted with conveniently using Python's Pandas library, or any other library used to process parquet.

For example:

import pandas as pd

data_path = "./mydojodata/runs/DSSAT/33cf1a60-2544-420f-ae08-b453a9751cfc/20211227140758/example-run-8654912_5cf84e06-c1ce-4a9d-a2e6-ede687293a26.1.parquet.gzip"

df = pd.read_parquet(data_path)


This will return:

441763200000KenyaNyamiraNorth MugirangoMagwagwa-0.45835.042HDAT_AVE461462400000
473385600000KenyaNyamiraNorth MugirangoMagwagwa-0.45835.042HDAT_AVE495504000000
504921600000KenyaNyamiraNorth MugirangoMagwagwa-0.45835.042HDAT_AVE526608000000
536457600000KenyaNyamiraNorth MugirangoMagwagwa-0.45835.042HDAT_AVE557366400000