API endpoints

POST /prediction

Request a prediction on data for a given problem.

Data to post:
  • data : list of UUID on which to apply the prediction
  • problem : UUID of the associated problem
Success Response content:
  • new_preduplets: number of newly created preduplets
POST /problem

Add a new problem

Data to post:
  • uuid : problem UUID
  • workflow : workflow UUID
  • test_dataset : list of test data UUID
  • size_train_dataset : nb of train data per minibatch (integer)
Success Response content:
  • new_problem: new problem
POST /algo

Add a new algorithm

Data to post:
  • uuid : algo UUID
  • name : algo name
  • problem : UUID of associated problem
Success Response content:
  • new_algo: new algorithm
  • new_learnuplets: number of newly created learnuplets
POST /data

Add data

Data to post:
  • uuid : data UUID or list of data UUIDs
  • problems : UUID or list of UUID of associated problems
Success Response content:
  • new_datas: list of new data
  • new_learnuplets: number of newly created learnuplets
POST /worker/(uplet_type)/(uplet_uuid)
  • (uplet_type) : learnuplet or preduplet
  • (uplet_uuid) : UUID of the learnuplet or preduplet

Update the worker of a learnuplet or preduplet and change its status to pending (only exposed to the Compute).

Data to post:
  • worker : worker UUID
Success Response content:
  • learnuplet/preduplet_worker_set: uuid of updated learnuplet or

preduplet

POST /learndone/(learnuplet_uuid)
  • (learnuplet_uuid) : learnuplet UUID

Post output of learning, which updates the corresponding learnuplet. Modify the model_start of subequent learnuplet if performance increase. Only exposed to the Compute.

Data to post:
  • status : status of the prediction task: done or failed
  • If status = done, post also: - perf : performance of the trained model on all test data - train_perf : list of performances (one for each train data file) - test_perf : list of performances (one for each test data file)
Success Response content:
  • updated_learnuplet: uuid of updated learnuplet
POST /preddone/(preduplet_uuid)
  • (preduplet_uuid) : preduplet UUID

Update status of a preduplet. Only exposed to the Compute.

Data to post:
  • status : status of the prediction task: done or failed
  • prediction_storage_uuid : UUID of the prediction stored on Storage
Success Response content:
  • updated_preduplet: uuid of updated preduplet
GET /(collection_name)/(document_uuid)
  • (collection_name) : problem, algo, data, learnuplet,

or preduplet - (document_uuid) :UUID of the requested document

Get a document of a collection

Success Response content:
  • document elements
GET /(collection_name)
  • (collection_name) : problem, algo, data, learnuplet,

or preduplet

Get all the documents corresponding to the (collection_name). Possible to add filter to your request (e.g. /learnuplet?uuid=blabla)

Success Response content:
  • problems/algos/datas/learnuplets/preduplets: list of corresponding

documents