Abacus.AI API Reference

Abacus.AI API client makes it easy to embed cutting-edge AI into your mobile, web, or/and desktop application(s). These API methods span from creating a project for your data science/engineering team to generating predictions from deployed models. We have divided the API reference section into categories based upon their broad functionality e.g., project, dataset, model, predict, etc. Our APIs are designed to provide maximum security, reliability, interoperability, robustness, modularity, and ease of use. To make the API methods easy to reference, we currently provide at least one example for each API method in both, CURL and Python. You can search for a specific API method using the API search dropdown at the top right of the screen.

Authentication

Abacus.AI uses key based authentication. You can view and manage your API keys in the API keys Dashboard. Your API keys carry many privileges, so be sure to keep them secure. Every private API call requires key-based authentication.

Creating an API key

Navigate to the API keys Dashboard. Click on the "Generate new API Key" button to generate a new key for the current organization. If you want your key to have a custom tag, enter a tag name into the box under the "Generate new API Key" heading and then hit the "Generate new API Key" button. For security reasons, do not put your key directly in your code and don't commit it to public places like GitHub.

Managing API keys

When viewing the API keys Dashboard, you will see a list of your current API keys along with the following information:

Deleting an API key

Once you delete a key, it can no longer be used to access Abacus.AI's services. Click on the "Delete" button corresponding to the key you want to delete. This will delete the key permanently. Abacus.AI will reject any subsequent API calls made using the deleted API key.

Making API Calls using CURL

All API methods under the API reference are marked with GET and/or POST requests.

GET Request Format

The general format for using curl to make a GET request call is as follows:

  curl -X GET -H "apiKey: YOUR_API_KEY" "API_METHOD_URL"

where the client needs to provide an API key in the HTTP request header (for authentication) and then use the API method url (to call the API method).

GET Request Example

An example showing how to make an API call to list all the projects in thr current organization: REQUEST:

curl -X GET -H "apiKey: 3210987**************86507975azz" "https://api.abacus.ai/api/v0/listProjects"

RESPONSE:

{
              "success": true,
              "result": [
                {
                  "projectId": "63dcd98a6",
                  "name": "Movie Recommendations",
                  "useCase": "USER_RECOMMENDATIONS",
                  "createdAt": "2021-01-05T21:22:23+00:00"
                },
                {
                  "projectId": "205a81a4",
                  "name": "TV Series Recommendations",
                  "useCase": "USER_RECOMMENDATIONS",
                  "createdAt": "2021-01-05T20:22:31+00:00"
                }
              ]
              }

POST Request Format

For POST requests, you MUST provide the Content-Type header. Url parameters or POST body parameters are supported interchangeably.

curl -H "apiKey: YOUR_API_KEY" 'METHOD_URL -d 'data1' 'data2'

POST Request Example

An example showing a post request to the rename dataset API method:

curl "https://api.abacus.ai/api/v0/createProject?name=Customer&useCase=" \
  -X POST \
  -H "apiKey:**********c74043986d6087446810a1"

Errors

The Abacus.AI API will return suitable errors if anything goes wrong with your API calls. Most errors will have a brief description of the cause, along with the error type.

Error format

We expose API errors in JSON format:

Error Glossary

The errors shown in the glossary are context dependent. The exact error message will differ and change depending upon the situation it is thrown at. The descriptions shown below are provided to give a general idea on the error codes:

Error Examples

Example 1: If wrong API key is passed - HTTP 403 error:

curl -X GET -H "apiKey:"  "https://api.abacus.ai/api/v0/listProjectDatasets?projectId=a42399ada"

Error Output:

{"title": "403 Forbidden"}

Example 2: If a parameter is missing - HTTP 400 error:

curl -X GET -H "apiKey:3210987**************86507975azz" "https://api.abacus.ai/api/v0/getTrainingConfigOptions?"

Error Output:

{
  "success": false,
  "error": "Missing required parameter 'projectId'",
  "errorType": "MissingParameterError"
}

Example 3: If an invalid method type is used for the HTTP request:

curl -X GET  -H 'apiKey:3210987**************86507975azz'  'https://api.abacus.ai/api/v0/addCustomColumn?projectId=58e9acad4&datasetId=f58641d08&column=new_gender&sql=CONCAT(gender,"A")'

{
  "success": false,
  "error": "Invalid HTTP Method (GET) used for action addCustomColumn. Method must be called using HTTP Method: POST",
  "errorType": "MethodNotAllowed"
}

Commonly Used API Methods

• Uploading a dataset from a local file
  You can create a dataset or dataset version by uploading your file using the upload_file utility. First, create an upload reference using the create_dataset_from_upload or create_dataset_version_from_upload APIs. Then, open your file in mode 'r' or 'rb', then pass your file in the upload_file method:

client.create_dataset_version_from_upload('<DATASET_ID>')
with open('<input_file_path>', 'rb') as input_file:
    upload.upload_file(input_file)

• Downloading batch predictions
  You can download the results of a batch prediction once you have a reference to the Batch Prediction Version. First, open a file in mode 'wb', then pass the file to the download_result_to_file method in the Batch Prediction Version object.

bpv=client.describe_batch_prediction_version('<BATCH_PREDICTION_VERSION>')
with open('output.csv', 'wb') as output_file:
    bpv.download_result_to_file(output_file)

• Uploading a CSV created in the in-product notebook as a dataset into the product
  You can upload a CSV created in your notebook as a dataset into Abacus.AI by using the following code:

client = ApiClient()
uploader = client.create_dataset_from_upload('MyOutputData-1', 'myoutputdata1', file_format='csv')
with open('my_output_data.csv') as dataset:
    uploader.upload_file(dataset)
dataset = client.describe_dataset(uploader.dataset_id)
dataset.wait_for_inspection()
print(f'{dataset.name} with id of {dataset.dataset_id} is now available at https://abacus.ai/app/dataset_detail/{dataset.dataset_id}')