logo

Quickstart

Authentication

Image without caption
There are ways to authenticate and interact with Censius AI Observability Platform:
Console Access
You can use your email ID and password to log in via the web console.
Programmatic Access
Every user gets an API key and tenant ID from us after signing up. API keys are needed when accessing Censius products programmatically (for example, when sending logs via the Python SDK).
🚀
If you don’t already have access to Censius, request it here.

Creating a Project

A Project is an organization structure that allows users to club together models under a similar category, client, or any other required preference.
A project can be created with the Censius UI through the + New Project button on the top-right corner of the Projects screen. The project will be assigned a unique ID after creation and start showing the activity details. You can perform opendeleterename, and share the project with your teammates.
💡
Every user can access a project if he/she is either a member of the project or is the one who created it.

Registering a Dataset

On entering a Project, there are two visible tabs: Models and Datasets
Models are dependent on datasets, so if the required dataset is not registered yet, it can be registered using the register_dataset() API to register a dataset to the Censius platform.
javascript
import pandas as pd dataframe_object=pd.read_csv("titanic.csv") client.register_dataset( name = 'titanic_dataset', file = dataframe_object, project_id = 121, features = [ { "name": "Survived", "type": DatasetType.INT }, { "name": "Pclass", "type": DatasetType.INT }, { "name": "Sex", "type": DatasetType.STRING }, { "name": "Age", "type": DatasetType.DECIMAL }, { "name": "SibSp", "type": DatasetType.INT }, { "name": "Parch", "type": DatasetType.INT }, { "name": "Fare", "type": DatasetType.DECIMAL } ], timestamp={"name":"Timestamp","type":DatasetType.UNIX_MS} )
Registering Dataset through Censius API

Registering a Model

Once the datasets are in place, you can register a new model either through the UI or the API.
When using the UI, there is an option to choose between existing models (from across projects) or create a new one. After that, the model just needs to be linked to the associated dataset and it will get registered instantly!
You can the register_model() API to register a new model to the Censius platform.
python
client.register_model( model_id = "titanic_m", model_name = "titanic model", model_type = ModelType.BINARY_CLASSIFICATION, model_version = "v1", training_info = { "method": Dataset.ID, "id": 262 }, project_id = 121, targets = ["Survived"], features = ["Pclass", "Sex", "Age","SibSp","Parch","Fare"] )

Auto-init of monitors

Once a dataset has been uploaded and a model associated with it, Censius automatically starts creating auto initialized monitors with pre-set thresholds based on the training data that was uploaded.
💡
Auto-initializing can take a while to process (30-60 mins depending on the number of features and the size of the training dataset). Feel free to go get coffee.

🖇️
Python SDK Reference Use the above guide for more details on how to use APIs to register models and datasets

Helpful?