New Releases, Enhancements, Changes + Arize in the News!

What's New

Local File Upload

Upload a local file directly through the UI on the 'Upload Data' page. Use Local Upload to verify the schema for a sample of your data and quickly onboard a model. This feature supports CSV, Parquet, and Avro files. Learn how to upload a file here.

Note: Local File Upload is limited to one 30MB file at a time.

Python SDK: Model Type & Metrics Validator

Revised model_type names (i.e. binary classification, regression) for enhanced clarity when sending in a model. Learn more about model types here and the applicable metric group per model type here.

(Optional) Provide the metrics_validation field to validate your schema and specify the desired metric group to be visualized in the Arize UI. Learn more about the metrics validator here.

response = arize_client.log(
    model_type=ModelTypes.BINARY_CLASSIFICATION, # new model type name
    metrics_validation=[Metrics.CLASSIFICATION], # new metrics validator

Python SDK: Embeddings Dictionary

embedding_column_names is now a dictionary object (previously a list). Use this dictionary to name embeddings represented in the UI for enhanced platform flexibility. Learn more here.

# before (embedding_column_names as a list)
embedding_column_names = [

# after (embedding_column_names as a dict)
embedding_column_names = {
    "embedding_display_name": EmbeddingColumnNames(

In The News

"What is" Series

Many executives wonder how generative AI will impact their business – particularly in a turbulent economic environment where growing productivity takes on an elevated importance. Here are a few tips for developing a generative AI strategy.

Key predictions for the state of AI in 2023:

  1. Generative AI will go mainstream (but we need all hands on deck)

  2. AI will likely take on elevated importance as the economy pressures companies to deliver greater efficiency and productivity.

  3. The days of central ML teams taking months or years to build and maintain proprietary feature stores or monitoring tools in-house are numbered.

  4. ML platforms that aren't built to handle unstructured use cases risk irrelevance.

Last updated