Directly upload raw data or send traces: send inputs and outputs of your AI system (text, tabular data, or embeddings) to Evidently Cloud, or use Tracing to instrument your app. Here, we charge based on the number of rows stored and processed. Retention policies apply.
Run evaluations locally: use the Python SDK to run checks in your environment on local datasets and upload only the results (JSON snapshots). You can evaluate millions of rows this way: if you don't upload raw data, we only charge by the total GB volume of snapshots stored.
You can combine both methods.
Can I get a trial of the enterprise version?
Yes, contact us to learn more about the enterprise version trial.
Usage limits
What are rows and snapshots?
Row is a row of data (e.g., inputs-output pair for a generative or predictive system) uploaded to Evidently Cloud. You can analyze, monitor and view raw data in the web interface.
Snapshot is a JSON file with aggregated data summaries and evaluation results for a specific period. If you upload only snapshots, you can view interactive reports, plot metrics, and get alerts. However, you cannot view the original raw data (individual predictions or conversations).
What is the size of one snapshot?
The exact snapshot size depends on the metrics, tests, columns, and rows in the dataset. You can use the Evidently Python library to generate sample snapshots and estimate their size.
For example: Data drift report for 50 columns and 10,000 rows: ~1MB Data drift report for 100 columns and 10,000 rows: ~3.5MB Data drift report for 100 columns and 100,000 rows: ~9MB
Sizes vary based on metrics used, so it's best to perform your own sizing.
What happens if I go over the limit?
If you are on a free tier and hit the data row limits, you will need to upgrade to continue using the tool. If you exceed your snapshot storage limits, you can either upgrade, or delete existing data to continue.
If you are on a paid tier, you will be billed according to your pricing plan.
Features
What are local evaluations?
You run local evaluations on your infrastructure using the Evidently Python library.
For example, you can pass a data batch and run regression tests during CI/CD processes. You can then send the evaluation results (JSON snapshots) to Evidently Cloud.
Local evaluation runs don’t count towards the monthly row limit; they only contribute to the snapshot storage.
What type of predictive ML tasks do you support?
We support all types of predictive ML tasks, with dedicated presets for classification, regression, recommendation systems, and ranking. Evidently works with text, tabular data, and embeddings.
What type of generative AI tasks do you support?
We support LLM-powered use cases such as chatbots, summarization, information extraction, RAGs, and AI-powered assistants. Evidently works with any input-output of a generative system, so you can adapt it to any case.
Is Evidently open-source?
Yes, the Evidently Python library is open source under the Apache 2.0 license, suitable for individual users and small teams, or teams with infrastructure knowledge to self-host. To start with Evidently open-source, check the documentation.
Evidently Cloud builds on the open-source version with additional features for larger teams and enterprises, including a no-code interface, alerting integrations, and user management.
Get Started with AI Observability
Book a personalized 1:1 demo with our team or sign up for a free account.
By clicking “Accept”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.