Frequently Asked Questions

I am not sure whether I need Datalore On-Premises or Datalore Cloud. Are there any differences?

Both editions of Datalore share most of the functionality between each other.

However, Datalore On-Premises has a few differences:

• By its nature, it is installed and operated on customers' own infrastructure, thus enabling an extended compliance and data governance scenarios (like air-gapped operations).

• Single sign-on authentication is currently supported for Datalore On-Premises only.

• With Datalore On-Premises, the base agent image can be customized, and custom JDBC connectors can be installed, adding more flexibility when needed.

We also recommend checking out a blog post which covers pros and cons of both approaches.

How frequent are Datalore releases?

While we do not have a fixed schedule for Datalore On-Premises releases, we aim to release a new version every quarter.

However, Datalore Cloud release schedule is not dependant on the release schedule of Datalore On-Premises; we release new features and improvements to Datalore Cloud on a regular basis.

We would like to sign up an NDA prior to discussing a sensitive topic. Is this possible?

Yes.

If the question is about the security-related topic - please, refer to JetBrains Trust Center for further guidance.

For any other inquiries - please, reach out to us, and we'll be happy to discuss this in details.

Is it possible to modify the baseline agent, so our custom libraries, software or a specific interpreter version can be pre-installed?

Yes. Refer to Customize or update environment for further guidance.

Datalore agents require too many permissions. Is it possible to run Datalore in a confined environment?

In principle - yes, but the reactive mode and neither the cloud storage connections nor Attached Files will work properly because of the current architectural limitations. We advise to contact our Support to discuss this requirement.

An automated image scanner has identified a vulnerability in one of your images. How this can be addressed?

Refer to CVE assessment and response for more details.

Can Datalore On-Premises be installed in our cloud or in our own datacenter?

Yes. The "On-Premises" here means that the installation is performed on a customer-controlled environment, regardless of its location (e.g., on-premises, in the cloud, or even in a hybrid environment). The only requirement for the On-Premises installation is the ability to run Kubernetes or Docker.

For more details, see Install Datalore On-Premises.

What is the right way to report a bug or issue with Datalore?

We'd love to hear from you! Reach out to our Support for further guidance.

What data does Datalore AI send to LLM providers?

Depending on how you use AI Assistant in Datalore, the following types of data are sent to LLM providers:

• Your prompts submitted to AI Assistant

• Notebook content selected when submitting prompts, including pieces of your code

However, Datalore's AI Assistant does not interact with your data directly - it can't access the data sources the notebook is connected to; its scope is limited to the notebook runtime variables and cell contents.

Therefore, if certain data is loaded to a runtime variable (Pandas dataframe, for example) - there is a possibility that AI Assistant will transfer a fraction of such data to JetBrains infrastructure.

In Datalore Cloud, we record users’ prompts so they have access to their prompt history.

If you use Datalore On-Premises with JetBrains AI, data is relayed through the JetBrains infrastructure, but we do not store it and we do not use it to train models. If your Datalore On-Premises instance uses Bring Your Own Key, all data is sent directly to LLM providers.

Can data sent by the AI Assistant to LLM providers be used for model training?

Datalore does not work with LLM providers that use customer data to train models.

For more details about LLM providers, see the JetBrains AI docs. Note that Datalore does not participate in Detailed Data Collection.

If your Datalore On-Premises instance uses Bring Your Own Key, the prompts and notebook content sent are governed by your AI provider’s data policies.

How secure is code suggested by AI Assistant?

AI Assistant does not check code suggested by its LLM providers for security vulnerabilities. Code suggested by AI Assistant is therefore no different in terms of security than code written by a developer and should be further reviewed to identify security vulnerabilities.

What LLM providers does AI assistance use?

AI Assistant in Datalore Cloud is powered by the JetBrains AI service, an independent LLM provider.

It uses Anthropic Claude Sonnet 4.5 and models trained by JetBrains by default, but can also use models from OpenAI and Google.

In Datalore On-Premises, instance administrators can select between models from OpenAI, Anthropic, or Google, or connect to their own AI provider via Bring Your Own Key.

Can I use my own model or my own LLM provider API key with Datalore?

Yes. In Datalore On-Premises, you can use any AI provider that supports an OpenAI-compatible API. That includes local models running on your infrastructure.

Can I avoid transferring data to JetBrains' premises?

If you use Datalore On-Premises with Bring Your Own Key, your data is sent directly from your Datalore instance to the LLM provider.

Is it possible to use Full-Line Code Completion in Datalore?

Not yet, but we're working on that.

Can I completely turn off AI assistance?

Yes. In Datalore Cloud, AI assistance is turned off by default unless you explicitly accept the AI terms and conditions. In Datalore On-Premises, AI assistance is turned off unless it is enabled instance-wide.

I am working with the database XXXX. Can I use it with Datalore's SQL cells?

Yes - as long as this database has a JDBC-compliant driver, it can be connected to Datalore. See Using custom or third-party JDBC drivers for more details.

I am working in Jupyter; is it possible to perform a "lift and shift" migration to Datalore?

Yes. Datalore is fully Jupyter-compliant, so any notebook created in Jupyter would work in Datalore with no additional configuration.

Will it be possible to migrate from Datalore to any other platform (for example, to Jupyter)?

Partially. While Datalore maintains back compatibility with Jupyter (so the Markdown and Code cells will migrate with no issues), there are also some Datalore-specifc features (like SQL cells, chart cells and so on); therefore, any other platform may not recognize them properly.

Which databases are currently supported natively by Datalore?

We support most of the modern databases, such as PostgreSQL, MySQL, MariaDB, SQLite, Oracle, SQL Server and many more. For full list of databases, see DataGrip documentation (as the DataGrip technology is used in Datalore for the database connections).

Custom JDBC connectors can also be used. See Using custom or third-party JDBC drivers for more details.

The notebook startup is slow, and the message Initialization may take longer than usual is observed. What can be done?

This message means that the machine is not taken from the pre-warmed pool, but is launched on-demand instead. Consider adjusting the initialPoolSize value of the selected agent configuration. See Configure Agents (Helm) or Configure Agent (Docker) for more information.

I would like to authenticate at my database with my personal SSO credentials, not with a workspace-wide service account. Is this possible?

No. However, we are well aware about the importance of this and already working on lifting this limitation.

There's no Azure Storage Account as a cloud storage option. Can I use it?

Yes. While the whole Storage Account cannot be connected, a part of it can be allocated as an SMB file share. See SMB Azure file shares for more details.

Once the cell is executed, an error "Checkpoint failed" is observed.

This is an error message, indicating that Datalore was not able to take a snapshot of the cell in question. This means that regardless of the reactive mode configuration, this specific cell will be re-evaluated during the next recalculation.

It's also recommended to check the logs (see Troubleshooting for further guidance). One of the reasons for this error could be the incorrectly configured permissions of the agent container. See Runtime for more details.

Can we use visualization libraries with a server-side architecture, like Dash with Datalore?

No. Since Datalore notebooks are running within an isolated environment, there's no option to configure the traffic ingress to the notebook runtime. Thefore, using a server-side app is technically possible, but it won't be accessible over the network.

If the owner of a workspace is leaving the organisation, can the owner of that workspace be changed?

If the owner is not changed, and the owner account is deleted, what happens to the workspace?

While the transfer of the workspace is not supported, the whole workspace can be cloned. See Workspace full clone for further guidance.

If the workspace owner account is deleted prior to the copying of the workspace, the data is irreversably lost.

Once the widget value is changed, does Datalore recalculate only the dependent cells (e.g., the cells to which the modified cell is referenced), or is the whole notebook recalculated?"

If the reactive mode is used, Datalore will recalculate only the modified cell and all of the subsequent cells. Otherwise, the whole notebook will be recalculated from the very beginning.

Is it possible to create multiple reports linked to the same notebook?

No. But we are looking into the options of making this possible.

Notebooks can contain multiple tabs to organize their content. Is it possible to create a report with multiple tabs as well?

Yes, reports can contain multiple tabs just like notebooks.

You can also replicate the notebook tabs in a report by selecting (More) | Replicate tabs at the top right of the report editor.

For more details, see Prepare a report.