Snowflake AI Data Cloud Features & Architecture · Topic 15 of 17

AI/ML and App Development, Snowflake Notebooks and Streamlit in Snowflake

Map Snowflake's AI, ML and app features, then learn how Snowflake Notebooks and Streamlit in Snowflake run, who can use them and when to pick each one.

  • 5 min
In a nutshell
  • Snowflake Notebooks: Python, SQL and Markdown cells in Snowsight, results shared between SQL and Python cells.
  • Notebooks in Workspaces is the next generation, with container compute, CPUs and GPUs, Git and a scheduler.
  • Legacy Notebooks have a Warehouse Runtime and a Container Runtime.

What it is

Snowflake brings code, AI and apps to the data instead of moving data out to other tools. Everything runs inside Snowflake's security boundary and is governed by the same role-based access control (RBAC) as tables.

At the Core level, the exam checks that you can match each feature to its job. This page gives the map, then covers the two interactive tools: Snowflake Notebooks and Streamlit in Snowflake.

FeatureWhat it is forWho typically uses it
Snowflake NotebooksInteractive Python, SQL and Markdown cells for analysis, data engineering and MLData scientists, data engineers
Streamlit in SnowflakeBuild and share interactive Python data appsDevelopers building apps for business users
SnowparkDataFrame API and UDFs in Python, Java and Scala that run in SnowflakeData engineers, developers
Snowflake MLTrain, register, serve and monitor ML modelsData scientists, ML engineers
Snowflake CortexManaged AI: LLM functions in SQL, Cortex Search, Cortex AnalystAnalysts, app developers
Governed datatables and stagesexplore inSnowflake NotebooksPython, SQLproductionize withSnowpark and Snowflake MLpipelines, modelspresent inStreamlit in Snowflakedata apps

How it works

Snowflake Notebooks

A Snowflake notebook is a cell-based development environment inside Snowsight. Cells can hold Python, SQL or Markdown, and they run next to your data with no driver or connection setup.

  • Results of a SQL cell can be used in a Python cell, and Python results can be used in SQL, so you can mix both in one analysis.
  • Markdown cells document the work for whoever reads it next.
  • Access is controlled by roles, like any other object.

Snowflake now offers Notebooks in Workspaces, the next generation of Snowflake Notebooks. It runs in a prebuilt container environment with managed CPUs and GPUs for AI and ML work, is Jupyter compatible, supports Git-integrated and shared workspaces, and has a native scheduler for running notebooks in production pipelines. The earlier notebooks are now called Legacy Notebooks.

Legacy Notebooks offer two runtimes:

RuntimeRuns onBest for
Warehouse RuntimeA virtual warehouseFast start, SQL and Snowpark work
Container RuntimeA compute pool, with CPU or GPU images and ML packages preinstalledML training and deep learning

In every runtime, SQL and Snowpark queries from the notebook run on a virtual warehouse, so warehouse credits apply.

SnowsightnotebookSQL cellWarehousequery resultspassed toPython cellpandas, Snowparkcharts inMarkdown and outputshared with role
Exam tip

If a scenario wants data scientists to explore data and train models interactively in Snowflake, without copying data to a laptop, the answer is Snowflake Notebooks.

Streamlit in Snowflake

Streamlit is an open-source Python library for building interactive data apps with very little code: charts, filters, forms and tables. Streamlit in Snowflake lets you build, deploy and share those apps directly in Snowflake.

  • You edit the app in Snowsight with a side-by-side editor and live preview.
  • The source code and environment configuration are stored in a Snowflake object, so RBAC governs who can see and run the app.
  • The app can call Snowpark, UDFs and stored procedures, and can be part of a Snowflake Native App.
  • You can choose a warehouse runtime or a container runtime to run the app.

Apps run with owner's rights by default: queries execute with the privileges of the app's owner role, not the viewer's role. The owner decides which roles may use the app. That lets a business user see the app's results without holding direct privileges on the underlying tables.

Developerowner rolebuilds in SnowsightStreamlit appSnowflake objectgrants use toBusiness roleopens appQueriesrun as owner
Watch out

Because the app runs with the owner's privileges, anything the owner role can read, the app can expose. Use a dedicated owner role with only the access the app needs.

Choosing between them

NeedPick
Interactive exploration, feature engineering, model trainingSnowflake Notebooks
A point-and-click app for business users who do not write codeStreamlit in Snowflake
Scheduled production pipeline from notebook codeNotebooks with a schedule, or Snowpark code in tasks
Natural language questions or LLM featuresSnowflake Cortex (covered on its own page)

How the exam asks about this

  • "Sales managers need an interactive app with filters over Snowflake data, without the data leaving Snowflake" -> Streamlit in Snowflake.
  • "A data scientist wants to combine SQL and Python cells to explore data and train a model in Snowsight" -> Snowflake Notebooks.
  • "App viewers should see results without being granted SELECT on the tables" -> Streamlit in Snowflake, because apps run with owner's rights.

Common mistakes

Thinking notebooks run outside Snowflake. Snowflake Notebooks run inside Snowflake, in Snowsight, against governed data. No data is exported to run them.

Assuming a Streamlit app runs as the viewer. By default it runs with the owner's rights. Viewer privileges on tables do not limit what the app can query.

Confusing Streamlit with a BI tool connection. Streamlit in Snowflake apps are Python code stored and run in Snowflake, not an external dashboard tool connecting over a driver.

Forgetting compute costs. Notebook SQL and Snowpark queries use a warehouse, and container runtimes use compute pools, so both consume credits.

What to remember for the exam

  • Snowflake Notebooks: Python, SQL and Markdown cells in Snowsight, results shared between SQL and Python cells.
  • Notebooks in Workspaces is the next generation, with container compute, CPUs and GPUs, Git and a scheduler.
  • Legacy Notebooks have a Warehouse Runtime and a Container Runtime.
  • Streamlit in Snowflake builds Python data apps stored as Snowflake objects and governed by RBAC.
  • Streamlit in Snowflake apps run with owner's rights by default.

Check yourself

  • Which cell types can a Snowflake notebook contain?

    Python, SQL and Markdown cells.

  • Whose privileges does a Streamlit in Snowflake app use by default?

    The owner's privileges, so viewers do not need direct access to the underlying tables.

  • Which tool fits an interactive app for business users over Snowflake data?

    Streamlit in Snowflake, which builds and shares Python data apps inside Snowflake.

Topic 15 of 17 in Snowflake AI Data Cloud Features & Architecture

Test yourself on Snowflake AI Data Cloud Features & Architecture

The full Snowflake SnowPro Core COF-C03 question bank has 845 questions, with timed practice exams and an explanation for every answer.

This study guide is independent and is not affiliated with, endorsed by, or sponsored by Snowflake. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.