7 Snowflake Workspace & Notebooks

Snowsight UX, compute, and what Workspace provides for R

Keywords

snowflake, R, RStudio, Posit, VS Code, workspace notebooks, snowflakeR, RSnowflake, mlops

7.1 Overview

Before bootstrapping R, understand where you work in Snowflake. Snowsight is the web UI. Workspace is the project-oriented development area within Snowsight. Notebooks are interactive, cell-based environments for Python, SQL, and (with this guide and package) R.

Workspace Bootstrap covers setup_notebook() next.

Important

Workspace Notebooks required. The multi-language R stack in this guide (setup_notebook(), %%R, snowflakeR) runs only in Workspace Notebooks on ML Container Runtime — not in legacy standalone Snowsight notebooks.

7.2 Learning Objectives

Navigate Snowsight → Workspace → Notebook
Choose and configure compute for a notebook session
Use Python, SQL, and Markdown cells — and know where R fits
Understand session lifecycle (start, active, restart, end)

7.3 Snowsight and Workspace

Snowsight provides:

SQL worksheets and query history
Dashboards and Streamlit apps
Admin and governance screens
Workspace — projects, files, Git repos, notebooks

Workspace adds a developer workflow:

Capability	Benefit
Project + files	Organize notebooks, scripts, configs (`snowflaker_config.yaml`)
Git integration	Clone, commit, push — team collaboration
Notebooks	Cell-based Python/SQL/Markdown (Jupyter-compatible in Workspaces)
Compute session	Attached warehouse + container runtime for kernel

Note

Terminology: “Workspace Notebook” means a notebook running inside your Snowflake account — not Jupyter on your laptop. Kernel and data stay in-account.

7.3.1 Snowsight navigation

Screenshots are cropped to the relevant UI area — to aid readability.

Snowsight left navigation with Projects selected and Workspaces flyout menu — Snowsight Home — Projects menu with Workspaces shortcut

Snowsight Workspaces home page with Welcome header and Notebook/SQL quick-create buttons — Welcome to Workspaces — quick-start actions

7.4 Workspace notebooks

All R content in this guide assumes Notebooks in Workspaces — the Jupyter-compatible experience with Git integration and ML Container Runtime. Create and open notebooks from Snowsight → Workspace.

7.5 Creating and opening a notebook

Typical flow — product walkthrough: Snowflake Notebooks in Workspaces; Workspaces (files, folders, Git).

Open Snowsight → Workspace (Workspaces overview)
Select or create a workspace (private or shared); optionally connect a Git repository (Integrate workspaces with Git)
Create notebook (+ Add new / upload .ipynb) or open an existing notebook file in the workspace (overview — file management)
Start session — connect to a notebook service so the kernel runs on container runtime (compute setup). When configuring the service (Connected → Edit), you can optionally select a Custom Runtime Environment (CRE) registered by you or your organisation (cre@<org_name>) so R, %%R, snowflakeR, and RSnowflake are pre-installed — see CRE & ML Jobs. If no CRE is available, use the default ML Container Runtime and bootstrap R in step 7.
Set execution context: role, warehouse, database, schema (editing and running — set execution context)
Run a SQL or Python smoke cell (SELECT CURRENT_USER(), etc.) (run cells)
If you did not select a CRE in step 4: run the Python bootstrap cell (setup_notebook()) — Workspace Bootstrap

Warning

Workspace does not auto-set database/schema in all versions — set context explicitly in YAML config or SQL, or fully qualify object names.

7.5.1 Notebook workflow in Snowsight

Snowsight Create dropdown in Workspaces with Notebook and SQL file options — Create menu — Notebook, SQL file, Streamlit, and more

Workspace notebook editor with Run and Connected toolbar, empty Python cell, and SQL smoke query with results — Workspace notebook editor — Run toolbar and Python/SQL cells

Workspace notebook Use role and Use warehouse dropdowns in the toolbar — Session context — role and warehouse selectors

Workspace notebook Run button and Connecting status while the kernel starts — Notebook session toolbar — Run and Connect while kernel starts

7.6 Notebook toolbar & session

Workspace session controls (reference):

Control	Purpose
Start / Active	Start kernel session; view memory/CPU; restart or end
Run all / Stop	Execute all cells or cancel
Package selector	Install Python packages (Anaconda channel on some runtimes)
Scheduler	Run notebook on a schedule via Tasks
Connected / Edit	Role, warehouse, database, schema, External Access
Export / duplicate	Share or fork notebooks

For R without a CRE: run the Python bootstrap cell first; packages install via micromamba + YAML tarballs — not the Anaconda package picker alone.

7.7 Cell types

Workspace notebooks expose three cell types. There is no separate “R cell” type — you run R in a Python cell whose first line is %%R.

Cell type	Kernel / compute	Use	R in this guide
Python	Python kernel on compute pool	Snowpark, pandas, `setup_notebook()`; R via `%%R` magic (first line of the cell)	Bootstrap once unless you attached a CRE in step 4
SQL	Virtual warehouse	Queries, DDL, context	Same as any SQL worksheet
Markdown	—	Documentation	Explain steps to collaborators

R = Python cell + %%R, not a fourth cell type

Add a Python cell, put %%R on the first line, then write R below it. IPython (via rpy2) runs that body in an embedded R interpreter inside the same Python process. Snowsight may display the cell with an R icon or label, but the notebook kernel is always Python — there is no standalone R kernel. See Architecture.

Execution model:

SQL cells run on the warehouse you selected in context, or switch to via USE WAREHOUSE. The warehouse can be changed at any time, and the context will be updated accordingly. This is useful for running SQL queries in different warehouse sizes.
Python kernel runs on ML Container Runtime (compute pool selected for the notebook session)
%%R in a Python cell — R code runs inside the Python process via rpy2. When R calls RSnowflake / dbplyr or snowflakeR sfr_query(), that SQL runs on the virtual warehouse (same as SQL cells), not on the container pool

See R Cells & Interop for data exchange patterns. Container internals (nbctl, uvenv kernel, /filesystem): Appendix G.

7.8 Compute setup

Notebooks use ML Container Runtime for the interactive kernel (compute setup docs):

flowchart LR
  subgraph notebook [Notebook session on compute pool]
    PY[Python kernel]
    Rcell["R via %%R and rpy2"]
  end
  WH[Virtual warehouse]
  PY --> Rcell
  PY -->|Snowpark from Python| WH
  Rcell -->|RSnowflake dbplyr sfr_query| WH
  SQLcell[SQL cells] --> WH

Component	Runs on	Pays
Python kernel (hosts `%%R` / rpy2)	Compute pool / container runtime	Container credits
SQL cells; Snowpark from Python	Virtual warehouse	Warehouse credits
SQL from R (`%%R` + RSnowflake, dbplyr, snowflakeR)	Virtual warehouse	Warehouse credits
Model training via snowflakeR ML APIs	Mixed — ML API may use warehouse + stages	Both

Practical tips:

Size warehouse for SQL-heavy notebooks (including R pipelines that collect() late); size compute pool for package install and kernel memory
Suspend warehouse when idle; end session when done with notebook
Enable EAI before bootstrap (not usually needed if you use a CRE with packages pre-baked) when installing from CRAN/GitHub/pip (Network & EAI)

7.9 What Workspace provides for R teams

Data locality — Query and train without bulk export
Git-backed projects — Reproducible notebooks and configs
Shared ML objects — Same Feature Store / Registry as Python colleagues
Governance — Roles, audit, network rules, tags
Scheduling — Promote notebook to recurring Task

Trade-offs vs desktop RStudio:

Without a CRE: cold-start setup_notebook() bootstrap and EAI setup for outbound installs (Workspace Bootstrap, Network & EAI)
With a CRE: faster startup — R stack baked into the image; bootstrap cell often unnecessary
Notebook UX vs full IDE (Connections Pane, debugger)
Container runtime package constraints for serving (conda-forge) vs interactive training

7.10 Alternatives in the same account

Option	Best for
Workspace Notebooks	Mixed Python/R/SQL, team Git repos, ML exploration
Posit Workbench Native App	Full RStudio/Posit IDE in Snowflake
ML Jobs	Non-interactive batch on container runtime
Local RStudio + RSnowflake	Familiar desktop IDE, `connections.toml` auth
RStudio Server in SPCS	Custom containerized IDE (you operate the image)

7.11 Next steps

Workspace Bootstrap — setup_notebook() and YAML.

Network & EAI — outbound access for installs.

R Cells & Interop — %%R, outputs, path discovery.

--- title: "Snowflake Workspace & Notebooks" subtitle: "Snowsight UX, compute, and what Workspace provides for R" --- ## Overview Before bootstrapping R, understand **where** you work in Snowflake. **Snowsight** is the web UI. **Workspace** is the project-oriented development area within Snowsight. **Notebooks** are interactive, cell-based environments for Python, SQL, and (with this guide and package) R. [Workspace Bootstrap](../06_workspace_bootstrap/index.qmd) covers `setup_notebook()` next. ::: {.callout-important} **Workspace Notebooks required.** The multi-language R stack in this guide (`setup_notebook()`, `%%R`, snowflakeR) runs only in **Workspace Notebooks** on ML Container Runtime — not in legacy standalone Snowsight notebooks. ::: ## Learning Objectives - Navigate Snowsight → Workspace → Notebook - Choose and configure compute for a notebook session - Use Python, SQL, and Markdown cells — and know where R fits - Understand session lifecycle (start, active, restart, end) --- ## Snowsight and Workspace {#sec-snowsight} **Snowsight** provides: - SQL worksheets and query history - Dashboards and Streamlit apps - Admin and governance screens - **Workspace** — projects, files, Git repos, notebooks **Workspace** adds a developer workflow: | Capability | Benefit | |------------|---------| | **Project + files** | Organize notebooks, scripts, configs (`snowflaker_config.yaml`) | | **Git integration** | Clone, commit, push — team collaboration | | **Notebooks** | Cell-based Python/SQL/Markdown (Jupyter-compatible in Workspaces) | | **Compute session** | Attached warehouse + container runtime for kernel | ::: {.callout-note} **Terminology:** "Workspace Notebook" means a notebook running **inside** your Snowflake account — not Jupyter on your laptop. Kernel and data stay in-account. ::: ### Snowsight navigation {#sec-screenshots-snowsight} Screenshots are **cropped to the relevant UI area** — to aid readability. ![Snowsight Home — Projects menu with Workspaces shortcut](../_images/05-snowsight-nav.png){fig-alt="Snowsight left navigation with Projects selected and Workspaces flyout menu" width=72%} ![Welcome to Workspaces — quick-start actions](../_images/05-workspace-landing.png){fig-alt="Snowsight Workspaces home page with Welcome header and Notebook/SQL quick-create buttons" width=72%} ## Workspace notebooks {#sec-workspace-notebooks} All R content in this guide assumes **Notebooks in Workspaces** — the Jupyter-compatible experience with Git integration and ML Container Runtime. Create and open notebooks from **Snowsight → Workspace**. ## Creating and opening a notebook {#sec-create} Typical flow — product walkthrough: [Snowflake Notebooks in Workspaces](https://docs.snowflake.com/en/user-guide/ui-snowsight/notebooks-in-workspaces/notebooks-in-workspaces-overview); [Workspaces](https://docs.snowflake.com/en/user-guide/ui-snowsight/workspaces) (files, folders, Git). 1. Open **Snowsight** → **Workspace** ([Workspaces overview](https://docs.snowflake.com/en/user-guide/ui-snowsight/workspaces)) 2. Select or create a **workspace** (private or shared); optionally connect a **Git** repository ([Integrate workspaces with Git](https://docs.snowflake.com/en/user-guide/ui-snowsight/workspaces-git#create-a-git-workspace)) 3. **Create notebook** (+ **Add new** / upload `.ipynb`) or open an existing notebook file in the workspace ([overview — file management](https://docs.snowflake.com/en/user-guide/ui-snowsight/notebooks-in-workspaces/notebooks-in-workspaces-overview#overview)) 4. **Start session** — connect to a notebook service so the kernel runs on container runtime ([compute setup](https://docs.snowflake.com/en/user-guide/ui-snowsight/notebooks-in-workspaces/notebooks-in-workspaces-compute-setup#setting-up-compute)). When configuring the service (**Connected → Edit**), you can optionally select a **Custom Runtime Environment (CRE)** registered by you or your organisation (`cre@<org_name>`) so R, `%%R`, snowflakeR, and RSnowflake are **pre-installed** — see [CRE & ML Jobs](../09_custom_runtime_and_ml_jobs/index.qmd). If no CRE is available, use the default ML Container Runtime and bootstrap R in step 7. 5. Set **execution context**: role, warehouse, database, schema ([editing and running — set execution context](https://docs.snowflake.com/en/user-guide/ui-snowsight/notebooks-in-workspaces/notebooks-in-workspaces-edit-run#set-the-execution-context)) 6. Run a SQL or Python smoke cell (`SELECT CURRENT_USER()`, etc.) ([run cells](https://docs.snowflake.com/en/user-guide/ui-snowsight/notebooks-in-workspaces/notebooks-in-workspaces-edit-run#run-cells)) 7. **If you did not select a CRE in step 4:** run the Python bootstrap cell (`setup_notebook()`) — [Workspace Bootstrap](../06_workspace_bootstrap/index.qmd) ::: {.callout-warning} Workspace does **not** auto-set database/schema in all versions — set context explicitly in YAML config or SQL, or fully qualify object names. ::: ### Notebook workflow in Snowsight {#sec-screenshots-create} ![Create menu — Notebook, SQL file, Streamlit, and more](../_images/05-create-notebook.png){fig-alt="Snowsight Create dropdown in Workspaces with Notebook and SQL file options" width=72%} ![Workspace notebook editor — Run toolbar and Python/SQL cells](../_images/05-notebook-editor.png){fig-alt="Workspace notebook editor with Run and Connected toolbar, empty Python cell, and SQL smoke query with results" width=90%} ![Session context — role and warehouse selectors](../_images/05-compute-context.png){fig-alt="Workspace notebook Use role and Use warehouse dropdowns in the toolbar" width=90%} ![Notebook session toolbar — Run and Connect while kernel starts](../_images/05-session-controls.png){fig-alt="Workspace notebook Run button and Connecting status while the kernel starts" width=90%} ## Notebook toolbar & session {#sec-toolbar} Workspace session controls ([reference](https://docs.snowflake.com/en/user-guide/ui-snowsight/notebooks-in-workspaces/notebooks-in-workspaces-edit-run)): | Control | Purpose | |---------|---------| | **Start / Active** | Start kernel session; view memory/CPU; restart or end | | **Run all / Stop** | Execute all cells or cancel | | **Package selector** | Install Python packages (Anaconda channel on some runtimes) | | **Scheduler** | Run notebook on a schedule via **Tasks** | | **Connected / Edit** | Role, warehouse, database, schema, **External Access** | | **Export / duplicate** | Share or fork notebooks | For R without a CRE: run the Python bootstrap cell first; packages install via micromamba + YAML tarballs — not the Anaconda package picker alone. ## Cell types {#sec-cells} Workspace notebooks expose three cell types. **There is no separate “R cell” type** — you run R in a **Python cell** whose first line is `%%R`. | Cell type | Kernel / compute | Use | R in this guide | |-----------|------------------|-----|-----------------| | **Python** | Python kernel on compute pool | Snowpark, pandas, `setup_notebook()`; **R via `%%R` magic** (first line of the cell) | Bootstrap once unless you attached a **CRE** in step 4 | | **SQL** | Virtual warehouse | Queries, DDL, context | Same as any SQL worksheet | | **Markdown** | — | Documentation | Explain steps to collaborators | ::: {.callout-important} ## R = Python cell + `%%R`, not a fourth cell type Add a **Python** cell, put `%%R` on the **first line**, then write R below it. IPython (via **rpy2**) runs that body in an embedded R interpreter inside the **same Python process**. Snowsight may *display* the cell with an R icon or label, but the notebook kernel is always Python — there is no standalone R kernel. See [Architecture](../02_architecture/index.qmd). ::: **Execution model:** - **SQL cells** run on the **warehouse** you selected in context, or switch to via `USE WAREHOUSE`. The warehouse can be changed at any time, and the context will be updated accordingly. This is useful for running SQL queries in different warehouse sizes. - **Python kernel** runs on **ML Container Runtime** (compute pool selected for the notebook session) - **`%%R` in a Python cell** — R code runs inside the Python process via rpy2. When R calls **RSnowflake** / **dbplyr** or **snowflakeR** `sfr_query()`, that SQL runs on the **virtual warehouse** (same as SQL cells), not on the container pool See [R Cells & Interop](../08_r_cells_and_interop/index.qmd) for data exchange patterns. Container internals (nbctl, uvenv kernel, `/filesystem`): [Appendix G](../appendices/G_workspace_container_internals/index.qmd). ## Compute setup {#sec-compute} Notebooks use **ML Container Runtime** for the interactive kernel ([compute setup docs](https://docs.snowflake.com/en/user-guide/ui-snowsight/notebooks-in-workspaces/notebooks-in-workspaces-compute-setup)): ```{mermaid} flowchart LR subgraph notebook [Notebook session on compute pool] PY[Python kernel] Rcell["R via %%R and rpy2"] end WH[Virtual warehouse] PY --> Rcell PY -->|Snowpark from Python| WH Rcell -->|RSnowflake dbplyr sfr_query| WH SQLcell[SQL cells] --> WH ``` | Component | Runs on | Pays | |-----------|---------|------| | Python kernel (hosts `%%R` / rpy2) | Compute pool / container runtime | Container credits | | SQL cells; Snowpark from Python | Virtual warehouse | Warehouse credits | | SQL from R (`%%R` + RSnowflake, dbplyr, snowflakeR) | Virtual warehouse | Warehouse credits | | Model training via snowflakeR ML APIs | Mixed — ML API may use warehouse + stages | Both | **Practical tips:** - Size warehouse for SQL-heavy notebooks (including R pipelines that `collect()` late); size compute pool for package install and kernel memory - **Suspend** warehouse when idle; **end session** when done with notebook - Enable **EAI** before **bootstrap** (not usually needed if you use a CRE with packages pre-baked) when installing from CRAN/GitHub/pip ([Network & EAI](../07_network_and_eai/index.qmd)) ## What Workspace provides for R teams {#sec-value} - **Data locality** — Query and train without bulk export - **Git-backed projects** — Reproducible notebooks and configs - **Shared ML objects** — Same Feature Store / Registry as Python colleagues - **Governance** — Roles, audit, network rules, tags - **Scheduling** — Promote notebook to recurring Task **Trade-offs vs desktop RStudio:** - Without a CRE: cold-start **`setup_notebook()`** bootstrap and **EAI** setup for outbound installs ([Workspace Bootstrap](../06_workspace_bootstrap/index.qmd), [Network & EAI](../07_network_and_eai/index.qmd)) - With a CRE: faster startup — R stack baked into the image; bootstrap cell often unnecessary - Notebook UX vs full IDE (Connections Pane, debugger) - Container runtime package constraints for **serving** (conda-forge) vs interactive training ## Alternatives in the same account {#sec-alternatives} | Option | Best for | |--------|----------| | **Workspace Notebooks** | Mixed Python/R/SQL, team Git repos, ML exploration | | **[Posit Workbench Native App](https://posit.co/products/enterprise/posit-workbench/)** | Full RStudio/Posit IDE in Snowflake | | **ML Jobs** | Non-interactive batch on container runtime | | **Local RStudio + RSnowflake** | Familiar desktop IDE, `connections.toml` auth | | **[RStudio Server in SPCS](../04b_rstudio_spcs/index.qmd)** | Custom containerized IDE (you operate the image) | ## Next steps {#sec-next} [Workspace Bootstrap](../06_workspace_bootstrap/index.qmd) — `setup_notebook()` and YAML. [Network & EAI](../07_network_and_eai/index.qmd) — outbound access for installs. [R Cells & Interop](../08_r_cells_and_interop/index.qmd) — `%%R`, outputs, path discovery.