Examples

Every notebook under examples/ in the repo demonstrates a specific Strata capability. Each example is also rendered into the docs site automatically: click through to read the cell sources and any cached outputs without cloning the repo.

To run examples locally, point the server's storage root at the repo's examples/ directory and start in personal mode:

STRATA_NOTEBOOK_STORAGE_DIR=$PWD/examples \
STRATA_DEPLOYMENT_MODE=personal \
  uv run strata-notebook

Every example then appears on the Strata home page. No copying required: open the notebook, click Run, edit, watch the cascade.

(If you'd rather keep examples separate from your own work, copy the directory you want under your existing storage root instead: cp -R examples/iris_classification ~/.strata/notebooks/.)

Walkthroughs: start here

Notebook	What you'll see
iris_classification	End-to-end ML in seven cells: load → split → train → evaluate → plot. The canonical "multi-cell DAG with caching" demo.
pandas_basics	Core DataFrame operations and rich display outputs.
titanic_ml	Feature engineering plus comparison of two classifiers, closer to a real ML workflow.
data_viewer	A DataFrame larger than the inline preview, in the interactive viewer: page through and sort the full cached artifact.
agent_demo	The notebook a coding agent builds live in the strata agent demo: an expensive training cell stays cached while only the evaluation re-runs.

Variant cells

Notebook	What you'll see
model_variants	Three classifier variants (logistic regression, random forest, gradient boosting) sharing one DAG slot. Switch tabs, re-cascade downstream; the others stay cached.
model_variants_sweep	Sweep mode: every variant runs and one downstream cell compares them as a {variant: value} dict, plus a # @per_variant fan-out.

Prompt cells and AI

Notebook	What you'll see
arxiv_classifier	AI-powered paper classification: prompt cells + distributed workers in one pipeline.
review_triage	Structured-output prompt cells with @output_schema + the validate-and-retry loop.
news_alpha_trader	Secret manager + AI pricing lookup + multi-cell DAG.

SQL cells

The server needs the sql and sql-sqlite extras for this one (uv tool install "strata-notebook[sql,sql-sqlite]"); the Docker image carries both.

Notebook	What you'll see
sql_orders_report	Named connections, bind parameters from Python upstream, schema-aware caching.

Loop cells

Notebook	What you'll see
loop_hill_climb	# @loop with carry state and an early-termination predicate.

Widget cells

Notebook	What you'll see
widget_playground	A widget cell (slider, number, dropdown) driving a downstream DataFrame, with # @live recomputing on every change.

Library cells

Notebook	What you'll see
library_cells	Cross-cell def / class sharing via the synthetic-module slicing path.

Mounts

Notebook	What you'll see
s3_mount	# @mount annotation makes an S3 prefix available as a local pathlib.Path inside the cell.

Markdown rendering

Notebook	What you'll see
markdown_showcase	Every markdown rendering path: headings, lists, tables, code blocks, security guards, dynamic Markdown(...) output.

R cells

Notebook	What you'll see
r_mtcars_analysis	A pure-R notebook - lm(), aggregate(), and inline ggplot2 + base-graphics plots, with data.frame and R-only (RDS) handoff between R cells.
r_lm_vs_sklearn	Mixed Python + R: fit a model with R's lm() and compare side-by-side with scikit-learn over a cross-language Arrow handoff.