Notebook Templates

What Notebook Templates are for

Templates provide a useful first investigation without locking you into a fixed report. Choose a template, bind the required Arizmic context, create your own project copy, and then edit the ordinary Marimo .py source as needed.

Every built-in template begins by showing:

• the selected object and exact version;
• relevant lineage and retained hashes;
• environment, SDK, and Marimo versions;
• retention, validation, and available-artifact state;
• whether a view is complete, paged, sampled, or truncated; and
• which values are engine-reported and which are calculated in the Notebook.

Built-in source is immutable and versioned. An update never overwrites your project. Create a new copy or review a comparison before adopting changes.

Launch template library

The V1 launch library contains four required templates. A fifth template, Data Quality, ships only if its Data Manager dependency is complete and certified.

Portfolio review, execution diagnostics, and provider-specific investigations are later candidates, not promised launch templates. Save a recurring custom investigation as a personal template instead of waiting for every use case to become built in.

Starter

Use Starter when you are new to Notebooks or want a blank investigation with the Arizmic connection already explained.

Its first useful view includes:

• connection and provenance status;
• the first page of recent Datasets and Prepared Data;
• the first page of recent Studies and Run Outputs;
• one typed, bounded Research SDK example; and
• links for starting a context-specific project.

Starter remains useful in an empty workspace. Instead of failing on an empty list or unreachable local API, it shows one clear state and the next action. It never loads an unbounded catalog.

Choose Starter when you want to build a custom investigation from the smallest supported foundation.

Run Review

Use Run Review for one exact Run Output. Select the Run in Results and choose Open in Notebook, or choose the template in Notebooks and bind the Run manually.

The template presents:

• Strategy and Strategy-version identity;
• Dataset and Prepared Data lineage;
• parameters, execution fidelity, retention, and validation state;
• engine-reported performance values;
• one bounded price or equity window;
• paged trades and supported trade analytics;
• a selected-trade drill-down window; and
• optional evidence or telemetry only when it is retained and supported.

The first render does not load every bar, event, or trade. Select a smaller window or page before requesting more detail. Values shown as engine-reported must match the corresponding Results tabs exactly.

Use this template to investigate why a run behaved as it did, build a custom visual, or prepare a review—not to replace the authoritative Run Output.

Study Review

Use Study Review for a Strategy or Portfolio Study and its retained Study Result.

The template can show:

• Study identity, status, method, subject, and input lineage;
• the engine-reported Study summary;
• paged trials and selected candidates;
• a bounded parameter surface when supported;
• folds or cross-asset blocks when retained;
• validation and missing-output states; and
• a handoff into Run Review for an exact candidate Run Output.

The template adapts through the Study's declared capabilities rather than special-casing a particular Study ID. Optional surfaces remain explicitly unavailable when the Study did not produce them.

Use Study Review to understand a search or validation exercise as a whole and then open individual Runs only when closer inspection is useful.

Robustness

Use Robustness when a Study or Result contains retained information about stability and sensitivity.

The template can organize:

• parameter-neighborhood and plateau views;
• candidate sensitivity;
• consistency across retained validation segments;
• cross-asset evidence where available; and
• optional local statistics with their formulas and sample limits visible.

Robustness does not manufacture missing evidence. Sparse or unavailable inputs stay visibly incomplete. A locally calculated surface, projection, or statistic is labeled notebook-derived and does not change an Arizmic validation state.

Use the template to ask better follow-up questions and design another Study, not to turn an attractive chart into a promotion decision.

Data Quality

Data Quality is conditional. It is available only when Data Manager exposes the bounded, calendar-aware quality summary required by the Notebook plan.

When supported, it can show:

• Dataset identity, provenance, schema, instruments, and coverage span;
• ordering, duplicate, overlap, gap, and coverage findings owned by Data Manager;
• checksums and related Prepared Data; and
• a capped drill-down into exact findings.

The template must not scan an entire Dataset locally to recreate quality rules or infer quality from metadata alone. If the backend-owned summary is missing, the template remains Blocked rather than presenting a shallow substitute.

Create a project from a template

1. Open Notebooks.
2. Check that the workspace and environment are ready.
3. Choose Create from template.
4. Review the template's accepted context kind and required SDK capabilities.
5. Select the exact Dataset, Study, Study Result, or Run Output where required.
6. Enter a project name and review the proposed .py path.
7. Create the project.
8. Review its context, template version, source hash, and dependency changes.
9. Complete the imported-code trust acknowledgement when required.
10. Launch the project and confirm its identity and status header before using the analysis.

A deep link from Data, Studies, or Results can prefill steps 4 and 5. It does not create or launch the project automatically.

Read availability and blockers

A template may be:

• Ready — its required SDK, context, environment, fixtures, and launch proof are current;
• Unavailable for this context — the selected object lacks a required output or uses the wrong kind;
• Blocked — an owning product capability, dependency, or certification is incomplete; or
• Update available — a newer built-in source exists, while your project remains unchanged.

A .py file that merely compiles is not enough to make a template Ready. Each built-in must also pass bounded-read, semantic parity, real-launch, interaction, export, and responsive accessibility checks.

Customize without losing provenance

Once created, the project is yours. You can add controls, cells, charts, calculations, and commentary. Keep the template origin and version visible so another reader can distinguish the built-in starting point from your changes.

When you reuse the investigation regularly, save it as a personal template. Review its dependencies, context requirements, and secret exclusions before making another copy.

Do not edit the built-in registry, interpolate executable text into a template, or import private Arizmic services. Use the public arizmic.notebooks SDK and normal project source.

Troubleshoot a template

• The template is blocked: open its blocker details. Fix the named environment, SDK, context, or owning-product requirement rather than copying data through an unsupported path.
• The context is unresolved: reopen the exact object, verify it still exists, and bind the current supported version.
• A chart is empty: check whether the requested artifact and time window were retained. Empty must not be treated as zero.
• The view is truncated: use the template's page or window controls instead of increasing an unbounded local read.
• Engine values differ from Results: stop using the project and compare the exact object/version, SDK response origin, and template version.
• An update is available: create a new copy or review the proposed diff. Never replace your only edited source.
• Data Quality is unavailable: use Data Manager's own quality and lineage views until the bounded Notebook contract is certified.

Next

Use Python Notebooks for projects, environments, the Research SDK, and Marimo behavior. Use AI in Notebooks to draft or revise a project through reviewed proposals.