nicsuzor/streamlit

Streamlit — Presentation Layer (academicOps)

This skill is the Streamlit-specific implementation of the presentation layer described in the tech-agnostic analyst skill (aops-tools). The analyst skill owns the principle (the presentation layer DISPLAYS pre-computed data — it never transforms, joins, aggregates, or applies business logic; that all lives in the transformation layer). This skill owns the Streamlit how-to.

Streamlit is one swappable choice of presentation layer. The display-only rule holds regardless of which dashboard tool you use; only the patterns below are Streamlit-specific.

Contents

• [[streamlit-workflow]] — single-step collaborative workflow for building Streamlit dashboards (load → STOP → chart → STOP → interactivity → STOP).
• [[streamlit-patterns]] — design patterns and best practices for research dashboards.
• [[streamlit]] — standard app structure and additional Streamlit patterns.

When to use

• The project has a Streamlit app (streamlit/ directory or .py files using st.).
• You need to display pre-computed metrics, render charts, or add interactive filtering on EXISTING columns.

Hard boundary

Streamlit may read (SELECT * FROM mart), filter on existing columns, format for display, and render charts. It must NEVER GROUP BY/aggregate, JOIN, apply CASE business logic, or compute derived metrics inline. If tempted to transform: STOP and add a model in the transformation layer (see the dbt skill) instead.