Building customer-visible data science dashboards with Altair / Vega / Vue
summary
Argues that shipping customer-facing dashboards requires a compromise: Python generates Vega/Vega‑Lite specs (Altair converts pandas data frames to JSON) while a minimal Flask backend serves them and a Vue front end uses vega‑embed to render and hot‑reload. Prototype needs a few Python lines plus a few JavaScript lines, about 800 npm deps to install, and it works — faster UI and happier front‑end teams; plotting millions of points is possible but becomes slow without server‑side aggregation. Flask backend returns Vega‑Lite JSON Altair generates specs from pandas Vue plus vega‑embed renders and hot‑reloads
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:03Why dashboards matter
- 2:31Python-only limits
- 4:58Front-end vs back-end needs
- 5:59Vega and Vega-Lite
- 7:26Front-end framework choice
- 9:12JavaScript tools
- 10:51Demo setup Flask Vega-Lite
- 13:05Altair generates Vega
- 14:30Altair in Jupyter
- 15:34JavaScript front-end trade-offs
- 16:16Vega-Embed rendering
- 18:25Interactive UI components
- 19:51Dashboard styling
- 20:33Tech stack summary
- 23:03Key takeaways
- 23:46Call for papers
- 24:27Q&A
related talks
28mScaling Altair visualizations with VegaFusion
42mAltair Declarative, Statistical Visualization for Python
30mMeta-Dashboards: Accelerating Geospatial Web Apps Creation with Voilà
42mInteractive Data Analysis: Visualization and Beyond
41mEffective Dashboard Design Using
28mBuilding Data-Intensive Visualization Applications with Vue
32mDash: Interactive Data Visualization Web Apps with no Javascript
34mTo The Web and Beyond