Commerce analytics code interpreter

Ask store questions. Generate Python. Preview trusted answers.

A polished first demo for a merchant-facing data analyst that turns BigCommerce questions into pandas analysis, matplotlib charts, and R2-backed artifacts through the Cloudflare Sandbox SDK path.

Dataset cards

Select a built-in dataset or stage a local CSV. The current upload control does not move bytes yet; it previews the intended R2 handoff safely in-browser.

Upload merchant CSV

Optional mock input. Select a file to update the staged dataset label.

Choose CSV

Analyst workspace

Natural language examples call /api/analyze for AI-generated Python and insight narrative, with local fallback output if AI is unavailable.

Ask a commerce question

Use a suggested query or write your own. Workers AI generates the analyst plan; fallback output keeps demos resilient.

Architecture notes

The demo is live-capable for planning and narrative generation, and the seams map directly to a production Cloudflare Sandbox SDK code interpreter flow.

Worker UI + API

Serves the analyst surface, validates requests, calls Workers AI, creates analysis sessions, and proxies status/results to the browser.

R2 datasets

Stores uploaded CSVs, generated notebooks, chart PNGs, and execution logs with signed URLs for result retrieval.

Sandbox SDK

Creates isolated code interpreter sessions, writes files, installs Python dependencies, and executes generated commands.

pandas + matplotlib

Runs dataframe transforms, aggregations, forecasting helpers, and chart rendering without executing code inside the Worker.