01

The Problem with AI + Data

Context is Everything

Your AI tools are only as good as the context they have. Today, most organizations are building that context by hand, maintaining lists of tables and SQL queries that go stale the moment someone writes them.

There is a better way.

02

First Principles

Humans should not have to create context.

Your organization already has high-quality, curated context about your data. It just needs to be extracted, not recreated.

  • Query knowledge (validated SQL from dashboards and BI tools)
  • Query standards (how your team writes joins, aggregations, filters)
  • Presentation standards (formatting, naming, visualization conventions)
  • Agentic behaviors (when to explore, when to checkpoint, when to ask)

Humans should not have to maintain context.

Any system that depends on people maintaining a golden record set of queries or standards will fail. It creates a dependency on individuals, differential quality across teams, and a single point of failure.

Context should improve over time and work for everyone.

Improving context should be part of people's organic workflows. AI is better at cataloging and maintaining knowledge than humans are. Context should compound as your team works, not decay when someone leaves.

03

A Better Way

Introducing Dante

A two-part system that extracts, maintains, and delivers context to AI coding tools automatically.

FREE / OPEN SOURCE

Dante-DS

For the individual data scientist

A Python library that creates local context from your BI platforms and warehouse schema. Connects to your dashboards, extracts the SQL behind every chart, and converts it into semantic embeddings that your AI tools can search.

Within minutes, Claude Code or Cursor will have the expertise of your entire analytics team.

$pip install dante-ds
Dante includes a built-in management UI for configuring connections, managing knowledge, and running ingestion.
04

Under the hood

From dashboards to embeddings.

Dante connects to your BI platforms, extracts the SQL behind every chart, simplifies it, and stores it as a searchable embedding. When you ask a question, the AI finds the closest matching queries and adapts them.

01 CONNECT

BI Platform

Looker, Redash, Superset, or your warehouse schema

02 EXTRACT

Dashboard SQL

Every chart's query, simplified and paired with a natural language question

03 EMBED

Vector Index

Stored locally. Searched automatically when you ask questions.

The more dashboards your organization has, the more context the AI starts with. No manual curation required.