My Insight Canvas

Inspiration

Normal dashboards make it hard to combine data from different databases (like sales in one place, inventory in another). People end up writing complicated code or doing everything manually. We wanted to fix that.

What it does

My Insight Canvas lets you ask normal questions in English like: “Compare sales and inventory across all regions last year”

The system automatically:

Finds the right data from multiple databases
Combines everything correctly
Shows nice charts

No SQL. No manual work.

How we built it

We created three AI helpers that work together like a team:

Planner — reads your question and makes a simple step-by-step plan
Data Worker — writes SQL for each database separately, gets the numbers, then uses Python to join everything
Chart Maker — looks at the final numbers and picks the best chart (bar, line, etc.) automatically

Users first create easy “Data Cubes” (simple descriptions of their tables) so the AI understands the data better.

Challenges we ran into

Joining data from totally different databases is hard → solved by using Python instead of forcing SQL
SQL sometimes fails → added automatic retry + fix logic
Choosing good charts is tricky → made the AI think carefully about the data

Accomplishments that we're proud of

Works across different databases without any manual joining
Almost never crashes — fixes its own mistakes
Creates professional-looking charts completely automatically
Very easy for non-technical people to use

What we learned

Breaking big problems into small steps (Planner → Worker → Chart) works much better
Using Python to combine data is way more flexible than SQL
AI needs to fix its own errors to be reliable
Simple data descriptions (“Data Cubes”) make the AI much smarter