Your first chat

This page walks you through one question from start to finish, using the example cancer genomics data that comes with Polaris.

1. Start a new chat

Select New chat in the sidebar, or press ⌘ ⇧ O (Ctrl Shift O on Windows and Linux). The chat input says “Ask anything about your data”.

2. Pick a model

The model picker sits in the chat input.

Auto uses the default model your admin chose.
Claude Opus 5.5 is the most capable, and the slowest and most expensive.
Claude Sonnet 5.5 is a good balance for most analysis.
Claude Haiku 5.5 is the fastest and cheapest.

Press ⇧ Tab in the input to cycle through models. See Models for details.

3. Ask a question

Type a question and press Enter. For example:

How many patients are in each TCGA cohort, and what share of them are female?

4. Read the answer

While Polaris works, you see each step it takes: reading context files, running SQL, drawing charts. When it finishes, you get:

A written answer.
A SQL card for each query. Select View query to read the SQL and View results to see the rows.
Charts or tables, when they help.
Up to three suggested follow-up questions.

5. Follow up

Ask a follow-up question in the same chat. Polaris keeps the earlier questions and results in mind:

Now break that down by tumour stage.

You can type a follow-up while Polaris is still working. It waits in a queue and is sent when the current answer finishes. To stop the current answer, select the stop button (■) or press Ctrl C.

More to try

Which genes are most often mutated in bladder cancer (TCGA-BLCA)?

Compare VHL mutation rates between KIRC and BLCA, and show it as a bar chart.

Which Open Targets targets for kidney cancer have an approved small-molecule drug?

How many imaging series does IDC have for the TCGA-KIRC patients in our cohort, by modality?

Next, learn what the example data contains, or read more about chats.