Skip to Content
TutorialsUsing the AI Chat

Using the AI Chat Agent

The AI chat agent can build workflows, query databases, and analyze data — all from plain English. This tutorial shows you how.

Starting a Chat

Open the chat panel on the right side of the screen. You’ll see the default agent is ready.

Example 1: Build a Workflow

Type:

@canvas build a Python pipeline that loads a CSV file, filters rows where age > 30, and shows a histogram of the remaining ages

The Canvas agent will:

  1. Create a new flow
  2. Add a Python node to load the CSV
  3. Add a Python node to filter the data
  4. Add a Python node to plot the histogram
  5. Connect all nodes with edges

Example 2: Query a Database

Type:

Search PubMed for papers about CRISPR gene editing published in 2024

The agent will use the query_entrez_database tool to search PubMed and return results.

Example 3: Analyze Data

Type:

I have a counts matrix at /app/data/counts.csv. Calculate the top 10 most variable genes and create a heatmap.

The agent will write and execute Python code to:

  1. Load the CSV
  2. Calculate variance per gene
  3. Select top 10
  4. Generate a heatmap with matplotlib/seaborn

Using the Canvas Agent

Prefix your message with @canvas to use the specialized Canvas agent:

  • @canvas add a Python node called Preprocess
  • @canvas connect the Preprocess node to the Analysis node
  • @canvas verify the flow structure

Tips

  • Be specific about file paths and parameters
  • The agent can see your current canvas state
  • Ask for explanations: “Why did you choose that method?”
  • Request modifications: “Change the threshold to 0.01 instead”

Next: Custom Nodes

Last updated on