Skip to main content

Slash Commands

Special modes that change how the AI interprets your requests.

What are Slash Commands?​

Slash commands are keywords prefixed with / that activate specific AI behaviors. Instead of immediately executing your request, these commands make the AI plan, recommend, or focus on particular tasks like visualization.

Currently available: /plan, /visualize, /recommendations, and /taxonomy

Slash Command Usage

Type a slash command at the start of your message

Why Slash Commands Matter​

Control AI behavior: Tell the AI how to approach your request, not just what you want.

Prevent mistakes: Use /plan to review the AI's approach before execution.

Focused modes: Get better results when the AI knows you specifically want charts or suggestions.

Efficiency: Skip back-and-forth by being explicit about intent upfront.

Available Commands​

/plan​

What it does: AI explains its assumptions and approach before creating Views.

When to use it:

  • Complex requests where the AI might misunderstand
  • When you want to verify logic before execution
  • Learning how the AI breaks down problems
  • Ensuring data joins and filters are correct

Example:

/plan @[orders] and @[customers] Show customer lifetime value,
segmented by acquisition channel

AI response:

Here's my plan:
1. Join orders with customers on customer_id
2. Sum total order value per customer
3. Group by acquisition_channel
4. Calculate average LTV per channel
5. Create summary View

Does this approach look correct?

You can then approve, modify, or clarify before the AI executes.

/plan Response

/plan shows AI's approach before executing

/visualize​

What it does: Focuses the AI on creating charts and visualizations.

When to use it:

  • When you want a chart, not just data
  • To specify chart type and styling upfront
  • For dashboard-ready visualizations

Example:

/visualize @[monthly_revenue] as a line chart with a trend line,
colored blue, showing the last 12 months

AI response: Creates the chart with all requested specifications, optimized for visual presentation.

Without /visualize, the AI might create data Views and ask if you want a chart afterward. With /visualize, it knows that's the goal.

/visualize Output

/visualize creates charts directly

/recommendations​

What it does: AI suggests relevant analyses you might want to try based on your data.

When to use it:

  • You're not sure what to analyze
  • Looking for inspiration or next steps
  • Want to discover insights you hadn't considered

Example:

/recommendations @[sales_data]

AI response:

Based on your sales data, here are suggested analyses:
1. Revenue trends over time (detect seasonality)
2. Top-performing products and categories
3. Customer cohort analysis
4. Geographic sales distribution
5. Average order value by customer segment
6. Inventory turnover rates

You can then pick one and ask the AI to execute it.

/recommendations Output

/recommendations suggests analyses to try

/taxonomy​

What it does: Automatically discovers and generates data-driven taxonomies for classifying data using AI tables.

When to use it:

  • You have unstructured or freeform text that needs categorization
  • You want to classify data into meaningful groups but aren't sure what categories exist
  • Creating AI tables for text classification (support tickets, product categories, etc.)
  • You need to explore data patterns before deciding on categories

Example:

/taxonomy @[support_tickets] Classify by customer intent based on subject and message body

AI response: The AI will:

  1. Explore your data - Analyze cardinality, frequency distribution, and patterns
  2. Propose categories - Suggest 3-7 meaningful categories with coverage statistics
  3. Validate with you - Present the proposed taxonomy for approval/adjustment
  4. Create AI table - Once approved, automatically create an AI table with the taxonomy

The process ensures you review category definitions before the AI starts classifying data.

/taxonomy Workflow

/taxonomy generates data-driven classification categories

How to Use Slash Commands​

Basic Syntax​

Start your message with the slash command:

/plan Your request here
/visualize Your chart request
/recommendations @[dataset_name]

Combining with @Mentions​

Slash commands work perfectly with @mentions:

/plan Join @[orders] with @[products] and calculate profit margins
/visualize @[monthly_users] and @[monthly_revenue] on the same chart
/taxonomy @[feedback_data] Group by topic

Common Use Cases​

Complex analysis verification:

/plan @[transactions] Calculate rolling 7-day average, then identify
anomalies more than 2 standard deviations from the mean

Dashboard chart creation:

/visualize @[kpis] as a dashboard with 4 metric cards showing
total revenue, user count, conversion rate, and average order value

Starting a new analysis:

/recommendations @[customer_feedback]

Creating a classification taxonomy:

/taxonomy @[support_tickets] Classify by customer issue type based on subject and description

Preventing misinterpretation:

/plan @[sales] Filter to last 90 days, group by week, then calculate
week-over-week growth rate

Tips & Best Practices​

Use /plan for complex requests: If your request has multiple steps or joins, plan first.

Use /visualize for specific chart needs: If you know exactly what chart you want, specify it with /visualize.

Use /recommendations when stuck: Don't know where to start? Get AI suggestions.

Use /taxonomy for classification: When you need to categorize text or create classification AI tables, let the AI discover patterns and propose categories.

Review plans carefully: When using /plan, read the AI's approach thoroughly before approving.

Iterate on recommendations: /recommendations gives ideas—pick one and chat to refine it.

Validate taxonomies before creation: When using /taxonomy, carefully review the proposed categories and definitions before the AI creates the AI table.

Combine commands thoughtfully: You can't use multiple slash commands in one message—pick the most appropriate one.

Understanding Plan Mode Responses​

When you use /plan, the AI will:

  1. Describe data sources it will use
  2. List transformations step by step
  3. Explain assumptions about your data
  4. Ask for confirmation before proceeding

You can respond with:

  • "Yes, proceed" or "Go ahead"
  • "No, instead do X" (to correct the approach)
  • Questions about specific steps

Plan Approval Flow

Approve, modify, or ask questions about the plan