Skip to main content

Advanced Features

Now that you've used both intents — Transform and SQL — let's explore the features that help you work more effectively with Visitran AI.

Context Rules​

Context rules let you provide standing instructions to the AI. Think of them as system prompts that shape every response.

Click the Context Rules icon (document icon) in the chat header to open the panel:

Personal Rules​

Your global instructions that apply to all projects. Use these for preferences like:

  • "Always use snake_case for column names"
  • "Prefer LEFT JOINs over INNER JOINs unless I specify otherwise"
  • "Include row counts in your responses when running SQL"

Project Rules​

Instructions specific to the current project. Useful for:

  • "The 'amount' column in raw_payments is in cents — divide by 100 for dollar values"
  • "Always filter out orders with status = 'returned'"
  • "Use 'analytics' as the target schema for all models"

Both rule types support up to 4,000 characters and include timestamps showing when they were last updated.

Managing Conversations​

Viewing Past Conversations​

Click the History icon (clock) in the chat header to see all conversations for the current project:

Each conversation shows its name, creation date, and a preview of the last message. Click to reopen any conversation.

Renaming Conversations​

Conversations are auto-named based on your first prompt. To rename, click the conversation title in the header and type a new name.

Starting a New Chat​

Click the + button in the header to start a fresh conversation. Your previous conversations are preserved in the history.

Token Usage​

Each prompt consumes tokens based on the length of your input, the context provided, and the response generated. You can monitor usage:

  • Per-message breakdown — See input/output token counts for each interaction
  • Balance indicator — Your remaining token balance is displayed in the chat header

Resetting the Onboarding​

Completed the onboarding but want to run through it again? Click the Reset Onboarding button (rocket icon) in the chat header. This resets all task progress and restarts the guided walkthrough.

Tips for Effective Prompts​

Be Specific About Your Data​

Instead of:

"Create a customer model"

Try:

"Create a customer model by joining raw_customers with raw_orders on customer id, and include total order count and most recent order date"

Reference Tables with @Mentions​

Use @table_name to explicitly reference tables. This removes ambiguity and helps the AI locate the right data:

"Join @raw_orders with @raw_payments on order_id"

Specify Column Names When Possible​

If you know the column names, include them:

"Group by payment_method from @raw_payments and calculate SUM(amount)"

Use Follow-Up Prompts for Refinement​

Don't try to get everything perfect in one prompt. Start simple and iterate:

  1. "Create a basic orders model joining orders with payments"
  2. "Add customer names by joining with the customers table"
  3. "Add a column for the number of days between order and payment"

Set Context Rules for Repeated Preferences​

If you find yourself repeating the same instructions, add them to your Context Rules instead:

  • Column naming conventions
  • Default schemas
  • Data quality filters
  • Business logic assumptions

What's Next?​

You've now covered all the core Visitran AI capabilities. Here are some next steps: