Column Comments & AI Context
When you connect a source and select a table, Dashport lets you add a comment to each column. These comments are passed directly to the AI — the clearer and more descriptive they are, the better and more accurate the answers you'll get from Dashport Chat.
Column names alone are often ambiguous — amt, rev, or cust_id mean nothing to an AI without context. A short comment explaining what a column represents, its unit, and any business logic behind it gives the AI exactly what it needs to answer your questions correctly the first time.
Where to add comments
After selecting your table in the connection wizard, Dashport displays a column preview panel. Each column row has a Comment field on the right-hand side — click it to type your description. Comments are saved with the connection and used every time Chat queries that source.
Examples: vague vs. clear
| Column name | Poor comment | Clear comment |
|---|---|---|
order_date | Date | The date the customer placed the order (UTC, YYYY-MM-DD). Use this as the primary time axis for all revenue trends. |
amt | Amount | Total order value in USD, after discounts, before tax. This is the main revenue figure used in financial reports. |
cust_id | Customer ID | Unique identifier for the customer. Join to the customers table on this column to get name, region, and plan tier. |
status | Status | Order lifecycle status. Values: pending, processing, completed, refunded, cancelled. Only 'completed' counts as recognised revenue. |
mrr | MRR | Monthly Recurring Revenue for this subscription in USD. Excludes one-time charges and trial accounts (trial = 1). |
product_sku | SKU | Internal product code. Maps to the products table. Use the product_name column from that table for user-facing labels. |
Tips for writing good comments
- State the unit. Always say whether a value is in USD, GBP, minutes, percentage points, etc. — never assume it's obvious.
- List possible values for status/flag columns. Enumerated fields like
statusorplan_tiershould have all valid values listed so Chat can filter them correctly. - Explain joins. If the AI needs to combine this column with another table to get a human-readable label, say so in the comment.
- Clarify business rules. If only certain rows count toward a metric (e.g. only
status = 'completed'is revenue), document it here so Chat applies the right filter automatically. - Note the time zone. For date and timestamp columns, specify whether values are UTC, local, or something else — this prevents off-by-one errors in daily and hourly charts.
- Keep it concise but complete. Two or three sentences is ideal. You don't need an essay — just enough that someone unfamiliar with your schema would understand the column immediately.