Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/dataease/SQLBot/llms.txt

Use this file to discover all available pages before exploring further.

Chat is the primary way you interact with SQLBot. You type a question in plain language — “What were our top 10 products by revenue last quarter?” — and SQLBot translates it into SQL, runs it against your connected database, and presents the results as a table and chart. You can keep the conversation going with follow-up questions that build on earlier answers, all without writing a single line of SQL.

​
Starting a new chat session

1

Click New Chat

From the left sidebar, click New Chat or the compose icon. A blank chat window opens.
2

Select a datasource

Use the datasource picker at the top of the chat window to choose which database you want to query. Only datasources connected to your workspace appear here.
If no datasources appear, ask your workspace admin to add one. You cannot start a chat without at least one available datasource.
3

Ask your first question

Type your question in the input box and press Enter or click Send. SQLBot begins processing immediately.
4

Review the results

SQLBot returns:
  • The data table with rows and columns from the query result.
  • A chart or visualization automatically selected based on the result shape (bar chart, line chart, pie chart, and so on).
SQLBot streams the response progressively. You will see the SQL being constructed and data loading in real time before the final chart renders.

​
What SQLBot returns

Each response in a chat session contains several parts:
A structured table of query results with sortable columns. For large result sets, the table is paginated. You can click any column header to sort.
SQLBot selects an appropriate chart type based on the structure of your results. A time-series result typically produces a line chart; a categorical breakdown produces a bar or pie chart. You can switch chart types using the controls above the chart.
Expand the SQL panel below the chart to see the exact query that was executed against your database. This is useful for verifying accuracy or copying the query for use elsewhere.
SELECT
    product_name,
    SUM(revenue) AS total_revenue
FROM orders
WHERE order_date >= DATEADD(quarter, -1, GETDATE())
GROUP BY product_name
ORDER BY total_revenue DESC
LIMIT 10

​
Asking follow-up questions

You do not need to start a new chat for each question. Ask follow-up questions in the same session to refine or extend your analysis:
  • “Break that down by region.”
  • “Show me the same metric for the previous quarter.”
  • “Which of those products had the highest return rate?”
SQLBot maintains context across the conversation, so follow-ups can reference your earlier questions implicitly.
If a follow-up question produces unexpected results, be more explicit. Say “of the top 10 products we just listed” rather than “of those products.”

​
Regenerating an answer

If you are not satisfied with a result, you can ask SQLBot to try again. Use the regenerate option on any response to produce a new SQL query for the same question. This is useful when the first attempt misunderstood an ambiguous term.

​
AI analysis and prediction

On any result that has produced a chart, you can trigger two additional AI actions:
  • Analysis: SQLBot provides a written interpretation of the data — identifying trends, outliers, and patterns.
  • Prediction: SQLBot generates a forecast based on historical patterns in the result data.
Both actions stream their output back to the chat window as a follow-up message.

​
Exporting results

To export the data from any response:
  1. Hover over the response you want to export.
  2. Click the Export icon (Excel/CSV).
  3. An .xlsx file downloads with the result data formatted into a spreadsheet, using the chart’s configured axis labels as column headers.
Only data you personally queried can be exported. You cannot export results from other users’ chat sessions.

​
Tips for asking better questions

Vague time references like “recently” or “last year” are interpreted by the LLM and may not match your intent. Prefer explicit ranges:
  • Instead of: “What were sales recently?”
  • Try: “What were total sales between January 1 and March 31, 2025?”
SQLBot performs better when you reference actual column names or business terms defined in your terminology library:
  • Instead of: “Show me performance by location.”
  • Try: “Show me GMV and order count by city for Q1 2025.”
When you want totals, averages, or counts, say so directly:
  • Instead of: “How are users doing?”
  • Try: “What is the average session duration per user in the last 30 days?”
Very broad queries (all rows across years of data without filters) can hit database timeouts. Add filters to keep result sets manageable:
  • Add: “WHERE region = ‘APAC’”
  • Add: “LIMIT to the top 20 results”

​
When SQLBot cannot answer

If SQLBot cannot generate valid SQL for your question, it tells you directly rather than returning incorrect results. Common reasons include:
  • The question references tables or columns that are not enabled in the datasource.
  • The question is too ambiguous for the LLM to resolve without additional context.
  • The required calculation is not expressible in standard SQL for your database engine.
When this happens, try rephrasing with more specific column or table names, or add relevant terms to the Terminology library so SQLBot understands your domain vocabulary.
SQLBot only executes read queries (SELECT). It cannot INSERT, UPDATE, DELETE, or modify your database in any way.

​
Renaming and organizing chats

Each chat session is saved automatically in the left sidebar. To rename a session, click the three-dot menu next to the chat name and select Rename. Giving sessions descriptive names makes it easier to return to previous analyses.

Build docs developers (and LLMs) love

Get started for freeTalk to us