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.

A datasource is SQLBot’s connection to your data. Before you can ask any natural language questions, you need to connect at least one database or upload a file. SQLBot uses your datasource to introspect the schema, surface table and field metadata to the LLM, and execute the generated SQL. This page walks through every step of adding and configuring a datasource.

​
Supported database types

SQLBot can connect to the following systems out of the box:

MySQL

PostgreSQL

Microsoft SQL Server

Oracle

ClickHouse

Elasticsearch

AWS Redshift

Apache Hive

Apache Doris

StarRocks

DM (达梦)

Kingbase

SQLite

Excel / CSV

Only workspace admins can add, edit, or delete datasources. Regular workspace members can query existing datasources but cannot modify connection settings.

​
Adding a database connection

1

Navigate to Datasources

In the left sidebar, click Datasources. The list shows all datasources already connected to your workspace.
2

Click Add

Click the Add button in the top-right corner. A configuration panel opens.
3

Choose the database type

Select your database engine from the type dropdown. The form adapts to show the correct connection fields for that engine (host, port, database name, username, password, and any engine-specific options).
4

Fill in connection details

Enter the connection parameters. For example, a PostgreSQL connection requires:
Host:     db.example.com
Port:     5432
Database: analytics
Username: sqlbot_reader
Password: ••••••••
SQLBot stores connection credentials encrypted. Use a read-only database account to limit what SQLBot can access.
5

Test the connection

Click Test Connection. SQLBot attempts to connect and returns a success or failure status. If the test fails, double-check host reachability, credentials, and firewall rules before saving.
6

Save the datasource

Click Save. The datasource appears in the list with a status indicator showing whether the connection is active.

​
Selecting tables

After saving a datasource, SQLBot fetches all available tables from the database. You must explicitly select which tables SQLBot is allowed to query. This controls both what the LLM sees in its schema context and what SQL it is permitted to generate.
1

Open the datasource

Click the datasource name in the list to open it.
2

Go to the Tables tab

Switch to the Tables tab. You will see every table in the connected database.
3

Enable the tables you want

Toggle on each table that SQLBot should be able to query. Untoggled tables are excluded from schema context entirely — the LLM will not know they exist.
4

Save your selection

Click Save to confirm the table selection. SQLBot immediately updates its schema context for this datasource.
If you add new tables to your database later, you must return here and manually enable them. SQLBot does not automatically detect schema changes.

​
Annotating table and field metadata

For each enabled table, you can add a description (called a custom comment) to both the table itself and each of its columns. These descriptions are injected into the LLM’s schema context and directly affect query accuracy. To edit metadata:
  1. Click a table name to expand it.
  2. Enter a plain-language description in the Table description field. For example: Stores one row per completed customer order, including totals and shipping status.
  3. For each column, add a Field description explaining what it contains. For example, a column named gmv might get the description Gross merchandise value in USD, before refunds.
  4. Click Save.
Descriptions are especially important for columns with non-obvious names, numeric IDs that join to other tables, or status codes stored as integers. The more context you provide, the more accurately SQLBot can map your questions to the right columns.
You can also export and bulk-import metadata via Excel:
  • Export schema downloads an .xlsx file with all your tables and field descriptions pre-filled.
  • Import schema lets you fill in or update descriptions in the spreadsheet and re-upload it, applying all changes in one batch.

​
Uploading Excel or CSV files

Excel and CSV files are treated as a special datasource type. SQLBot imports the file into its internal PostgreSQL store, making the data queryable just like any other database.
1

Click Add and select Excel/CSV

On the Add datasource form, choose Excel/CSV as the type.
2

Upload your file

Drag and drop or browse to your .xlsx, .xls, or .csv file. SQLBot previews the detected columns and inferred data types before importing.
3

Confirm column types

Review the inferred field types (string, integer, decimal, and so on). Adjust any types that were detected incorrectly, then click Import.
Supported file formats: .xlsx  .xls  .csv
4

Name the datasource and save

Give the datasource a recognizable name and click Save. The file’s sheets become queryable tables.
Each sheet in an .xlsx file becomes a separate table. For .csv files, a single table named Sheet1 is created. Table names are auto-generated with a short unique suffix to avoid collisions.

​
Syncing schema changes

If your database schema changes after initial setup — new columns are added, tables are renamed — you can resync the field list for a table without removing and recreating the datasource:
  1. Open the datasource and select the table.
  2. Click Sync fields.
  3. SQLBot fetches the latest column list from the live database and reconciles it with what it has stored. New fields are added; removed fields are marked inactive.

​
Managing existing datasources

From the datasource list you can:
  • Edit a datasource to update connection credentials or rename it.
  • Test an existing connection at any time to verify it is still reachable.
  • Delete a datasource, which also removes all associated table and field metadata. Any chat sessions that referenced the datasource will no longer be able to regenerate queries.
Deleting a datasource is permanent. All table selections, field descriptions, terminology entries, and SQL training examples linked to it will be lost.

Build docs developers (and LLMs) love

Get started for freeTalk to us