Run and use Excel files as table data

You can convert Excel data into a structured database format to maximize search accuracy and consistency. You can view and edit tables within Excel sheets directly from the dashboard, and the completed schema is automatically converted into a database so that LLMs can analyze structured data more effectively. This minimizes the interpretation problems that occurred when handling large tabular data and provides more accurate and reliable answers.

Excel file upload and supported file formats

On the dashboard, you can upload Excel files in the following formats:

  • .csv

  • .xls

  • .xlsx

  • .xlsm

If the uploaded file is one of the four formats above, an icon that allows you to automatically enter the table edit screen will be displayed.

Caution 1. .docx Other formats besides Excel will not show the table edit screen.

📊 Structuring table data into structured data

To use table-form structured data in the Table Agent app, you must reconstruct the data using the 'Table Edit' feature.

How to edit

  1. Click the table edit icon

    • When clicked, you will move to an edit screen where you can set the table range, headers (column names), data types, and more.

  2. Configurable items

    • Table range: Specify the cell area that contains the data

    • Header (column name): Specify the name for each column

    • Data type: Specify the value format for each column (text, number, date, etc.)

Caution 1. The table edit icon is enabled only for members who have edit permissions on the document. Members without permission cannot use this feature.

📊 Composing table data

The table edit screen is divided into left and right areas, and the roles of each area are as follows.

Left area: uploaded file preview

  • A preview of the uploaded Excel fileis displayed.

  • You can view thecolumn and row information together.

    Caution 1. If there are many rows or columns, the preview display may be limited. In this case, the following message will appear at the top of the screen: "Refer to the original file to add table data" If this message appears, please construct the table data based on the original file.

Right area: table data editing

  • Based on the file information in the left area, you can directly edit the table data.

  • Editable items:

    • Specify table range

    • Edit header (column name)

    • Set data type (text, number, date, etc.)

Caution 1. Currently only one table on the first sheet is supported, so once one table is added, the + add button will be disabled.

Table configuration steps

  1. Name the table Enter a table name that matches the purpose of the data. (e.g., Employee Details or Employee HR Information, etc.)

  2. Specify table range Specify the cell range within the sheet to be used as data. For example, entering A1:N1001 will use data from column A to N and row 1 to 1001 as the table data.

  3. Enter table description Briefly describe the purpose or structure of the table. (e.g., salary, hire/termination dates, and job group information for 2025 hires) If necessary, you can also write input rules or data constraints for each column. However, this description is directly reflected in the model prompt, so keep only the core rules concise. Unnecessarily long sentences or redundant examples may confuse the model.

Set detailed content within the table Now set each header (column information) to be included in the table. Headers are key elements that define the data structure and will be referenced during searches and SQL queries.

How to add headers

  1. Enter header cell number Enter the row number that will be used as the header. (e.g., if the first row is the header, enter the corresponding alphabet and number such as A1) If the table range is A1 to D99, you must enter all four headers when specifying headers.

If hidden columns are included in the file

When a document is uploaded with hidden columns included, those columns will remain hidden in the preview screen and will automatically be excluded when converting to table data. In this case, because column B was uploaded as a hidden column, you can enter only A, C, and D in the header information. If you want to include hidden columns in the table data, refer to the original file and manually enter the headers for the hidden columns. The table will be restructured based on the headers you enter, and if you manually enter them, the hidden column data will also be included in the table data.

  1. Enter header name Enter the header name to use for each column. ※ It does not have to exactly match the original Excel column name, and it is recommended to name it in a way that is useful for search or filtering.

  2. Specify data type Select the type of data that will go into each column. Data type directly affects search accuracy and SQL query results, so it is important to specify it accurately according to the actual data characteristics. For example, if you set the type to DATETIME but the actual value is a STRING, it may be stored as null, so be careful.

Descriptions and examples by data type

STRING

General text data composed of characters. Even numbers should be set as strings if they are for display rather than calculation.

Employee name: Kim Suhyun, Employee ID: A-102

NUMBER

Integer or floating-point numeric data. Use for data that requires calculations such as sums or averages.

Salary: 120, Ratio: 0.85

DATETIME

Data that contains date or time information. Can be used as a period filter in SQL searches.

Hire date: 2025-10-16, Termination date: 2025-09-01 14:32:00

BOOLEAN

Data in True/False form. Used to represent selection or status values.

Active status: TRUE, Notification completed: FALSE

Briefly write the meaning or intended use of the header's data. This description is used directly by the model to understand the column's meaning, so enter only the short and clear essentials.

Caution 1. If column values include currency units, the unit information may be lost when structuring the data. If necessary, include the currency unit in the header description.

If you want to handle merged cells

How to handle merged cells Merged cells will be automatically unmerged when converting to table data, and the original value will be duplicated into all cells of the merged area.

Therefore, enter as follows.

Example:

  • If B2:C2 are merged and have the value "Department" → enter B1 = "Department", C1 = "Department2", etc.

  • Since cell names must be unique, it is recommended to add a symbol like _2.

If the header position you want to save as data is different

If the rows of cells for the table you want to configure do not align, assign the numbers for the area you want to designate as the header. If you enter it as shown above, actual empty values will be treated as NULL and remain intact without breaking the data. Execute as table data

When all header information is entered, click the save button.

After all information is entered, click the 'Execute as table data' button.

You can now configure the app to use that table data for answers.

If saving as table data succeeds and it can be used in the app, it will be shown in green.

If you only save your entries and do not execute as table data (so it cannot be used in the app), it will be shown in orange.

Configure the app

Note 1. The Table Agent app is included by default in the app you are creating, so it is recommended to copy or edit that app when using it

  1. Click the app edit button.

  2. In the LLM input node, select the Excel file you just configured within the document range. (You must perform this step to use it as an app. If you publish and run the app you are creating without specifying the document, an error will occur because there will be no table data to reference.)

Caution 1. The variable must be set to the system variable EXCEL_DOCUMENTS. Do not change this to any other variable.

2. The recommended model is OPENAI CPT-4o.

  1. Publish the app.

Check operation - test the app

You can now search for information and generate answers based on the configured table data.

Click View Answer Process to see the detailed process of how the SQL query was generated.

※ Notes when managing Excel files in the dashboard Documents tab

Caution 1. The document update button provided as a basic function in Alli can only be used for Excel files that have not been executed as table data. Once an Excel file is configured as table data, the update button will not be displayed. 2. The table edit icon is shown for all Excel files. However, if you do not configure it as a table in the actual edit screen, the Excel file will be parsed into markdown (plain text) and can be used freely as a document answer source as before. If you want to use one document in both table data and markdown formats simultaneously, upload two copies with different document names and configure one of them as table data.

Last updated