Tabular Data Tool

Tabular Data Tool handles precise questions from a CSV, XLSX workbook, or Google Sheet. Codeer can reach it through automatic Knowledge Base discovery or an explicitly selected source. It reads structured rows directly, so it is better suited than document search for filtering, sorting, counting, and calculating totals or averages.

Each explicit tool connects to one tabular source. Whether the source is discovered automatically or selected explicitly, a query uses one table, worksheet, or sheet tab at a time.

Prepare the data source

Add the source to Knowledge Base before configuring the tool:

Upload a CSV or XLSX file directly.
Import a Google Sheet from Google Drive.
Wait until processing finishes and the source is ready.
Open the source detail and review its tables, row counts, columns, inferred types, and warnings.

The source is mapped as follows:

Source	Tables available to the tool
CSV	The file becomes one table
XLSX	Each visible worksheet becomes one table
Google Sheets	Each visible sheet tab becomes one table

Use clear table and column names. They help the agent choose the correct table and fields when it prepares a query.

Workbook formulas

Codeer does not run spreadsheet formulas. For XLSX files, it reads the value saved in the workbook; a formula without a saved value is treated as empty and may produce a warning. Hidden worksheets and tabs are not available to the tool.

Choose automatic discovery or an explicit source

An agent with a configured Knowledge Base can automatically find and query a relevant CSV, XLSX worksheet, or Google Sheets tab. You do not need to add a separate Tabular Data Tool for every source.

Setup	Use it when	How it behaves
Automatic discovery	A general-purpose agent should choose the relevant table from its configured Knowledge Base	Configure `Knowledge Base` and do not add an explicit `Tabular Data Tool`
Explicit source	The source must stay fixed, or several sources have similar names or content	Add one `Tabular Data Tool` for each source the agent may use

With automatic discovery, the agent queries a single clear match directly. If several tables are plausible, it lists a small set of choices and asks the user to confirm before querying.

Publishing makes configured tabular data available to channel users

When you publish an agent, tabular sources included through its Knowledge Base or explicit Tabular Data Tool can be queried for users on the Web client, Widget, Browser Extension, Slack, LINE, Messenger, and Instagram. This includes exact rows, filtered results, and aggregates—not only document excerpts. Before publishing, include only sources approved for external users.

Explicit tools take priority

Once an agent has any explicit Tabular Data Tool, automatic discovery is disabled for that agent. It only uses the sources selected in its explicit tools and does not fall back to other tables in the Knowledge Base.

Sources added before automatic discovery

An older tabular source may need to be processed again before the agent can find it automatically. If it is not selected as expected, replace the CSV or XLSX, refresh the Google Sheet, or import it again, then retest the agent. Explicit Tabular Data Tool connections continue to work.

Add an explicit tool to an agent

Open the agent in Editor.
Under Tools, click Add Tool.
Choose Tabular Data Tool.
Select one ready CSV, XLSX, or Google Sheets source.
Write a narrow When to Use instruction.

For example:

Use this tool when the user asks for customer records, order totals, or grouped sales figures from the 2026 sales workbook. Use the Customers table for customer lookups and the Orders table for sales calculations.

If an agent needs more than one source, add a separate tool for each source and make their When to Use instructions distinct.

Test the expected questions

Use Live Test with questions that reflect real requests:

Which customer has account ID C-1042?
Show orders with an amount above 1,000, highest first.
What is the total order amount by region?
How many distinct customers placed an order?

Confirm that the answer uses the intended source and table, and verify totals against the original spreadsheet before publishing the agent.

Update the connected data

For CSV and XLSX, use the source's replace action and choose a file of the same format.
For Google Sheets, use Refresh to fetch the current version.
If processing fails, use Retry after correcting the source or access problem.

Replacing or refreshing keeps the existing agent tool connection. If a usable version already exists, the agent continues using that last good version while the update is processing. A failed update does not replace the last good version.

Limits and safety

Tabular Data Tool is read-only. It can inspect tables, select rows, filter, sort, limit results, and calculate supported counts, totals, averages, minimums, maximums, and grouped results.

It does not:

modify the source
run SQL, Python, macros, spreadsheet formulas, or arbitrary expressions
join multiple tables or sources in one query
return an entire large dataset as a bulk export

If a question requires combining tables, prepare a single table with the required fields before importing it, or ask separate questions and verify the results independently.

Common mistakes

Using document search for an exact calculation

Use Knowledge Base for relevant passages and source material. For exact rows, filters, sorting, or aggregates, let the Knowledge Base discover a relevant table automatically or configure an explicit Tabular Data Tool when the source must stay fixed.

Writing a vague `When to Use`

Name the business questions, source, and tables the tool should handle. Avoid instructions such as Use when needed.

Replacing with a different file format

CSV can only replace CSV, and XLSX can only replace XLSX. Add a new source when the format changes.

Expecting a cross-table join

One query targets one table. Reshape the source when a reliable answer requires fields from multiple tables.