Skip to main content
When you upload a file to the Agent chat, you can choose one of two options: The remainder of this page describes Process as Table uploads. When you choose Process as Table, Prophecy uploads the file to the SQL warehouse configured in your fabric and converts it into a structured table that the Agent can read and use in transformations.
Each file can be up to 30 MB, and the total size of all files in a single upload can be up to 100 MB.
Uploaded files are scanned for safety before they are processed. Prophecy rejects uploads if they contain disallowed file types, executable content, or archives that exceed size, nesting, or entry-count limits. Rejected uploads show an error message describing the reason.
You can upload the following file formats:
  • CSV
  • Excel (XLS, XLSX)
  • JSON
  • Parquet
  • XML

​
Upload a file

1

Start the upload

Click the paperclip icon in the chat footer or drag a file onto the pipeline canvas. When prompted, choose Process as Table to continue with the workflow described on this page.
2

Select your file

Choose the file you want to upload from your local filesystem and drag it into chat. The upload dialog opens automatically.
3

Configure storage location

Determine where the file should be stored in the SQL warehouse:
  • Choose the database.
  • Choose the schema.
  • Choose the table name. You can select an existing table or create a new one.
Click Next to continue.
If you select an existing table, Prophecy deletes and recreates the table with your uploaded file data.
4

Review and update schema

Review the inferred schema (column names and data types):
  • Keep the auto-detected schema, or modify column names and data types as needed.
  • Select any table properties you want to configure.
  • Click Next to proceed.
Excel files (XLS, XLSX) expose additional options at this step:After you configure these options, click Infer Schema to populate the column list. Column names, types (for example, String or Bigint), and other metadata are editable.You can also mark multiple regions of an Excel workbook as separate tables. Each marked table becomes its own output, and you can edit its schema, set drift-handling behavior, or unmark it independently of the others.
5

Preview and complete

Review the table preview to verify the data structure and content. If the preview looks correct, click Done to complete the upload.
For Excel files, the preview grid loads additional rows and columns as you scroll, rather than fetching the entire sheet at once. Merged cells and blank rows in the source sheet are handled automatically.
The file is converted to a table in your SQL warehouse.

​
Make the table available to the Agent

For the Agent to discover the table, you need to add it to the knowledge graph. After uploading a file, the knowledge graph indexer will automatically index the table. However, you can also manually reindex the knowledge graph, in case something goes wrong.
  1. Open the Environment tab in the left sidebar.
  2. Below your connections, locate the Missing Tables? callout.
  3. Click Refresh to trigger the knowledge graph indexer.
Once indexed, the Agent can reference the table using @ mentions or discover it through natural language queries. The Agent will use the Table gem to read the file into your pipeline.