The “Extract data from text” step allows you to analyze the entered text without having to attach a file. Based on the text, AI recognizes key information and saves it to the specified tables according to the established data schema.
Where to use the action:
Workflow
Parameters:
- Data source name (type: text): the data source ID previously entered in the external data source configuration.
- Text (type: text): the address of the cell containing the text from which the AI is to read and extract data.
- Table for identifiers (type: text): address of the column where the IDs of the processed files (GUID) will be saved.
- Prompt preview address (type: text): address of the cell indicating the location from which the system retrieves the default prompt (set of instructions) used during document processing.
- Prompt override (type: text): cell address allowing you to enter your own prompt to replace the default one. This allows you to customize the AI's behavior for specific document types or specific business needs.
Example application reading data from text
In this version of the application, the user enters the content of the email, and the system, using AI, automatically extracts key information. Based on this information, it determines the type of problem reported, its priority, and indicates the department to which the report should be forwarded, filling in the appropriate fields.
First, we create a list A1_EMAIL_PROCESSING with the following columns:
Data columns:
- ISSUE_TYPE (type: text) - title or type of reported issue
- DESCRIPTION (type: text) - summary of the email with the most important information. In the description of this column in gridview, include information that this is a summary of the email and list the most important elements of the message.
- PRIORITY (type: text) - priority of the report, automatically determined based on the content of the email
- DEPARTMENT (type: text) - the department to which the report should be sent, automatically determined based on the content of the email. The description of this column in the gridview should list the possible types of departments so that the AI system can correctly assign the report to the appropriate department
Technical columns:
- GUID (type: text)- identifier of the entire text document
- TEXT GUID (type: text)- unique identifier assigned to a specific text fragment
- LANGUAGE (type: text) - language of the email message
- INPUT_TOKENS_USED (type: number) - number of tokens used in processing the input
- OUTPUT_TOKENS_USED (type: number) - number of tokens generated at the output,
Next, we create a spreadsheet with one editable cell into which the user pastes, for example, the content of an email.
In the next step, we create a single-column table in which TEXT_GUID – a unique identifier for the received text – will be stored. We name the column, for example, TEXT_GUID.
Additionally, we create a spreadsheet with one editable cell in which the user will be able to see the generated prompt.
In order for the data to be correctly linked, in the A1_EMAIL_PROCESSING table, the TEXT_GUID column must be mapped using the lookup function with the table in which we save the returned TEXT_GUID. This will allow the system to know which processed file to assign the data to.
We add a workflow widget.
In the workflow, create an action step called “Extract data from text,” in which you should enter the appropriate parameters:
The end result:
0 Comments