Introduction Fabasoft Approve

Fabasoft Approve offers powerful document and quality management in technical projects and supports companies in the efficient management of technical documentation, quality-relevant documents and processes, as well as project-related correspondence - traceable, structured and compliant.

Fabasoft Approve is divided into two solutions with different scopes of services, Fabasoft Approve DMS and Fabasoft Approve QMS.

Fabasoft Approve DMS

Fabasoft Approve DMS contains the following application area:

• Correspondence Management (see chapter “Application Area: Correspondence Management”)
• Technical Records Management (see chapter “Application Area: Technical Records Management”)

Fabasoft Approve QMS

Fabasoft Approve QMS contains the following application areas:

• Correspondence Management (see chapter “Application Area: Correspondence Management”)
• Quality Management (see chapter “Application Area: Quality Management”)
• Technical Records Management (see chapter “Application Area: Technical Records Management”)

Integration of artificial intelligence with Mindbreeze AI

Fabasoft Approve integrates AI functionalities to support users and reduce the workload. These AI functionalities are based on the Fabasphere AI Core and require the appropriate licensing. Each AI-supported functionality is marked with “AI function”.

Ask Questions (AI Feature)

If documents have been indexed in the AI index, it is possible to chat with them. The “Ask Questions About” action is available in the context menu of an object for this purpose. Clicking on “Ask Questions About” opens a submenu where you can select the scope you would like to ask questions about.

Generate Summary (AI Feature)

In the configuration or in the places where templates can be defined, it is possible to create templates for summaries. For selected objects from the respective application areas, it is then possible to generate an AI-supported summary by selecting the context menu command “Generate Summary” and selecting one of the available templates. The AI then generates a summary for the object.

If the selected object is a document, the document is used as the basis for the summary. If it is a composite object, the documents directly associated with it are used as the basis for the summary.

Document Check (AI Feature)

In the configuration or in the places where templates can be defined, it is possible to create templates for document checks, which consist of a list of document check entries. For selected objects from the respective application areas, it is then possible to perform an AI-supported document check by selecting the context menu command “Perform Document Check” and selecting one of the available templates. The AI then evaluates each of the document check entries.

If the selected object is a document, the document is used as the basis for the document check. If it is a composite object, the documents directly associated with it are used as the basis for the document check.

Generating and optimizing translations (AI Feature)

In the configuration settings, or in places where templates can be defined, it is possible to create templates for translation folders consisting of a list of translators. Similarly, within a translator, it is possible to define a translation base, the source language, the target languages and glossary objects. Glossary classes can also be used to help the AI ensure consistency in line with app conventions.

For the selected translation base, the AI then generates translations for the respective selected languages if they do not yet exist; if they do exist, they are optimized.

Answer Form (AI Feature)

Fabasoft Approve allows you to fill out forms using AI. To do this, the form in question must be assigned the category “AI-supported form.” You must then select “Base Form for AI-Supported Form” as the base form. The following fields are then available in the properties:

• Only Generate Answers for
  Allows you to limit the fields to be filled in by the AI to the selected fields.
• Context Objects
  Allows you to specify how to identify the context documents that are used as the knowledge base for populating the form fields.
• Expression for Retrieving Context Objects
  If the value “Expression” was selected in the Context Objects field, the corresponding expression must be defined in this field.
• Use LLM Only
  Uses the LLM directly to populate the form fields without first having Mindbreeze AI Core identify the relevant text passages in the context documents. This requires that the context documents be passed to the LLM in blocks.
• Consolidate Answers
  If this field is enabled, the responses generated by the LLM for the various blocks of the context documents will be consolidated by the LLM to provide only the best responses.

For an object with the corresponding category, the “Answer Form” menu option will then appear in the context menu to initiate the AI-based response to the form.

The following default fields (which can be assigned to context documents via a category) allow you to configure additional settings for the context documents:

• Priority
  In this field, you can set the priority of the context document. Responses from high-priority context documents are given preference during the consolidation of responses.
• Type of Analysis
  Here, you can specify whether the context document is passed to the LLM as text, an image, or both text and an image. For complex tables, it is recommended to pass the document as an image.
• Pages for AI Metadata Extraction
  In this field, you can limit the page range of the context document that is passed to the LLM. This allows you to exclude, for example, the table of contents or sections containing examples, and send only the pages relevant to filling out the form fields.

Metadata Extraction (AI Feature)

Fabasoft Approve supports the extraction of metadata from documents using AI.

This chapter provides a brief overview of the definition and use of metadata extractions. For further details, refer to chapter “Providing Metadata Extractions: new window” in the administration help.

Defining Metadata Extractions and Extraction Definitions

Metadata extractions and extraction definitions are managed in the configuration in the “AI Elements” list.

Defining Metadata Extractions

A metadata extraction defines the objects to which it can be applied, describes the general framework, and must contain one or more extraction definitions that specify which properties are to be extracted using AI.

For metadata extraction, select the object class or category in the Applicable to property for which the metadata extraction can be used.

In the General Prompt for Metadata Extraction property on the “Advanced” tab, you can enter detailed instructions to provide the AI with additional context for the extraction task.

In the Placeholders property, you can define placeholders that can then be used in the General Prompt for Metadata Extraction property for metadata extraction, as well as in the Prompt for Extracting the Value property within the extraction definitions. When used, placeholders must be enclosed in double curly braces: “{{placeholder}}”.

Defining Extraction Definitions

To extract the value of a property using AI, an extraction definition must be created for that property and assigned to metadata extraction.

For each extraction definition, the target property must be selected in the Property for Extracting the Value field to specify the property for which the AI is to extract a value.

In the Prompt for Extracting the Value property, a prompt that is as specific as possible must be defined to instruct the AI on how to extract the desired value.

AI-Assisted Population of Extraction Definitions

Using the “Populate with AI” context menu, you can have the AI generate the Prompt for Extracting the Value, Synonyms, and Examples for the selected extraction definitions.

Releasing Metadata Extractions

Once the metadata extraction and its associated extraction definitions have been defined, they must be made available for use via the “Release for Usage” context menu option.

Preparation of Context Documents

Metadata extraction generally uses the documents directly subordinate to the target object as the knowledge base for the metadata extraction.

Using a category, the following properties can be assigned to the context documents to influence the metadata extraction:

• Priority
  This field allows you to prioritize context documents as “High,” “Medium,” or “Low.” Results from high-priority documents are weighted more heavily than results from low-priority documents when determining the best answer for an extraction definition. Documents classified as “Not Relevant” are ignored during metadata extraction.
• Type of Analysis
  Here, you can specify whether the context document should be passed to the AI as plain text or whether its pages should be converted into images and sent to the AI's vision model. Using the vision model results in a longer processing time but delivers better results if the document contains complex tables, charts, or graphics.
• Pages for AI Metadata Extraction
  Here, you can specify which pages or page sections should be included in the metadata extraction. It is recommended that you exclude all irrelevant sections (such as tables of contents, general examples, or indexes).

Using Metadata Extractions

If a metadata extraction is applicable to an object, the “Extract Metadata” option appears in the context menu for that object, allowing you to start the AI-based metadata extraction.

Once metadata extraction is complete, the properties of the target object are opened. If the AI was able to determine a value for a property, that field is already populated with the extracted value in the dialog, and the property is highlighted in color. Clicking “Next” saves the changes; clicking “Cancel” discards them.

Importing Metadata Extractions

Metadata extractions can be imported from a JSON file. To import metadata extractions from a JSON file, navigate to the AI Elements list in the configuration and select “Import Metadata Extractions”.

The following is an example JSON that can be used for importing metadata extractions.

Basic Information About the Fabasphere

Fabasoft Approve builds on the many functions of the Fabasphere, which are documented in the User Help Fabasphere AI Core: new window and the Administration Help Fabasphere AI Core: new window.