Dataset modeling and GenAI prompt management

Dataset modeling and in‑session GenAI prompts enable you to transform LN data directly into structured inputs for analysis and GenAI.

Dataset modeling allows you to define which LN data (tables and fields) must be extracted and how that data must be structured. GenAI prompts are then created as pre-designed instructions that users can run directly from an LN session, using selected records as the input context. This provides in-context insights, because users can run prompts where they work in LN without having to switch tools.

Shared datasets and prompts promote consistency and standardization across users and teams, while clear ownership and lifecycle control support governance. Reusing predefined structures and prompts reduces manual effort, improves efficiency, and ensures reliable and repeatable results.

Prerequisites and enablement

Feature enablement
The GenAI Embedded Experience tenant feature must be toggled on in CloudSuite Self‑Service Portal (CSSP). If this feature is not enabled, GenAI prompts are not available in LN sessions.
You must have the Admin role in CSSP to enable features. By default, your company's administrators listed in Infor Concierge have this role in CSSP.
Access by role
Users require access to the relevant LN sessions and permission to run GenAI prompts.
Content owners, such as power users or a center of excellence, create and maintain datasets and prompts to ensure consistency and governance.
Administrative or IT roles typically manage the available Large Language Models (LLMs) and support security and access configuration.

Procedure

  1. Enable the GenAI Embedded Experience feature for the tenant.
    1. Navigate to Infor Concierge > Navigation menu > Applications > CloudSuite Self-Service Portal.
    2. In CloudSuite Self-Service Portal, select Features from the side menu.
    3. From the Feature Type list, select GenAI.
    4. From the Tenant list, select the tenant for which to enable the GenAI features.
    5. Under GenAI Embedded Experience, locate the rows for the GENAI and LN applications. Enable the features by selecting the check box.
      Note: Enabling GENAI activates the LLM Passthrough APIs and Prompt Catalog functionality. These are required for the GenAI Embedded Experience to operate correctly.
    6. Click Save.
  2. Create and model the dataset.
    1. In the session, click New to initiate a new dataset configuration and start the Dataset Configuration screen.
    2. Select a Main table that aligns with the target business session and set the Selection Index to match the session index.
    3. In the , select the fields required for meaningful output. Include descriptive fields from referred tables where appropriate.
    4. Define dataset properties such as ordering, collection and row names to ensure a clear XML structure.
    5. Optionally, add a dataset filter. If the Input field is set to Yes in the Dataset Filter (ttadv9140m000) session, the filter value can be provided through a GenAI prompt parameter, allowing the dataset to be reused across multiple prompts with different filtering behavior.
  3. Verify prompt readiness for the dataset.

    From the Datasets (ttrpi5100m000) session, start the session and verify that a parameter of type Query Filter was created from the dataset filter.

  4. Refresh available Large Language Models (LLMs).
    1. In the session, click Refresh to load the available models and versions.
    2. Optionally, review available versions in the GenAI LLM Versions (ttrpi5561m000) session. The Is default indicator is informational and determined by the GenAI service.
  5. Verify the output formatter.

    Start the session and verify that an output format named XML exists and is linked to library ttrpidlloutxml, which is the standard-delivered XML formatter.

    If a different output than XML is required, create a custom formatter library (typically in the tx package) and register it in the Dataset Output Formats (ttrpi5110m000) session. KB3669822 on the Infor Customer Portal provides the source code of the standard XML formatter, which can be used as a reference for building custom output formats, such as JSON or CSV.

    As an alternative to separate formatter libraries, you can centralize formatting through the Dataset.CustomOutputFormat process extension. At runtime, the extension receives the requested format through i.dataset.format, as specified in the prompt placeholder (for example, #data,JSON#).

  6. Create the GenAI prompt.

    In the session, create a prompt.

    1. Assign a Label with the Context of Label set to Form Command (required to make the prompt available in sessions).
    2. Link a dataset to the prompt.
    3. Select a Usage Type to control intent and icon display.
    4. Maintain the Instruction Text (user‑facing) and Prompt Text (LLM instruction). Include a dataset placeholder using the syntax #data,[formatter],[optional root node name]#. For example: #data,XML#.

      If the formatter name referenced in the placeholder does not exist in the Dataset Output Formats (ttrpi5110m000) session, LN routes formatting through the Dataset.CustomOutputFormat process extension.

    5. If the dataset contains input‑enabled filters, click Parameters to set values in the GenAI Prompt Parameters (ttrpi5151m000) session. Select Refresh from Dataset if parameters are not visible.
  7. Publish the prompt to a session.

    In the session, link the prompt to the target business session.

    • Specify the Session , for example, Business Partners (tccom4500m000).
    • Set Active to Yes.
    • Select a value for Authorization , for example, Display.
    • Select a value for When Enabled , for example, Records Selected.
  8. Run and validate.
    1. Open the target business session, for example Business Partners (tccom4500m000). Select one or more records, and run the GenAI Form Command.
    2. Review the response and copy or download the results as needed.

    After modifying the dataset, close and reopen the session to ensure cached, session‑linked configuration is refreshed.

Moving configuration data between tenants

Prompts and datasets can be transferred between tenants using LN Configuration Data Management or CSSP Local Packages. Use the configuration object set ttPromptsDatasets in the Configuration Object Sets (ttdcm0550m000) session. If this object set is not available, run Import default LN Configuration Data Definition to load the standard DCM objects and sets. This object set groups the required configuration objects— datasets, dataset formats, prompts, and prompts-by-session links —allowing content to be promoted cleanly across environments.

For more information, see KB2252363 on the Infor Customer Portal and the Infor Configuration Management Reference Guide for Local Packages – LN in the Infor LN Enterprise Server Library (Cloud) on docs.infor.com.