AI Search - Administrator Guide

This guide covers configuring and monitoring AI Search indexing inside Innovator. It assumes AI Search is already enabled. See Get Started with AI Search for enablement steps, and Installation of AI Search or Installation of AI Search on SaaS for deployment-specific installation steps.

The AI Search Configurator is available only to Innovator administrators.
On the dedicated AI Search page, administrators see a gear icon in the top-right corner that opens the configurator.
Non-administrators do not see this icon and cannot access the admin panel.
To return to search after configuring, use the Back to Search arrow in the Indexing Scope header. This returns to the AI Search page and preserves the current URL context (such as an active search query).

Open the AI Search page in Innovator and select the gear icon in the top right corner to open the AI Search Configurator.

The Select an ItemType... field is a searchable selector.
Type any part of an ItemType name to filter the list, then select the ItemType and click Add.
The list is sourced live from your connected Innovator instance and includes all ItemTypes registered in your database, including PLM data types, administrative types, and system types alike. Use the search field to find the data types relevant to your use case.

The list includes both regular ItemTypes and Relationship ItemTypes.
Relationship ItemTypes are marked with a Rel badge and can be added in two ways:

• Directly to scope: add the Relationship ItemType itself as an indexable type.
• As file properties on another ItemType: configure relationship-sourced file properties under a regular ItemType in the File Properties section (see below).

Each added ItemType appears as a collapsible row.
Select the row header to expand it and configure settings.
ItemType detail loads the first time you expand a row; a loading indicator is shown while detail is fetched. If detail fails to load, an inline error is shown with a Retry option.
Each row has a delete icon to remove the ItemType from scope. If a row fails to render, an error boundary offers to reset by removing that ItemType.

Expand an ItemType row and use the Select a property... field to add properties one at a time.
The property selector is also searchable; type any part of a property name to filter the list.
Each added property appears in a table showing its field name and type, with a drag handle for reordering and a delete icon for removal.
Removing a property from the table disables it for indexing; it is not permanently deleted and can be re-added later from the property selector.

At least one property, file-indexable property, or file relationship is required per ItemType. Attempting to apply a configuration with none of these configured for a given ItemType returns the following error:

Apply failed
IndexConfiguration.IndexableItemTypes[N].IndexableProperties:
  - At least one indexable property is required

Each ItemType supports an optional title template and subtitle template, which control how that ItemType’s results are presented on result cards.

The File Handling Settings area lets an administrator select which file types attached to an ItemType’s items should be included in search.

The Select file types... field is searchable.
Type any part of a file format name to filter the list.
The available file types are sourced live from your Innovator instance’s FileType table and reflect whatever formats are registered in your database.
A standard Innovator installation includes common engineering formats such as AutoCAD, CATIA, SolidWorks, STEP/IGES, as well as office formats (PDF, DOCX, XLSX) and text formats. Custom file types registered in your Innovator instance also appear.

Below the property table, each ItemType may show a File Properties section when file-related properties are available.
This section is separate from File Handling Settings (which selects file types).
File Properties controls which file-related fields are used for text extraction, vectorization, and search.
Each file property row shows the field name, data type, and an enable/disable switch. Properties sourced from a relationship show a Rel badge with the relationship name.
Enable at least one file property when file content indexing is required for an ItemType that has file handling configured.

Administrators can export the current configuration as a JSON schema file, edit it, and re-import it. The action bar at the bottom of the page shows: Import Schema JSON, Export Schema JSON, and Verify on the left; Cancel and Apply on the right.

Import Schema JSON loads a JSON file into the configurator editor and shows a success notification (Configuration imported).
The imported configuration is not saved until you select Apply.

• Cancel reverts all unsaved changes, including an import, back to the last configuration saved on the server.
• Export Schema JSON downloads the current editor state (including unsaved changes).
• Verify validates the current editor state without saving.

Verify checks that every ItemType referenced in the configuration exists in the current Innovator instance and that the property constraint described above is satisfied.

Verify does not save the configuration. Always select Apply after a successful Verify to persist changes.

The configurator tracks unsaved edits. While changes are pending:

• Apply is enabled; Cancel is enabled.
• With no unsaved changes, both Apply and Cancel are disabled.
• While the configurator is loading or a save/verify is in progress, action bar buttons are disabled.

If you navigate away using Back to Search with unsaved changes, a confirmation dialog appears:

Unsaved Changes
You have unsaved changes. Are you sure you want to leave? All unsaved changes will be lost.

Select Leave Page to discard changes, or Cancel to stay. Closing the browser tab with unsaved changes also triggers a warning.

Selecting Apply saves the configuration to the server and starts the apply job in the background. The configurator returns to an interactive state promptly; progress is visible in the Indexing Status section while the job runs.

While an apply is in progress, the configurator blocks additional configuration changes. Apply and Cancel are disabled until the current apply completes.

When you have unsaved changes, a re-indexing warning appears above the action bar:

Re-indexing required
Changing the indexed scope will require re-indexing.
This operation may take time and temporarily reduce search freshness.

The warning indicates that the change triggers indexing work. The scope of that work depends on what changed. See What Apply Does for Each Change Type below.

If the AI Search package is not installed on the Aras Innovator instance, the Indexing Scope section shows:

AI Search Not Installed 
This Innovator instance does not have the AI Search package installed. 
Please install it on the instance to configure indexing settings.

Apply detects the difference between the previously saved configuration and the new one and performs only the work that difference requires.

Two invariants apply across all operations, including Apply and Reindex All:

• No operation deletes file embeddings. Files already parsed and embedded are preserved regardless of what changes. Existing file embeddings are reused without re-parsing when they come back into scope.
• ItemTypes not affected by a change remain fully searchable throughout any apply. Search returns only embeddings within the active scope; embeddings that are preserved in storage but outside the current scope are not returned.

Apply and Reindex All is the secondary action on the Apply split button. Select the dropdown arrow to the right of Apply to access it.
This action saves the current configuration and triggers a full rebuild of all item embeddings. All item embeddings for the active configuration are deleted before rebuilding. Search results for affected ItemTypes are incomplete during the rebuild.

Use Apply and Reindex All when item embeddings need to be rebuilt from scratch: for example, after an embedding model change or to recover from a corrupted item index state. For routine configuration changes, the standard Apply handles re-indexing automatically and with less disruption.

On Aras Innovator instances where file properties are enabled for ItemTypes with a large number of records, adding a non-clustered index to the Innovator database for each configured file property column improves search query time.

This applies when both of the following are true:

• One or more ItemTypes in the indexing scope have file properties enabled in the File Properties section of the AI Search Configurator.
• The source table for that ItemType in the Innovator database contains a large number of records.

For each qualifying ItemType, create a non-clustered index on each configured file property column. Run the following on the SQL Server instance that hosts the Innovator database, substituting the actual ItemType name and file property name:

CREATE INDEX [IX_{ItemTypeName}_{filePropertyName}]  
	ON [innovator].[{ItemTypeName}] ([{filePropertyName}]);

Example for a Part ItemType with a drawing_file file property:

CREATE INDEX [IX_Part_drawing_file]  
	ON [innovator].[Part] ([drawing_file]);

Repeat for each file property configured for that ItemType. If file properties are configured on a Relationship ItemType, create the index on the relationship table instead:

CREATE INDEX [IX_{RelationshipName}_{filePropertyName}]  
	ON [innovator].[{RelationshipName}] ([{filePropertyName}]);

Run the following after index creation to confirm the index is in place, substituting the actual table name and column name:

SELECT
    i.name AS IndexName,
    c.name AS ColumnName,
    i.type_desc AS IndexType
FROM sys.indexes i
INNER JOIN sys.index_columns ic
    ON i.object_id = ic.object_id AND i.index_id = ic.index_id
INNER JOIN sys.columns c
    ON ic.object_id = c.object_id AND ic.column_id = c.column_id
INNER JOIN sys.tables t ON i.object_id = t.object_id
INNER JOIN sys.schemas s ON t.schema_id = s.schema_id
WHERE s.name = N'innovator'
  AND t.name = N'Part'         	-- substitute ItemType table name
  AND c.name = N'drawing_file'; -- substitute file property column name

A row in the result confirms the index is in place. If the query returns no rows, the index does not exist and must be created.

The Indexing Status section, below Indexing Scope on the same page, shows the live state of indexing for the active configuration.
The section auto-refreshes every 30 seconds and displays a Last updated timestamp. Select the refresh control next to the timestamp to refresh data.

If a refresh fails but previously loaded data is available, a warning is shown: Could not refresh indexing status. Showing last known data.
If the initial load fails entirely, an error is shown with Retry.

Indexing Status reflects the currently saved configuration on the server. Unsaved changes in the Indexing Scope editor are not reflected here until you select Apply.

A row of status indicators at the top right of the Indexing Status section shows the health of the six backend services that AI Search depends on. Each indicator shows a colored dot and service name:

These indicators are read-only monitoring. They do not trigger repairs or configuration changes.
If health data cannot be loaded, Health unavailable is shown with a Retry link.

Below the health indicators, the Indexing overview table shows one row per ItemType in the active scope, plus a Files row when file handling is enabled for at least one ItemType. Each row shows:

• Total: the discovered item count, or an em dash if the total is not yet known.
• Indexed: the number of items successfully indexed.
• Failed: 0 as plain text when there are no failures, or a clickable chip labeled with the error count (for example, 3 Errors) when items failed.

Select a Failed chip to open the Failed Items dialog for that row. See Handling errors and failed items below.

The initial indexer details table is hidden by default. Select Show Initial Indexer details below the Indexing Status panel to expand it. Select Hide Initial Indexer details to collapse it.

The table is organized into two sections:

• ITEM INDEXERS: one row per ItemType in the active scope. A total item count is shown at the bottom of this section.
• FILE INDEXER: a single row covering all file indexing activity. This row appears when file handling is enabled for at least one ItemType.

Each row in the indexer table shows one of five states:

Each row also shows the Started and Completed timestamps for the most recent run, a Progress counter, and a progress bar.
Progress messages include:

For the File Indexer, an info icon may appear when some files reused existing embeddings and did not require reprocessing.
A total item count is shown at the bottom of the table. The table supports pagination; use the rows-per-page selector and navigation controls to view all indexers when many ItemTypes are configured.
The status of the most recently completed run persists between configuration changes and page reloads, so administrators see the last known outcome rather than a blank state.

If no ItemType has been added to the indexing scope, this section shows:

No Indexers Configured
No item types have been configured for indexing yet.

If the AI Search package is not installed, this section shows:

AI Search Not Installed
Install the AI Search package to view indexing status.

A Failed or Partially Completed status on an indexer row indicates that the corresponding indexing run did not complete without errors.

Select the Failed chip in the Indexing overview table (see above) to open the Failed Items dialog for that row. The dialog title reads {ItemType or Files} — Failed Items, and it lists each failed item or file with its Name and failure Reason.

For a single row, select Retry to re-queue that item or file for indexing immediately, or Resolve to dismiss it from the failed items list without indexing it. Select multiple rows with their checkboxes to enable bulk actions in the dialog’s action bar: Retry N selected and Resolve N selected.

Resolve requires confirmation. The Resolve Failed Items dialog reads: “N item(s) will be dismissed from the failed items list without being indexed. This does not delete the underlying item -- it stops the platform from retrying it automatically. You can still retry it later if it reappears.” For files, an additional sentence clarifies that resolving dismisses only file-indexing reconciliation, and that a separate File ItemType reconciliation failure, if present, must be resolved from the item-type view. Select Resolve to confirm, or Close to cancel. Retry has no confirmation step and queues immediately.
Select Close to dismiss the Failed Items dialog. It also closes automatically once its failed items list becomes empty.
All ItemType versions are indexed.

If issues persist after following the steps above, contact Aras Technical Support via the Subscriber Portal with your tenant ID, the ItemType configuration in question, and any error details copied from the Failed Items dialog.