About the permanentid field

Leverage Machine Learning Coveo Relevance Cloud Developer System Administrator Product Documentation

The permanentid field contains a value that uniquely and permanently identifies each item with respect to its original repository. The field value remains the same, even for an item that’s indexed more than once with different sources. This field allows Coveo Machine Learning (Coveo ML) models to learn user behavior on stable item IDs.

Prior to the addition of the permanentid field, Coveo JavaScript Search Framework pages and Coveo ML models used the urihash field by default to identify index items.

permanentid field value

The method to get the permanentid field value may be different depending on the repository type:

  • For most standard sources, the permanentid field is:

    • Based on the item URI, because it’s an appropriate unique and permanent identifier.

    • A 60 hexadecimal character hash of the item URI (to optimize index performances).

  • For other repository types, such as Google Drive and YouTube, where one item can have more than one URI, the permanentid field is typically a hash of a string containing:

    (repository type identification) + (unique separator) + (source dependent unique item identifier)
    Example

    In a Box source, the item permanentid field value is a hash of the string:

    "https://www.box.com/" + "@@@" + file_id

    where file_id is the Box unique identifier for each item.

  • For Salesforce sources, the Salesforce item ID is used to set the permanentid for Coveo for Salesforce editions using a Coveo index. Coveo for Salesforce editions using a Salesforce index also pass the Salesforce item ID as the unique ID for all usage analytics events.

    Consequently, if you migrate from a Coveo for Salesforce edition using a Salesforce index to one using a Coveo index, the item ID remains the same, keeping the Coveo ML learned behavior history for all Salesforce items.

    Example

    The URI for a Salesforce case is:

    https://na26.salesforce.com/5003200004BSRT6BAP

    The permanentid value will be: 5003200004BSRT6BAP

Taking advantage of the permanentid field

The permanentid field usage should be mostly transparent, but you may need to perform some tasks to fully take advantage of the custom aspects benefits.

  • Standard source types

    The introduction and usage of the permanentid field is meant to be mostly transparent for standard source types:

    • All standard sources automatically include the permanentid metadata and field mapping.

    • Since April 2017, a JavaScript Search Framework search page automatically detects the existence of the permanentid field for each clicked item. It sends this field and its value with each usage analytics event, in the contentIdKey and contentIdValue metadata fields respectively. The page always sends the urihash field and its value.

    • The Coveo ML models update take the contentIdKey field and the contentIdValue metadata passed in usage analytics events to identify each item to learn from, using the urihash as a fallback. The transition from the urihash to the permanentid fields is therefore automatic. Within an index or even a source, items may be identified using either fields.

      The models can map older usage analytics events, which contain only the urihash field, to newer ones with both the urihash and permanentid fields. This allows each item’s click history to be preserved through the transition.

      Following the transition from the urihash to the permanentid fields, your end users may experience a minor and temporary degradation of Coveo ML Automatic Relevance Tuning (ART) and Content Recommendations (CR) model performance, but only if you push custom usage analytic click events that don’t include both the urihash and the permanentid.

      This is because items whose unique identifier suddenly changes will appear as new items to Coveo ML models. With time, however, new usage analytics events on these items will accumulate, rebuilding their usage history and allowing Coveo ML models to learn from them again.

  • Custom sources (Push API)

    For custom sources populated through the Push API, when the URI isn’t a unique identifier and you want to allow Coveo ML models to learn usage of pushed items, you can:

    1. Push metadata that uniquely and permanently identifies each pushed item.

      The content of the metadata can be anything such as a URI or a GUID, as long as it’s unique and never changes.

      Using an alphanumeric string of at most 60 characters (without spaces) optimizes index performance. However, you can use any value you want. Using a value that isn’t hashed can help with troubleshooting.

    2. Map this metadata to the permanentid field (see Manage source mappings).

About the urihash field

Prior to the addition of the permanentid field, the urihash field was used to identify index items.

The value of the urihash field for a given item is based on the item’s URI value. The item’s URI is hashed to create a standardized urihash value.

For some sources, such as a repository that allows to share items, an item could have more than one URI, and therefore more than one urihash field value. This can cause Coveo ML models to interpret these values as distinct items when they actually refer to the same item.

Example

In a Box source, the item URI include the Box user ID, so a shared file or folder has many URIs. The permanentid field value is based on the Box file_id value, which remains the same even when an item is shared and accessible with various paths by various users.