AI Assistant Operations

Managing AI Assistants

Open the Administration Tool

  

Activate the Utilities tab.

  

Choose the option below.

  

The AI assistants are listed.

  

. . .

  

The options are listed below.

Icon	Description
	View the complete AI Assistants settings.
	Copy the complete AI Assistants settings.
	List the cached embeddings for the selected AI assistant.
	Download the cached entries for the selected AI assistant.
	Import cached entries for the selected AI assistant.
	Delete all of the cached entries for the selected AI assistant.

Embedding Value Caching

Embeddings for any particular string are effectively constant for a given service and model combination. As such you can avoid AI service costs by caching the embeddings. There are 2 global settings associated with the caching of embedding values.

• defaultCacheEmbeddings for no alias named entry and

• maxCacheTextEmbeddingLength.

The global defaultCacheEmbeddings flag must be true to even consider caching of the fetched binary and text resulting embeddings.

Individual entries can then have these settings:

• cacheEntries flag,

• allowFetchUrlEmbedding,

• allowFetchUrlEmbeddingBinaryCaching is used when the content type contains and

• ‘image’ or ‘binary’.

The allowFetchUrlEmbeddingTextCaching setting is used when the content type contains ‘text’, ‘json’, or ‘html’.

For other content types the setting matches the text one.

After changing the model associated with an alias then the Qarbine cache, if any, must also be cleared. If this is not done then the following error may occur.

  

Choose an AI Assistant

  

Click the delete image

  

The request’s status is shown.