N/llm Module

Note:

The content in this help topic pertains to SuiteScript 2.1.

The N/llm module supports generative artificial intelligence (AI) capabilities in SuiteScript. You can use this module to send requests to the large language models (LLMs) supported by NetSuite and receive responses to use in your scripts.

If you're new to using generative AI in SuiteScript, see SuiteScript 2.x Generative AI APIs. That topic contains essential information about this feature.

The following list summarizes the main features that are available using the N/llm module:

• Content generation – You can request generative AI content from a supported LLM using llm.generateText(options). You can provide a prompt that describes the content you want to generate, and the module sends the request to the Oracle Cloud Infrastructure (OCI) Generative AI service to generate a response.

• Prompt evaluation – If you use Prompt Studio to manage existing prompts in your NetSuite account, you can use llm.evaluatePrompt(options) to send a prompt from Prompt Studio to the LLM for evaluation. This method uses the information from the prompt definition in Prompt Studio (such as the model and model parameters), and it lets you provide values for any variables the prompt uses before sending it for evaluation. For more information about Prompt Studio, see Prompt Studio.

• Prompt and Text Enhance action management – By using the N/record module, you can create, update, and delete prompts and Text Enhance actions in your scripts. For more information, see Managing Prompts and Text Enhance Actions Using the N/llm Module.

• Retrieval-augmented generation (RAG) support – You can give source documents to the LLM when calling llm.generateText(options). The LLM uses information from the source documents to augment its response. The LLM also returns citations that identify which source documents it used. For an example of how to implement a RAG use case using the N/llm module, see Provide Source Documents When Calling the LLM.

• Embedding support: The llm.embed(options) method converts text to vector embeddings. Your SuiteScript applications can use vector embeddings for use cases such as semantic searches, recommender systems, text classification, or text clustering. For an example of how to generate and use embeddings, see Find Similar Items Using Embeddings. For more information about embedding models, refer to About the Embedding Models in Generative AI in the Oracle Cloud Infrastructure Documentation.

  Embedding methods have their own monthly free usage quota, separate from generate methods. If you use both in a month, you'll see two rows for the month, labeled by usage type. For more information, see View SuiteScript AI Usage Limit and Usage.

• Streaming support – Using the llm.generateTextStreamed(options) and llm.evaluatePromptStreamed(options) methods, your code gets content as the LLM generates it, instead of waiting to receive it all at the same time. For an example how to work with streamed content, see Receive a Partial Response from the LLM.

• Method aliases – You can use the following aliases in your code instead of the method names:

  • llm.chat(options) is an alias for llm.generateText(options).

  • llm.executePrompt(options) is an alias for llm.evaluatePrompt(options).

  • llm.chatStreamed(options) is an alias for llm.generateTextStreamed(options).

  • llm.executePromptStreamed(options) is an alias for llm.evaluatePromptStreamed(options).

  Promise versions are also available for these methods.

  When aliases are available for methods, you'll see them listed in the main table of the method's help topic. For an example, see llm.generateTextStreamed(options).

Using OCI Credentials to Obtain Additional Usage

By default, NetSuite provides a free monthly usage pool of requests for the N/llm module. Successful calls to generate methods (such as llm.generateText(options) and llm.evaluatePrompt(options)) and embed methods (such as llm.embed(options)) consume usage from this pool, and the pool is refreshed each month. Usage is tracked separately for generate methods and embed methods. You can track your current monthly usage on the AI Preferences page in NetSuite. For more information, see View SuiteScript AI Usage Limit and Usage.

Each SuiteApp installed in your account gets its own separate monthly usage pool for N/llm methods, and these SuiteApp pools are independent from the usage pool for your regular (non-SuiteApp) scripts. For example, if you install two SuiteApps, each with scripts that use N/llm methods, each SuiteApp draws from its own unique usage pool. This approach means you get twice the total SuiteApp usage (one pool per SuiteApp). Any other scripts outside of SuiteApps use a separate usage pool, and SuiteApp usage doesn't count against it. This setup ensures that SuiteApps can't use up all your monthly allocation and block your own scripts from calling N/llm methods.

If you want more monthly usage, you can provide the Oracle Cloud Infrastructure (OCI) credentials for an Oracle Cloud account that includes the OCI Generative AI service. When you provide these credentials, usage is drawn from the provided OCI account instead of the free usage pool. For more information, see Using Your Own OCI Configuration for SuiteScript Generative AI APIs.

Important:

As you work with this module, keep the following considerations in mind:

• Generative AI features, such as the N/llm module, use creativity in their responses. Make sure you validate the AI-generated responses for accuracy and quality. Oracle NetSuite isn't responsible or liable for the use or interpretation of AI-generated content.

• SuiteScript Generative AI APIs (N/llm module) are available only for accounts located in certain regions. For a list of these regions, see Generative AI Availability in NetSuite.

In This Help Topic

• N/llm Module Members

• ChatMessage Object Members

• Citation Object Members

• Document Object Members

• EmbedResponse Object Members

• Response Object Members

• StreamedResponse Object Members

• Usage Object Members

N/llm Module Members

Member Type	Name	Return Type / Value Type	Supported Script Types	Description
Object	llm.ChatMessage	Object	Server scripts	The chat message.
	llm.Citation	Object	Server scripts	A citation returned from the LLM.
	llm.Document	Object	Server scripts	A document to be used as source content when calling the LLM.
	llm.EmbedResponse	Object	Server scripts	The embeddings response returned from the LLM.
	llm.Response	Object	Server scripts	The response returned from the LLM.
	llm.StreamedResponse	Object	Server scripts	The streamed response returned from the LLM.
Method	llm.createChatMessage(options)	Object	Server scripts	Creates a chat message based on a specified role and text.
	llm.createDocument(options)	Object	Server scripts	Creates a document to be used as source content when calling the LLM.
	llm.embed(options)	Object	Server scripts	Returns the embeddings from the LLM for a given input.
	llm.embed.promise(options)	void	Server scripts	Asynchronously returns the embeddings from the LLM for a given input.
	llm.evaluatePrompt(options)	Object	Server scripts	Takes the ID of an existing prompt and values for variables used in the prompt and returns the response from the LLM. When you're using unlimited usage mode, this method also accepts the OCI configuration parameters.
	llm.evaluatePrompt.promise(options)	void	Server scripts	Takes the ID of an existing prompt and values for variables used in the prompt and asynchronously returns the response from the LLM. When you're using unlimited usage mode, this method also accepts the OCI configuration parameters.
	llm.evaluatePromptStreamed(options)	Object	Server scripts	Takes the ID of an existing prompt and values for variables used in the prompt and returns the streamed response from the LLM. When you're using unlimited usage mode, this method also accepts the OCI configuration parameters.
	llm.evaluatePromptStreamed.promise(options)	void	Server scripts	Takes the ID of an existing prompt and values for variables used in the prompt and asynchronously returns the streamed response from the LLM. When you're using unlimited usage mode, this method also accepts the OCI configuration parameters.
	llm.generateText(options)	Object	Server scripts	Takes a prompt and parameters for the LLM and returns the response from the LLM. When you're using unlimited usage mode, this method also accepts the OCI configuration parameters.
	llm.generateText.promise(options)	void	Server scripts	Takes a prompt and parameters for the LLM and asynchronously returns the response from the LLM. When you're using unlimited usage mode, this method also accepts the OCI configuration parameters.
	llm.generateTextStreamed(options)	Object	Server scripts	Takes a prompt and parameters for the LLM and returns the streamed response from the LLM. When you're using unlimited usage mode, this method also accepts the OCI configuration parameters.
	llm.generateTextStreamed.promise(options)	void	Server scripts	Takes a prompt and parameters for the LLM and asynchronously returns the streamed response from the LLM. When you're using unlimited usage mode, this method also accepts the OCI configuration parameters.
	llm.getRemainingFreeUsage()	number	Server scripts	Returns the number of free requests in the current month.
	llm.getRemainingFreeUsage.promise()	void	Server scripts	Asynchronously returns the number of free requests in the current month.
	llm.getRemainingFreeEmbedUsage()	number	Server scripts	Returns the number of free embeddings requests in the current month.
	llm.getRemainingFreeEmbedUsage.promise()	void	Server scripts	Asynchronously returns the number of free embeddings requests in the current month.
Enum	llm.ChatRole	enum	Server scripts	Holds the string values for the author (role) of a chat message. Use this enum to set the value of the options.role parameter in llm.createChatMessage(options).
	llm.EmbedModelFamily	enum	Server scripts	The large language model to be used to generate embeddings. Use this enum to set the value of the options.embedModelFamily parameter in llm.embed(options).
	llm.ModelFamily	enum	Server scripts	Holds the string values for the large language model to be used. Use this enum to set the value of the options.model parameter in llm.generateText(options).
	llm.Truncate	enum	Server scripts	The truncation method to use when embeddings input exceeds 512 tokens. Use this enum to set the value of the options.truncate parameter in llm.embed(options).

ChatMessage Object Members

Member Type	Name	Return Type / Value Type	Supported Script Types	Description
Property	ChatMessage.role	string	Server scripts	The author (role) of the chat message.
	ChatMessage.text	string	Server scripts	Text of the chat message. This text can be either the prompt sent by the script or the response returned by the LLM.

Citation Object Members

Member Type	Name	Return Type / Value Type	Supported Script Types	Description
Property	Citation.documentIds	string[]	Server scripts	The IDs of the documents where the cited text is located.
	Citation.end	number	Server scripts	The ending position of the cited text.
	Citation.start	number	Server scripts	The starting position of the cited text.
	Citation.text	string	Server scripts	The cited text from the documents.

Document Object Members

Member Type	Name	Return Type / Value Type	Supported Script Types	Description
Property	Document.data	string	Server scripts	The content of the document.
	Document.id	string	Server scripts	The ID of the document.

EmbedResponse Object Members

Member Type	Name	Return Type / Value Type	Supported Script Types	Description
Property	EmbedResponse.embeddings	number[]	Server scripts	The embeddings returned from the LLM.
	EmbedResponse.inputs	string[]	Server scripts	The list of inputs used to generate the embeddings response.
	EmbedResponse.model	string	Server scripts	The model used to generate the embeddings response.

Response Object Members

Member Type	Name	Return Type / Value Type	Supported Script Types	Description
Property	Response.chatHistory	llm.ChatMessage[]	Server scripts	List of chat messages.
	Response.citations	llm.Citation[]	Server scripts	List of citations used to generate the response.
	Response.documents	llm.Document[]	Server scripts	List of documents used to generate the response.
	Response.model	string	Server scripts	Model used to produce the LLM response.
	Response.text	string	Server scripts	Text returned by the LLM.
	Response.usage	llm.Usage	Server scripts	Token usage for a request to the LLM.

StreamedResponse Object Members

Member Type	Name	Return Type / Value Type	Supported Script Types	Description
Property	StreamedResponse.chatHistory	llm.ChatMessage[]	Server scripts	List of chat messages.
	StreamedResponse.citations	llm.Citation[]	Server scripts	List of citations used to generate the streamed response.
	StreamedResponse.documents	llm.Document[]	Server scripts	List of documents used to generate the streamed response.
	StreamedResponse.model	string	Server scripts	Model used to produce the streamed response.
	StreamedResponse.text	string	Server scripts	Text returned by the LLM.

Usage Object Members

Member Type	Name	Return Type / Value Type	Supported Script Types	Description
Property	Usage.completionTokens	number	Server scripts	The number of tokens in the response from the LLM.
	Usage.promptTokens	number	Server scripts	The number of tokens in the request to the LLM.
	Usage.totalTokens	number	Server scripts	The total number of tokens for the entire request to the LLM.

Related Topics

• N/llm Module
• SuiteScript 2.x Generative AI APIs
• SuiteScript 2.x Modules
• SuiteScript 2.1

General Notices