Create an AI Profile

Steps to Create an AI Profile

To open Data Studio Settings, go to any tool and click the gear icon labeled Settings in the left navigation pane.

1. In Data Studio Settings go to the AI Profile tab.

2. Click Validate to check that the Generative AI service is accessible and the AI Profile is functioning correctly.

3. Click the plus (+) button next to the AI Profile name and select a mode.

   • Create Essential AI Profile: to create a basic AI profile with limited features.

   • Create NL2SQL AI Profile: to use object lists of table metadata for generating queries.

   • Create RAG AI Profile: to use Vector Index for semantic search over unstructured data.

   It opens the Create AI Profile page and provides the settings for the specific mode.

4. Specify the mandatory Service Settings for the AI Profile.

   • Profile Name: Enter a name for the AI Profile. You cannot modify the name after the profile is created.

   • AI Provider: Select the AI service provider:

     • OCI
     • Open AI
     • Azure OpenAI Service
     • Cohere
     • Google
     • Anthropic
     • Hugging Face

     You cannot modify AI Provider after the profile is created.

   • Credential: Select the credential to access the API service provider APIs. These are stored in the database and helps to securely connect to external AI service providers.

   • OCI Region: The location of the Generative AI cluster that you want to use.

     Each OCI region supports different LLMs, for example us-chicago-1, uk-london-1, and ap-osaka-1 offers many LLMs, while others have fewer. The page displays the models that are available in the selected region. If an AI Model does not appear, you can switch regions to refresh the model list.

     This field is available only if you select OCI as AI Provider. For more information, see Regions with Generative AI, Pretrained Foundational Models in Generative AI.

   • AI Model: Select the AI Model you want to use to generate responses. The AI Models available for use depend on the AI Provider you selected.

   You can use the Show Code toggle button to view the SQL statement generated automatically for the selected settings.

5. Under Advanced Settings you can provide additional attributes.

   • Service Settings

     Provide the settings for cloud services that the AI Profile can use. Service Settings is available only for OCI and Azure OpenAI Service. Depending on the AI Provider you selected, the page displays the corresponding service settings:

     If you select OCI as AI Provider, you can configure the following:

     • OCI API Format: Select the format the API can use to send and receive data. This attribute applies to OCI Generative AI Chat Models. Values are None, Cohere, Generic.

     • OCI Compartment ID: Specify the OCID of the compartment you want to access for using the OCI Generative AI service. The default is the compartment ID of the Pluggable Database (PDB).

     • OCI Endpoint ID: Indicates the OCID of the endpoint on a dedicated AI cluster. Specify this field to use a dedicated endpoint instead of an on‑demand model on a shared infrastructure.

     If you select Azure OpenAI Service as AI Provider, you can configure the following:

     • Azure Resource Name: Enter a name for the Azure OpenAI Service resource instance. It forms a part of the endpoint URL for accessing the service. For example:

       https://[your-resource-name].openai.azure.com/

     • Azure Deployment Name: Enter a name for the language model you want to deploy within the Azure OpenAI resource.

     • Azure Embedding Deployment Name: Specify a name for the deployed embedding model, for example text-embedding-ada-002, text-embedding-3-large, and so on. This is primarily used for creating vector representation of text.

     For more information, see Azure OpenAI Embedding skill - Azure AI Search | Microsoft Learn.

   • Generation Settings

     Provide the settings for Generative AI Services that the AI Profile can use.

     You can configure the behavior and limits of the AI Model accessed through the AI Profile. It allows customization for different scenarios, such as Generative AI Chat, SQL generation, or Natural Language Queries in Autonomous AI Database.

     • Conversation: This typically indicates whether the AI Profile supports conversational context, allowing the model to retain context over multiple interactions or prompt exchanges—useful for chat-style AI applications.

     • Case Sensitive Values: This setting indicates whether the input values or parameters should be treated as case-sensitive by the AI processing. Enable this to enforce exact case for inputs and parameter values.

     • Temperature: This is a slider parameter controlling randomness in the AI's responses. A lower temperature (e.g., 0.2) makes output more focused and deterministic, while a higher temperature (e.g., 0.8) yields more creative and diverse text generation.

     • Stop Tokens: These are token sequences that instruct the AI model where to stop generating further text output. When the model generates any of these tokens, it stops. This helps in controlling the response length and preventing unwanted trailing text.

     • Max Tokens: This defines the maximum number of tokens the AI model can generate in response to a prompt. It limits the length of the generated output to avoid overly long or expensive responses.

     • Target Language: Select the language that the AI Provider uses to respond.

     • Source Language: Select the language the AI Provider accepts in the input text.

If you selected Essential AI Profile, click Create AI Profile to create a basic AI Profile.

NL2SQL Settings

To create an NL2SQL AI Profile, specify the following Table Metadata.

• Object List Mode: Select from the following.

  • All: This option sends metadata for all tables in the specified schemas to the LLM, so it can generate SQL against any of them.

  • Automated: This option lets Select AI automatically choose a subset of relevant tables by running vector search on schema metadata, so only the most likely tables for a given prompt are used.

  • Selected Tables: This option limits metadata to only the specific tables you list in the AI Profile, so SQL generation is restricted to those tables.

  • None: This option does not send any table metadata from an object list, so the AI either cannot generate SQL over tables or rely on other configuration that does not use an explicit object list.

• Object List: Specify the database objects the LLM can use.

  Selecting this displays another option Enforce Object List. Enable this to restrict the LLM to generate SQL queries using only the objects mentioned in the object list.

• Annotations: Enable this to allow the AI to provide additional metadata about the database objects when generating results, for example column descriptions or tags.

• Comments: Enable this to allow the AI to use the table and column comments from the data dictionary to better understand the data.

• Constraints: Enable this to control to allow the AI to access information about primary keys, foreign keys, and other constraints and understand the relationship between tables for improving query accuracy.

If you selected the NL2SQL mode, click Create AI Profile to create the AI Profile for NL2SQL.

RAG Settings

For a RAG AI Profile, specify the following:

• Vector Index: Select the Vector Index to use for storing and searching the embeddings in your documents. This enables the AI to use the vectors in the database and find accurate response for the query. See Create Vector Index for more information.

• Embedding Model: It specifies which embedding model is used to convert your text into vector representations that are stored in that index, ensuring the semantic search behavior matches the model you configure for RAG.

• Enable Sources: Enable this to include the source document links or file names that contributed to the RAG result.

• Enable Source Offset: Enable this to include the start and end offsets within the converted text of the document in the RAG result.

• Enable Custom Source URI: Enable this to specify a custom source URI to show in the RAG result instead of the Object Storage URL.

If you selected the RAG mode, click Create AI Profile to create the AI Profile for RAG.