7.10 Manage Data Science Conversations

The Data Science Agent Conversations page lists all the conversations you created. Here, you create and manage conversations.

Create your Data Science Agent conversation to interact with the agent on these areas of data science and machine learning:

• Data profiling
• Data wrangling and transformation
• Statistical analysis of variable relationships
• Feature Importance, Classification, Regression, XGBoost, Clustering, and Anomaly Detection.
• Model training and evaluation
• Inference on new data

Here is the Data Science Conversations page. It lists all the conversations that you created. Click a conversation title to open it and resume it.

Figure 7-6 Data Science Conversation listing page

Description of "Figure 7-6 Data Science Conversation listing page"

This page lists the following details about the conversations:

• Title: This is the name of the conversation you provided while creating the conversation.
• Profile Name: This is the AI Profile you selected while creating the conversation. An AI Profile contains information about the user and their attributes, such as the provider, credential_name, and object_list. You can create and manage your AI profiles through DBMS_CLOUD_AI package.
• Messages: This indicates the number of interactions in the conversation with the agent. As shown in the screenshot, the conversation titled SQL has 2 messages or interactions as of the last updated date.
• Created on: This is the date on which the conversation was first created.
• Last Updated: This is the date on which the conversation was last used or updated.
• Status: There are two statuses ACTIVE and IDLE.

You can perform the following tasks here:

• Create a Data Science Agent Conversation
  A conversation is a set of interactions with Data Science Agent in the chat interface. Before you start a conversation with the Data Science Agent, you must create a conversation.
• Delete a Data Science Agent Conversation
  You can delete a conversation from the Data Science Agent Conversations listing page.

7.10.1 Create a Data Science Agent Conversation

A conversation is a set of interactions with Data Science Agent in the chat interface. Before you start a conversation with the Data Science Agent, you must create a conversation.

Note:

You can use the same Database user credentials to access the same conversation in multiple browsers. However, Oracle does not recommend this as it may lead to unexpected behavior. If you attempt this, Data Science Agent will display a warning, but you will have the option to override it.

To create a Data Science Agent conversation:

1. On the Data Science Agent Conversation page, click Create.

   Figure 7-7 Data Science Agent Conversations listing page

   
   

   
   Description of "Figure 7-7 Data Science Agent Conversations listing page"

   
   

   This opens the New Conversation dialog.
2. In the New Conversation dialog, enter the following details:

   Figure 7-8 Create Conversation dialog

   
   

   
   Description of "Figure 7-8 Create Conversation dialog"

   
   
3. In the Title field, provide a name for your conversation. In this example, create a conversation named Predict Subscription.
4. In the AI Profile drop-down menu, click on the down arrow and select a profile. Select the profile GROK_4_3_PROFILE.
5. Click Test to test the selected AI profile. AI profiles may show warnings if the parameters model, temperature, or max_tokens are outside the recommended DSAgent ranges.

   Note:

   A warning does not necessarily mean the profile cannot be selected. Review the warning before continuing.

   Note:

   Profile test failures can be caused by Access Control List (ACL), missing or deleted credentials, invalid credentials, invalid model, timeout, or unexpected DBMS_CLOUD_AI.GENERATE errors. To resolve any errors, check the ACL access, credential validity, model availability, and the request ID shown in the error.
6. Click OK. The conversation is created only if the selected profile test succeeds. Once the conversation is created, it opens the Data Science Agent chat interface. Here, you can start chatting with Data Science Agent.

   Note:

7. In the chat interface, Data Science presents tips to begin your conversation. In the Send a message field, type in your prompt in natural language and press Enter.

   Figure 7-9 Chat with Data Science Agent

   
   

   
   Description of "Figure 7-9 Chat with Data Science Agent"

   
   

7.10.2 Delete a Data Science Agent Conversation

You can delete a conversation from the Data Science Agent Conversations listing page.

To delete a conversation:

1. On the Data Science Agent Conversations listing page, select the conversation you want to delete and click Delete. The Delete conversation dialog opens.
2. In the Delete Conversation dialog:
   • Select Views and results table to drop views and tables associated with the conversation.
   • Select Machine learning models to drop machine learning models associated with the conversation.

   Figure 7-10 Delete Conversation

   
   

   
   Description of "Figure 7-10 Delete Conversation"

   
   
3. Click Delete. This completes the task of deleting the conversation.