A.1 FAQs

Topics:

A.1.1 Must tables and table columns have enhanced metadata, such as table names, column names, comments, or annotations to help the LLM understand the data better?

It depends on several factors, for example how clear and meaningful the table and column names are. While annotations and comments are not always required, they can provide valuable context and improve the agent’s understanding. The metadata enrichment capability in Data Studio - Transforms can further enhance the available context.

See Best practices to improve NL2SQL accuracy with Oracle Select AI for more information.

Parent topic: FAQs

A.1.2 Is Data Science Agent available in on premise databases?

Data Science Agent is available on Autonomous AI Database Serverless and Dedicated Region, only.

Parent topic: FAQs

A.1.3 What are the best practices for table and column annotations?

The table and column annotations must be semantically meaningful.

Parent topic: FAQs

A.1.4 Are there limits on the number of tables or database objects Data Science Agent can consume at one time?

Data Science Agent supports database object search and data discovery. This has been successfully tested with thousands of objects.

Parent topic: FAQs

A.1.5 Is NL2SQL used behind the Discovery stage?

For object discovery, the LLM receives relevant database object metadata together with the user’s prompt.

Parent topic: FAQs

A.1.6 How can we determine which models are best suited for use with Data Science Agent?

A useful way to evaluate model performance is to consult the Artificial Analysis Intelligence Index. As a general guideline, models with a score below 30 may not be suitable for the intended use case.

See:

Parent topic: FAQs

A.1.7 Can we integrate the Data Science Agent console with tools/agent outside of the database?

At present, integration of Data Science Agent with tools and agents outside the database is not possible. All tools and agent infrastructure run inside the database.

Parent topic: FAQs

A.1.8 Will Data Science Agent be available on Autonomous AI Database on Dedicated Infrastructure or AI Autonomous, Data Lakehouse?

Data Science Agent is an Oracle Machine Learning UI feature in Autonomous AI Database - Serverless that enables users to run complete data science workflows using natural language. Data Science Agent is available on Autonomous AI Database Serverless, which includes AI Lakehouse. It is not available on Autonomous AI Database - Dedicated.

Parent topic: FAQs

A.1.9 In a multi-agent architecture, how do you handle permissions and access control between agents?

Multi-agent architecture is on the roadmap.

Parent topic: FAQs

A.1.10 How does Data Science Agent filter object metadata for the LLM when hundreds of tables and thousands of columns could exceed the token budget?

The Data Science Agent design minimizes token consumption by performing object discovery outside the agentic loop of analysis and actions. It uses a lightweight catalog to select relevant objects. It is a temporary, compact summary created during object discovery. It describes accessible database objects at a high level, such as their names, types, descriptions, representative columns, and relationships, to help identify relevant objects. After that, it sends the full metadata retrieved for the selected objects into the loop. Depending on the object type, the full metadata can include all column names, data types and descriptions, row and column counts, object descriptions, and model details.

Note:

The lightweight catalog is created temporarily by Data Science Agent. It is not stored or managed as a standalone catalog by the customer.

Manually added objects are always included. See Add Objects to Data Science Agent Conversation for more information.

Parent topic: FAQs

A.1.11 Where are resources consumed for AutoML and Data Science Agent? Do they use database ECPUs, or do they run on separate dedicated machines?

All data analysis and machine learning model building take place in the database.

Parent topic: FAQs

A.1.12 Can we compare Data Science Agent with Codex using Autonomous AI Database MCP server?

Data Science Agent is an Oracle Machine Learning UI component in Autonomous AI Database - Serverless that enables users to run end-to-end data science workflows using natural language. It is designed specifically for data science workflows and includes specialized tools for machine learning, data analysis, and related tasks. By running entirely in-database, Data Science Agent delivers high performance, security, and operational simplicity.

Oracle Autonomous AI Database MCP Server is a built-in, managed service that lets AI agents securely interact with an Oracle Autonomous AI Database using the Model Context Protocol (MCP). Codex with Autonomous AI Database MCP server is better suited for broader development workflows, such as writing code, interacting with database resources, and integrating data operations into applications. The appropriate choice depends on the task, required tools, and desired level of specialization.

For more information, see Connecting Codex to Enterprise Data with Oracle Autonomous AI Database MCP Server

Parent topic: FAQs

A.1.13 Can we detect data drift, analysis, and model re-generation?

You can use OML Services from REST endpoints, or the Oracle Machine Learning Data Monitoring and Model Monitoring UI interfaces along with Oracle Data Science Agent, to detect drift.

See:

Parent topic: FAQs

A.1.14 Can Data Science Agent create notebooks?

Currently, Data Science Agent doesn't support the creation of notebooks. However, this feature is on the roadmap.

Parent topic: FAQs