12 Select AI Agent Framework Interfaces

Select AI Agent Framework interfaces support environments such as SQL, PL/SQL, and Python.

Topics:

• Select AI Agent Framework for SQL
  Use Select AI Agent Framework for SQL to create and run AI agents that interact with database data using natural language. Instead of manually writing and coordinating SQL queries, describe your objective and let the agent determine the appropriate steps, generate SQL, and return the results.
• Select AI Agent Framework for Python
  Select AI Agent Framework for Python builds on the Select AI for Python client library, select_ai, enabling you to leverage DBMS_CLOUD_AI features within Autonomous AI Database directly from Python. The module extends support for advanced generative AI workflows and agent-based automation through the select_ai.agent submodule.

12.1 Select AI Agent Framework for SQL

Use Select AI Agent Framework for SQL to create and run AI agents that interact with database data using natural language. Instead of manually writing and coordinating SQL queries, describe your objective and let the agent determine the appropriate steps, generate SQL, and return the results.

Select AI Agent Framework for SQL uses generative AI capabilities, including large language models (LLMs), to enable agent-based interactions with data. Agents can interpret natural language requests, plan and execute database tasks, use available tools, and generate SQL queries to retrieve and analyze data.

Select AI Agent Framework for SQL uses the DBMS_CLOUD_AI and DBMS_CLOUD_AI_AGENT packages to define, manage, and run AI agents directly within the database.

For more information, see the documentation for DBMS_CLOUD_AI and DBMS_CLOUD_AI_AGENT.

12.2 Select AI Agent Framework for Python

Select AI Agent Framework for Python builds on the Select AI for Python client library, select_ai, enabling you to leverage DBMS_CLOUD_AI features within Autonomous AI Database directly from Python. The module extends support for advanced generative AI workflows and agent-based automation through the select_ai.agent submodule.

What You Can Do

You can design and orchestrate agentic workflows in Python with the following classes:

• select_ai.agent.Tool
• select_ai.agent.Task
• select_ai.agent.Agent
• select_ai.agent.Team

These classes enable you to programmatically define tools, construct tasks, configure agents, and assemble multi-agent teams in Python. This approach closely mirrors the structure and capabilities provided by the DBMS_CLOUD_AI_AGENT package in the database, giving you flexible control and seamless integration with Autonomous AI Database AI operations.

Async Select AI Agent Framework Support

The select_ai.agent module also includes asynchronous versions of its core classes, enabling you to build and run agent workflows using Python’s async and await. These async classes are designed for co-routine-based applications and enable non-blocking interaction with the database.

You can use the following async classes:

• select_ai.agent.AsyncTool

• select_ai.agent.AsyncTask

• select_ai.agent.AsyncAgent

• select_ai.agent.AsyncTeam

• Class-level delete methods: These class-level methods delete database objects directly without requiring initialization of a proxy object.

  • Agent.delete_agent(agent_name)

  • Task.delete_task(task_name)

  • Tool.delete_tool(tool_name)

  • Team.delete_team(team_name)

See "Async AI agent framework examples" in Select AI for Python to explore.

These asynchronous classes support the same core functionality as their synchronous counterparts, including:

• Creating tools for natural-language-to-SQL generation (NLSQL), web search, retrieval-augmented generation (RAG), PL/SQL, notifications, and custom functions

• Configuring task logic and tool usage

• Assigning agent roles and profiles

• Assembling and running agent teams programmatically

This enables you to build scalable AI pipelines that integrate naturally with Python async applications.

For complete API reference, see Select AI for Python guide.