Select AI enhances data interaction and enables developers to build AI-driven applications directly from SQL, transforming natural language prompts to SQL queries and text responses, supports chat interaction with LLMs, enhances response accuracy with current data using RAG, and generating synthetic data.

Use cases include:

• Generate SQL from natural language prompts

  Developer productivity: Select AI significantly enhances developer productivity by providing "starter" SQL queries quickly. Developers can input natural language prompts, and Select AI generates SQL based on your database schema tables and views. This reduces the time and effort needed to write complex queries from scratch, allowing developers to focus on refining and optimizing the generated queries for their specific needs.

  Natural language queries for end-users: Select AI empowers end-users to interact with your application's underlying data tables and views using natural language queries. This functionality allows users without SQL expertise to ask questions and retrieve data directly, making data access more intuitive and user-friendly relative to the capabilities of LLM being used and the quality of the schema metadata available.

• Customized media generation

  Select AI can be used to generate personalized media content such as emails tailored to individual customer details. For instance, in your prompt you could instruct the LLM to create a friendly and upbeat email encouraging a customer to try a set of recommended products. These recommendations could be based on customer demographics or other specific information available in your database. This level of customization enhances customer engagement by delivering relevant and appealing content directly to the customer.

• Code generation

  Select AI leverages the chat action, you can use Select AI to ask your specified LLM to generate code from natural language prompts. This feature supports various programming languages such as SQL, Python, R, and Java. Examples include:

  • Python Code: "Write the Python code to compute a confusion matrix over a DataFrame with columns ACTUAL and PREDICTED."
  • SQL DDL: "Write the DDL for a SQL table with columns name, age, income, and country."
  • SQL Query: "Write the SQL query that will use the Oracle Machine Learning in-database model named CHURN_DT_MODEL to predict which customers will churn and with what probability."

• Retrieval Augmented Generation (RAG)

  Use vector store content for semantic search retrieval to enhance prompt accuracy and relevance in LLM responses.

• Synthetic data generation

  Generate synthetic data using LLMs that conforms to your schema for solution testing, proofs of concept, and other uses. Synthetic data can support better testing of your applications in the absence of real data, leading to overall quality of your application.

  Synthetic data generation can also be used to populate an Autonomous Database clone or a metadata clone. A metadata clone is a copy of a database that includes only its structure and not the actual data. Select AI supports generating synthetic data for such clones. Using synthetic data helps to protect sensitive data while enabling development, testing, and validating user experiences. It’s also useful for AI and machine learning projects needing sample data for model training or test data for scoring.