Building NL2SQL Pipelines With Oracle DBMS_CLOUD_AI and Oracle Integration Cloud

2026-07-05 · Source: HackerNoon · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Cloud Computing & IT Infrastructure · Depth: Advanced, long

Summary

Oracle's DBMS_CLOUD_AI package, also known as "Select AI," provides a native Natural Language to SQL (NL2SQL) capability within the Oracle Autonomous Database. This built-in feature allows developers to configure AI profiles that directly link database objects to external Large Language Model (LLM) providers like Cohere. The database internally manages prompt augmentation, schema metadata injection, and query execution, thereby reducing architectural complexity and mitigating LLM hallucination risks common in enterprise NL2SQL deployments. It offers four execution modalities: showsql for query validation, runsql for direct data retrieval, narrate for natural language summaries, and explainsql for query transparency. The system can be extended via Oracle Integration Cloud (OIC) to expose NL2SQL functions as RESTful APIs for broader enterprise access. However, data governance and security are critical, especially when transmitting schema metadata or result sets to external LLMs, prompting consideration for internal solutions like OCI Generative AI Service.

Key takeaway

For AI Architects and Database Administrators evaluating NL2SQL solutions, Oracle's DBMS_CLOUD_AI offers a robust, integrated approach within Autonomous Database. You should prioritize its native prompt augmentation and schema linking to reduce hallucination risks and simplify deployment. Consider leveraging Oracle Integration Cloud to expose these capabilities as secure RESTful APIs across your enterprise. Always conduct a thorough data governance assessment, especially when using external LLM providers, to ensure compliance with regulatory frameworks like GDPR or HIPAA.

Key insights

Oracle's DBMS_CLOUD_AI natively integrates NL2SQL within Autonomous Database, streamlining LLM interaction and mitigating common enterprise challenges.

Principles

• Embed NL2SQL directly within the database.
• Automate prompt augmentation for LLMs.
• Enforce least privilege for LLM access.

Method

Deploying DBMS_CLOUD_AI involves five steps: obtaining an LLM API key, granting execute privileges, configuring network access, securely storing credentials, and defining an AI profile binding LLM, credential, and database objects.

In practice

• Use "showsql" for SQL query auditing.
• Integrate "runsql" for direct application data retrieval.
• Employ "narrate" for conversational agents.

Topics

• NL2SQL
• Oracle Autonomous Database
• DBMS_CLOUD_AI
• LLM Integration
• Data Governance
• Oracle Integration Cloud

Best for: AI Engineer, Software Engineer, AI Architect

Related on AIssential

• Build AI models that know your enterprise. - Mistral AI — Mistral AI's Forge enables enterprises to build custom, domain-aligned LLMs from proprietary data with full lifecycle control.
• Oracle converges the AI data stack to give enterprise agents a single version of truth — Oracle's converged database architecture aims to solve data fragmentation and consistency issues for enterprise AI agents.
• Access OpenAI models and Codex through your Oracle cloud commitment — OpenAI models are now accessible to Oracle Cloud Infrastructure customers via existing Universal Credits.
• Chart your AI and agent strategy with Microsoft Marketplace — Azure is enhancing its cloud infrastructure with specialized AI hardware and integrated database capabilities.
• Large Databases Need Small, Open-Weight Language Models — Quantized, open-weight language models offer a cost-effective and performant alternative to proprietary APIs for database integration.
• MongoDB - Mistral AI — Mistral AI's SMoE architecture combined with MongoDB's unified data platform enables efficient, real-time enterprise AI applications.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by HackerNoon.