Agentic AI How-to Guide

Agentic AI How-to Guide

Visulate for Oracle integrates Google Gemini AI to provide autonomous agents that assist with database analysis, query generation, and documentation.

Chatbot Interface

The primary way to interact with Visulate’s AI is through the chat window.

Select a Database and Schema: The AI needs context to provide accurate answers.
Open Chat: Click the chat icon or look for the chat panel on the right.
Ask Questions: You can ask plain English questions like:
- “Show me all tables related to ‘EMPLOYEES’.”
- “Explain the business logic in the ‘SALARY_CALC’ package.”
- “Are there any circular dependencies in this schema?”

Natural Language to SQL (NL2SQL)

Visulate can translate your natural language requests into executable SQL based on your schema.

Request SQL: In the chat, ask for data. Example: “List the top 5 customers by total order volume in 2023.”
Review & Execute: The AI will generate the SQL. Review the code for correctness.
Apply Filters: Use the “Run it” button or copy the SQL into the Query Editor to download the results as CSV.

Automated ERD Generation

Visualize your schema relationships on demand.

Request a Diagram: Ask the AI to visualize a set of tables. Example: “Generate an ERD for the order processing tables.”
Specialized Agent: The Root Agent will delegate this to the ERD Agent, which analyzes foreign keys and uses Mermaid to render a diagram.
Refine: You can ask to add or remove tables from the diagram: “Add the ‘PAYMENTS’ table to this diagram.”

Schema & Object Analysis

Get high-level summaries or deep-dives into database objects.

Schema Summary: Ask “Analyze the ‘HR’ schema” to get a report on object counts, storage usage, and overall complexity.
Problem Detection: Ask “Identify any tables missing primary keys” or “Find invalid objects in my schema.”
Object Deep Dive: Navigate to a specific object (e.g., a Package) and ask the AI to “Summarize what this package does” or “Find potential performance issues in this code.”

AI-Powered Code Comments

Document your legacy PL/SQL code in bulk.

Request Documentation: Ask the AI to “Generate documentation for all tables in the ‘FINANCE’ schema missing comments.”
Background Processing: The Root Agent uses the Comment Generator Agent to query the data dictionary and prepare a set of COMMENT ON statements.
Review & Download: The results are presented as a downloadable CSV or a list of SQL commands you can execute in your preferred tool.

Session Context

Visulate maintains conversational context. You can refer to previous answers:

“Now show me the DDL for the first table you mentioned.”
“Run that same query but filter for ‘active’ status only.”

[!TIP] Always ensure your GOOGLE_AI_KEY is correctly configured in your Quickstart setup to enable these features.