AI Rule Agents v7.6 beta
Overview
The AI Rule Agents module provides two AI-powered agents that automate data quality rule generation across your connection profiles:
| Agent | How It Works |
|---|---|
| RRR Agent | Applies rules from your Reusable Rule Repository (RRR) to new connection profiles. Uses your existing curated rule library as a template to generate consistent, organization-approved DQ rules at scale. |
| AI Recommendation Agent | Analyzes raw metadata — table structures, column types, data patterns — and auto-generates tailored DQ rules including null checks, range validations, uniqueness constraints, referential integrity checks, and more. |
Both agents reduce manual rule authoring effort. The RRR Agent enforces your existing standards; the AI Recommendation Agent discovers new rules you haven't written yet.
Target Persona: Data Stewards, DQ Engineers, Data Engineers.
Prerequisites: - At least one connection with completed metadata extraction. - Reusable rules defined in the RRR (required for the RRR Agent; improves AI Recommendation output).
How This Module Helps
Instead of manually writing DQ rules for every table and column, the AI Rule Agents auto-generate rules based on your existing rule library or raw metadata. This: - Accelerates DQ setup from weeks to minutes - Ensures comprehensive coverage — both agents check every applicable column - Leverages your Reusable Rule Repository for consistent, standards-aligned rule generation - Supports ongoing monitoring — activate an agent to auto-generate rules when metadata changes
| Connects To | Relationship |
|---|---|
| Metadata Extract & Enrich | Agent reads extracted metadata |
| Reusable Rule Repository | Agent references repository rules |
| Review & Confirm (DQ) | Generated rules go here for review |
| DQ Jobs | Approved rules become executable |
| DQ Results | Rule execution results |
Navigation
| Method | Path |
|---|---|
| Sidebar | Observability → RightSight → AI Rule Agents (under Data Quality) |
| URL | /ai-rule-agent |
| Wizard | Augmented DQ Config → Step 5 (AI Rule Agents) |
| Deep link | /ai-rule-agent?fromWizard=true&connectionId=<id> |
Features & Workflows
1. Connection Profiles Grid
Each card displays connection info and agent status.
| Card Element | Description |
|---|---|
| Source icon | Database type |
| Profile name | Connection display name |
| Rules count | Number of AI-generated rules |
| Tables count | Number of tables analyzed |
| Agent status | Active (●) or Inactive (○) |
2. Configuring the AI Agent
Steps
- Click Config on a connection card.
- The Agent Configuration Drawer opens.
- Configure the agent settings:
| Setting | Type | Description |
|---|---|---|
| Rule Generation Mode | Dropdown | NL (Natural Language), Expression, Expert |
| AI Model | Dropdown | Select which AI model to use |
| Scope | Multi-select | Which schemas/tables to analyze |
| Rule Types | Checkboxes | Which DQ dimensions to generate (Completeness, Accuracy, etc.) |
| Sensitivity | Slider | How strict rules should be |
| Auto-Execute | Toggle | Enable/disable automatic execution on metadata changes |
| Max Rules per Table | Number | Limit rules generated per table |
- Click Save to apply configuration.
- The agent is configured for this connection.
- If Auto-Execute is enabled, the agent will run when metadata changes are detected.
3. Executing the AI Agent
Steps
-
Click Execute on a connection card.
-
The Agent Execution Drawer opens and connects via SignalR for real-time logs.
-
The agent begins analyzing metadata and generating rules. Progress is shown in real-time:
| Progress Element | Description |
|---|---|
| Stage tracker | Shows current phase (Analyzing, Generating, Validating) |
| Log stream | Real-time log entries |
| Rule count | Live counter of rules generated |
- On completion, a summary appears.
- Generated rules appear in Review & Confirm (DQ) with status = Pending.
- The card's Rules count updates.
- A toast notification confirms completion.
4. Viewing Execution Logs
Steps
- Click Logs on a connection card.
- The Log Drawer opens.
- Filter logs by level (Info, Warning, Error).
- Click any log entry for full details.
5. Viewing Results
Steps
-
Click Results on a connection card.
-
You are navigated to Review & Confirm (DQ) with the connection pre-selected and AI-generated rules filtered.
6. Activating / Deactivating the Agent
Steps
-
Click the context menu (⋮) on a connection card.
-
Select Activate or Deactivate.
- A toast confirms the status change.
- Activate — Agent monitors for metadata changes and auto-generates rules.
- Deactivate — Agent stops monitoring; rules are only generated on manual execution.
7. Searching and Filtering
- Type in the Search bar to find connections by name.
- Use Filter Badges to narrow by agent status.
Cross-Module Links
| From AI Rule Agents | Navigate To |
|---|---|
| Generated rules → review | Review & Confirm (DQ) |
| Connection → extraction config | Metadata Extract & Enrich |
| Rules → execution | DQ Jobs |
| Rules → results | DQ Results |
| Wizard → next step | Review & Confirm (DQ) |
Tips & Best Practices
- Define business terms and reusable rules before running the AI agent — the agent references them to generate contextually relevant rules.
- Start with a single connection to understand the rule quality before scaling.
- Use the Configuration Drawer to narrow scope to critical schemas/tables first.
- Review all AI-generated rules in Review & Confirm before activating them.
- Activate the agent for ongoing monitoring once you're satisfied with initial rule quality.