# Understanding Rule Intelligence

> Rule Intelligence turns requirement documents and API specs into structured, testable requirements—review them, then feed them into test generation with coverage and traceability.

Source: https://totalshiftleft.ai/help-center/product-documentation/understanding-rule-intelligence

## Overview

**Rule Intelligence** turns human-written requirement documents and API specs into structured, testable requirements. You upload BRDs, PRDs, Swagger/OpenAPI, Excel, Confluence exports, and backlogs; the platform extracts discrete **requirements** (field constraints, business rules, error handling, and more); you review and approve them; and they drive test generation so your tests reflect the documented rules.

This article maps the whole module and its five tabs. Explore the capability at [requirements extraction](/features/requirements-intelligence/requirements-extraction).

## Before you begin

Rule Intelligence appears in two places, and both open the same five-tab workspace:

- **Inside the project editor.** Open a project for editing and click the **Requirements** tab in the editor's tab bar. The workspace mounts directly on the tab.
- **The standalone Rule Intelligence hub.** The hub opens on a **Select a project** screen headed *"Rule Intelligence"* with the note *"Import requirement documents (BRD, PRD, Swagger, Excel, Confluence, etc.) to extract business rules for test generation."* Click a project card (each card shows a folder icon, the project name, and description) to open the workspace for that project.

You also need:

- At least one **project**. If none exist, the hub shows *"No projects found. Create a project first."*
- An **AI provider** configured for your workspace (**Settings → AI**) for the AI Parse step. Uploading files and extracting text work without it; parsing does not.

## Step 1 — Learn the five tabs

Inside the workspace, a secondary tab bar runs across the top. From left to right:

| Tab | What it does |
|-----|--------------|
| **Requirement Documents** | Upload BRD/PRD/Word/Excel/etc., extract text, AI-parse into typed requirements, then review and approve them. |
| **Data for Generating Tests** | Upload sample Excel/CSV data; its rows become **fixtures** so generated tests use realistic inputs. |
| **Generation Setup** | Check **Generation readiness** per endpoint and (optionally) override how fixture fields map into requests and assertions. |
| **Coverage** | The **Coverage & Extraction Fidelity** dashboard — a per-endpoint × dimension × strategy matrix plus per-artifact extraction quality. |
| **Learning** | The read-only **What your workspace learned** dashboard, driven by reviewer edits and test-run outcomes. |

The active tab is underlined in blue. Switching tabs clears any success or error banner from the tab you left, so a message from one action doesn't follow you into another tab.

## Step 2 — Follow the document-to-tests flow

The core loop lives on **Requirement Documents**, where each uploaded file is a row with a three-step, always-visible, re-runnable action set:

1. **Upload** a document. It starts at status **Uploaded**.
2. **Extract text** — the green **play** button. Status becomes **Text Extracted** and the row shows *Ready to parse*.
3. **AI Parse** — the blue **sparkle** button (enabled once text is extracted). The platform extracts typed requirements, status becomes **Parsed**, and the **Extracted Requirements** panel opens automatically.
4. **Show / Hide requirements** — the purple **list** button (enabled once parsed) toggles that same review panel.
5. **Review** each requirement — **Approve**, **Reject**, **Edit**, or **Delete**.

**Only APPROVED requirements drive generation.** Pending (AI-extracted) and rejected requirements are ignored by the generator.

## Step 3 — Understand requirement types

Extracted requirements are typed. The review panel labels them with short chips: **Field Constraint**, **Business Rule**, **State Transition**, **Cross-Field**, **Security**, **Performance**, **Data Format**, **Error Handling**, and **Integration** — plus backlog-shaped types **User Story**, **AC** (acceptance criterion), **Capability**, and **Use Case**.

## Notes and accuracy guardrails

- **Supported protocols are REST, SOAP, and GraphQL.** Rule Intelligence draws facts from OpenAPI, WSDL, and GraphQL sources. gRPC is not supported.
- **Confidence gate.** Any AI-extracted requirement below **70%** confidence is flagged **Needs review** rather than treated as ready.
- **Re-parse is destructive.** Re-parsing a document that already has approved or human-reviewed requirements wipes and replaces them; the platform shows a **Re-parse will replace reviewed requirements** confirmation first.
- **Two kinds of "data."** *Data for Generating Tests* supplies sample inputs to the generator; data used when **running** tests lives elsewhere (Project Settings → Datasets).

## Related articles

- [Importing and reviewing requirements](/help-center/product-documentation/importing-and-reviewing-requirements)
- [Coverage and fidelity](/help-center/product-documentation/coverage-and-fidelity)
- [Requirements traceability matrix](/help-center/product-documentation/requirements-traceability-matrix)
- [Test Generation Settings](/help-center/product-documentation/test-generation-settings)