LyraTwo - AI-Assisted Product Requirements Management

AI Copilot (Cmd+J in editor)

Use slash commands to expand requirements into detailed sub-requirements, generate acceptance criteria from prose, decompose features into hierarchical tasks, or have the AI research a topic and insert findings directly into your document. The AI operates within your project context using semantic search over your existing documents.

Change Simulation and Impact Analysis

Before committing a change to an approved or active requirement, simulate its impact across the dependency graph. LyraTwo traces typed links (depends on, blocks, implements, derives from, supersedes, realizes) transitively and surfaces every downstream requirement that would be affected. Teams can review the full impact radius before a change is approved, reducing the risk of undetected regressions in dependent requirements.

Version History and Rollback

Every document change creates an immutable snapshot. The full version history is available at any time and any prior version can be restored with a single action. Review comments, status transitions, and change rationale are preserved alongside the version trail, giving teams a complete audit record of how a requirement evolved and why.

ReqIF Import and Export

LyraTwo speaks ReqIF -- the ISO standard interchange format used by enterprise requirements toolchains including IBM DOORS, Polarion, Jama, and Codebeamer. Import existing requirement sets from any ReqIF-compliant tool and export your documents back out in the same format. Teams migrating from or integrating with enterprise tooling do not need to manually recreate their requirements.

Agentic Skills and Third-Party Integration

Extend LyraTwo with custom AI skills installed directly from Git repositories. Skills are pydantic-ai agents that can call external APIs, query databases, or run arbitrary logic -- allowing teams to integrate LyraTwo with their own tooling, CI systems, project trackers, or internal knowledge bases. Pre-built skills are included for common operations; the skill registry makes it straightforward to add your own without modifying the core application.

Project Directives and Custom Requirement Designs

Each project can carry a directives document -- a guidance file that instructs the AI copilot on how requirements should be structured, what terminology to use, which acceptance criteria patterns to follow, and what standards apply. This allows teams with established requirement methodologies (SAFe, EARS, MoSCoW, ISO 26262, DO-178C, or custom internal standards) to encode those conventions once and have the AI enforce them consistently across all documents in the project.

Rich Document Editor

Built on the Plate editor with full rich text support including tables, code blocks, inline comments, and suggested edits. Documents auto-save and every change creates an immutable version snapshot.

Requirement Lifecycle

Documents move through a structured lifecycle: draft, proposed, in review, approved, active, deprecated. Each transition is recorded with a rationale and full audit trail.

Export Beyond ReqIF

Export individual documents or full project reports as PDF, DOCX, or Markdown in addition to ReqIF. Generate executive summaries and engineering handoff documents. Import from structured documents and Confluence in addition to ReqIF.

SSO and Access Control

Local accounts for simple deployments. Full OIDC SSO support for Azure AD, Google Workspace, Okta, Auth0, and Amazon Cognito. Role-based project membership: owner, manager, contributor, viewer.

Bring Your Own LLM

LyraTwo uses LiteLLM for model abstraction. Configure any supported model -- Claude, GPT-4, Azure OpenAI, or a locally hosted model -- via a single environment variable. Your data never leaves your AWS account.

DEPLOYMENT MODEL

LyraTwo deploys as a single ECS Fargate service backed by RDS PostgreSQL, sitting behind an Application Load Balancer. The included CloudFormation template provisions all required infrastructure into your existing VPC. No new VPC is created. Stack creation completes in approximately 12 minutes. The application serves both the API and the React frontend from the same container.

PREREQUISITES

• AWS account with permissions to create ECS, RDS, ALB, IAM, and Secrets Manager resources
• Existing VPC with at least 2 public subnets and 2 private subnets across different Availability Zones
• Private subnets with outbound internet access via NAT Gateway (or deploy with VPC endpoints enabled)
• API key for your chosen LLM provider (Anthropic, OpenAI, Azure OpenAI, or compatible)

SUPPORT

Documentation and support at https://thevgergroup.com