Skip to main content

What is REQQA?

Short answer: REQQA — Requirements Quality Assurance with AI — is a multi-tenant web application that helps teams write better software requirements by applying AI-driven analysis to them. You manage requirements, user stories, personas and glossary terms in one place, and REQQA uses large language models to analyse them for faults using its DeFOSPAM technique. It runs on the py4web framework with a MySQL back end, and analyses run as background jobs so you can keep working while they complete.

Detail​

REQQA exists to close the gap between writing requirements and knowing they are any good. Most teams write requirements in documents and discover the ambiguities, omissions and contradictions much later — in design, in test, or in production. REQQA brings the analysis forward: you author a requirement, ask REQQA to analyse it, and get back a structured list of issues to fix before anyone builds against it.

What you manage in REQQA​

Everything lives under an application, and every application begins with a mission statement — the top of the hierarchy that gives every requirement and story its shared context. Beneath that you work with:

  • Requirements — structured, versioned statements of what the system must do, each with a full change history.
  • Stories — user stories (Gherkin-style) generated from or linked to requirements.
  • Personas — the user classes that give requirements and stories their human context.
  • Glossary terms — a shared dictionary so the same word means the same thing across every requirement.

For how these fit together, see Key concepts and The mission.

How the AI analysis works​

REQQA's analysis technique is DeFOSPAM — REQQA's own method for finding faults in requirements and Gherkin stories. When you run an analysis, REQQA sends the requirement (plus its mission and glossary context) to a large language model and records each issue it finds, with a severity, against the requirement.

REQQA is not tied to a single AI provider. Each organisation configures its own model: in practice teams run both OpenAI GPT models and Anthropic Claude models, set per organisation along with the temperature and API key. Every AI call is logged with its token usage and timing, so the cost and behaviour of analysis are auditable rather than hidden.

note

REQQA never silently changes your requirements. Analysis produces issues for you to act on; the synthesis and cleanup steps that fold those issues back into requirement text are deliberate, reviewable actions — see The analysis engine and Synthesis and cleanup.

Built for teams, scoped by organisation​

REQQA is multi-tenant: every application, requirement, story and analysis belongs to an organisation, and data never crosses organisational boundaries. Access is invite-only — you don't sign up unilaterally, you're added to an organisation by someone who already belongs to one. See Access and onboarding and the FAQ on why REQQA is invite-only.

Because analysis can be slow, REQQA runs it as background jobs rather than making you wait on a page. You start an analysis, it queues, and you watch its progress and pick up the results when it finishes — see Background jobs.

Where to go next​