For Assessment Designers
Define exam flows, rubric criteria, and policies in a structured, versioned format. Compile to runtime configurations automatically.
Exam Runtime IR Specification
The canonical, versioned, executable specification of a published oral assessment. Bridging assessment theory and AI runtime execution.
For Runtime Engineers
Enforce hard constraints on transitions, follow-ups, time budgets, and evidence collection. Every state change produces a structured event.
For AI Examiners
Autonomous within bounded creative space. Follow up naturally, judge evidence, handle repairs — but cannot jump topics, reveal rubric, or bypass guardrails.
For Researchers
Grounded in oral assessment theory (Joughin 1998, Akimov & Malin 2020, Bayley et al. 2024, Fenton 2025). Design space systematically explored.
Architecture
Section titled “Architecture”┌─────────────────────────────────────────────────────────┐
│ AUTHORING STUDIO │
│ Exam Flow Model ──compile──► ExamRuntimePackage (IR) │
└──────────────────────────┬──────────────────────────────┘
│
┌──────────▼──────────┐
│ Exam Runtime IR │
│ (canonical spec) │
└──┬──────┬──────┬───┘
│ │ │
┌────────────▼┐ ┌─▼──────▼──────────┐
│ Pipecat │ │ Runtime Controller│
│ Adapter │ │ (policy engine) │
└──────┬───────┘ └────────┬───────────┘
│ │
┌────────────▼───────────────────▼────────────┐
│ PIPECAT RUNTIME │
│ LiveKit · STT · LLM · TTS · FlowManager │
└────────────┬───────────────────┬────────────┘
│ │
┌────────────▼────────┐ ┌──────▼──────────────┐
│ Event Store │ │ Evidence Ledger │
└─────────────────────┘ └──────────────────────┘Normative Language Legend
Section titled “Normative Language Legend”MUST Absolute requirementSHOULD RecommendedMAY OptionalMUST NOT Absolute prohibition