ZYK Chat · 智聊
Private multi-model educational chat with role-aware access, subject modes, approved knowledge sources and institution-defined controls.
ZYK platform
An institutional AI environment for education — not just six AI tools.
The proposed ZYK architecture combines user-facing AI applications with approved school knowledge, assessment services, governance controls, model choice and integration. Models can change over time while the institution keeps control of identity, policy, data and educational workflows.
User experiences
Six focused interfaces for everyday work.
ZYK Create · 智创
Governed image, video and media creation for educational use, with review and AI-generated-content labelling workflows where appropriate.
ZYK Analyze · 智析
OCR, document analysis, writing feedback, extraction and question-generation workflows operating within school-controlled environments.
ZYK Voice · 智声
Pronunciation practice, speech, audiobook and language-learning workflows, with sensitive capabilities subject to explicit policy and consent.
ZYK Code · 智码
Curriculum-aligned coding assistance for Python, JavaScript, C++ and other languages approved by the institution.
ZYK Agent · 智助
Bounded assistants for defined institutional roles such as admissions, counselling, library, IT helpdesk, HR and finance.
Education intelligence
Knowledge and assessment turn generic AI into school AI.
ZYK Knowledge
A proposed institutional RAG and knowledge layer for approved curriculum, rubrics, policies, handbooks, procedures and school resources. Its job is to govern what institutional information an AI experience can retrieve and cite.
Explore ZYK Knowledge →ZYK Assess
A proposed assessment layer covering diagnostic, placement, internal and mock testing; question banks; listening and speaking workflows; AI-assisted feedback; progress analytics and future adaptive testing.
Explore ZYK Assess →Preserved extensions
Useful ideas from earlier concepts remain in the architecture — without being presented as finished products.
Class & team workspaces
Shared class folders, teacher-team spaces and role-aware collaboration can organise AI workflows around real classes and departments rather than isolated personal accounts.
Parent portal & communication
Future workflows may support parent-facing explanations, consent, AI-literacy resources and appropriate learning or usage information. A parent portal should not be presented as a completed product until validated.
SIS / MIS integration
Beyond LMS, SSO and APIs, institutions may need deeper student-information, management-information, roster or examination-system integration according to scope.
Real-time translation
The Voice direction preserves real-time translation as a possible capability. Supported languages, latency, accuracy and educational suitability require independent validation.
Consent-controlled voice cloning
Earlier concepts included voice cloning. Any implementation should be treated as high-risk: explicit consent, revocability, restricted purposes and institutional controls rather than default availability.
Similarity / plagiarism assistance
Analyze can preserve similarity and academic-integrity support workflows, but outputs should remain evidence for educator review rather than authoritative declarations of plagiarism.
Two different control functions
Govern users and policy with Control. Govern model routing with the Gateway.
ZYK Control · governance plane
Control is the proposed administrative layer for people and institutional policy: users, roles, permitted capabilities, quotas, audit, consent, retention, filtering, approval workflows and operational visibility.
Explore ZYK Control →AI Gateway · model-serving plane
The Gateway is the proposed technical routing layer behind approved experiences. It selects or restricts model endpoints according to task, data policy, approved-model status, capacity and performance rules defined by the institution.
Explore AI Gateway →The distinction matters: Control answers who may do what under which institutional rules; the Gateway answers which approved model endpoint may serve that authorised request. Neither replaces the infrastructure layer that actually hosts or connects to models.
Infrastructure & integration
Keep the school experience stable while the technology underneath evolves.
Model portability
The goal is to avoid rebuilding the school experience around every new model release. Models and serving technology can evolve while identity, policy, knowledge and workflows remain stable.
Identity, LMS & APIs
Role-based access, SSO, roster, LMS, API, database and—where appropriate—SIS/MIS integrations can connect approved AI workflows to existing institutional systems according to project scope.
Operations
Monitoring, backups, updates, performance management, lifecycle planning and support make the platform an ongoing institutional service rather than a one-time server purchase.
Explore infrastructure →Knowledge, Assess, Control and AI Gateway are content-v2 product architecture concepts. Class/parent workspaces, SIS/MIS integration, real-time translation, voice cloning and similarity assistance are preserved earlier or extension concepts. Until implementation status is validated, all should be treated as product-development directions rather than fully launched capabilities.