The Product Alpha that operates enterprise AI project by project.
The operating control plane—Console, Customer Portal, Knowledge/RAG, Gateway, Usage/Billing, Agent Approval and Incident/Alert—is implemented as Product Alpha. The Policy Router Lab below is only an anonymous public simulation and does not call external models. Live provider traffic is gated by credentials, commercial rights and activation.
Policy Router Lab
POST /api/aifabric/route
objective + workload → policy decision
SYSTEM MODULES
Core operating modules implemented in Product Alpha
Unified Gateway
OpenAI-compatible endpoint, auth, rate limits and fallback
Policy Router
Select providers by workload, cost, latency and governance
Org / RBAC
Company, team, user, role and API-key isolation
Usage Ledger
Token/GPU metering, cost basis and customer billing
Data Control
Korea boundary, PII policy, retention and transfer hooks
Audit Trail
Evidence of who used which model under which policy
RAG / Agent
Retrieval, tool permission, agent workflow and evaluation
Dedicated Plane
Customer-isolated GPU/model/endpoint and SLA operations
ARCHITECTURE
Target Architecture
Public Policy Lab
An anonymous public lab for the routing-policy interface. Authenticated customer APIs and provider runtime operate separately inside the project control plane.
POST /api/aifabric/route
{ "objective": "governance", "workload": "reasoning" }PRODUCT ALPHA LIVE · PROVIDER ACTIVATION GATED — the operating control plane is implemented; each external provider still requires credentials, commercial rights, contract and activation validation before live traffic.