CTS Model Router Overview

The CTS Model Router is an enterprise-grade, OpenAI-compatible model gateway designed for high-throughput inference, intelligent provider failover, and localized billing.

Evolving from the foundational FaceGuard proxy, CTS Router acts as an Anycast intelligence layer between your client applications and frontier model providers (Anthropic Claude, OpenAI, DeepSeek, Google Gemini).

[ Client Application ]
         │ (OpenAI-compatible request)
         ▼
[ CTS Model Router (Port 8081 / Anycast Edge) ]
   ├── FaceGuard Policy Proxy (PII redaction & content safety)
   ├── Pre-Debit Token Ledger (Paystack / Naira / USD balance check)
   └── Dynamic Router Engine
         ├── Tier 1: Local / Dedicated Cluster (vLLM / MLX)
         ├── Tier 2: Cost-Optimized Cloud (DeepSeek-R1 / Gemini Flash)
         └── Tier 3: Frontier Fallback (Claude 3.7 Sonnet / GPT-4o)

---

Core Capabilities

OpenAI-Compatible Ingress

Drop-in replacement for OpenAI SDKs. Point your baseURL to https://api.creativetechstudios.org/v1 with your API key.

FaceGuard Security Proxy

Inline PII redaction, prompt injection defense, and cryptographic zero-retention transit. Prompts are never stored to persistent disks.

Local Naira & Paystack Ledger

Native support for African enterprise payments. Real-time token ledger with pre-debit holds and post-completion reconciliation.

Sub-Second Dynamic Failover

If an upstream provider suffers rate limiting (HTTP 429) or degradation (HTTP 5xx), requests automatically fail over within 180ms.

---

Next Steps

Quickstart GuideMake your first completions call in under 2 minutes with Python or cURL.

Architecture & LatencyUnderstand Anycast edge routing, provider failover, and memory guarantees.

Supported Model CatalogBrowse the complete catalog of routed models, pricing, and latency tiers.

API ReferenceFull specification for chat completions, embeddings, and billing endpoints.