[ API specifications // Sub-15ms endpoints ]

Low-Latency, Zero-Egress REST & WebSocket Specs

Every endpoint is designed to return before a round trip to a public model would have completed, and to guarantee that no raw identifier crosses the boundary in either direction.

Specification preview This page documents the intended interface. The endpoints are not yet live, and the request and response shown below are illustrative rather than captured from a running service.

Request / response

Request POST/v1/ingest/scrub

{
  "record": {
    "name": "A. Yılmaz",
    "email": "a.yilmaz@example.com",
    "phone": "+90 5xx xxx xx xx",
    "account": "TR33 0006 1005 1978 6457 8413 26"
  },
  "intent": "catchment_scoring",
  "policy": "local_first"
}

Response

{
  "token": "tkn_9f2a…c41",
  "intent": "catchment_scoring",
  "masked": ["name", "email", "phone", "account"],
  "routed_to": "edge/local",
  "raw_retained": false
}

Endpoints

POST /v1/ingest Submit a record for in-enclave scrubbing and tokenisation.
POST /v1/route Route a tokenised intent across the multi-LLM layer on live policy.
WS /v1/stream Sub-15ms bidirectional stream for continuous scoring workloads.
AUTH Bearer enclave token Short-lived credentials issued inside the perimeter; never brokered externally.

Specification

OpenAPI 3.1 SpecificationSoon Machine-readable contract for every endpoint
WebSocket Protocol NotesSoon Framing, backpressure and reconnect semantics

[ Sovereign deployment ready // zero-trust enclave ]

Ready to sovereignize your enterprise AI architecture?

Bridge subterranean data harvesting with high-velocity edge routing. Deploy Traffelo AI inside your private cloud or on-premise perimeter to scrub PII in-flight, support regulatory compliance, and orchestrate Multi-LLM workloads at enterprise scale.

Schedule Enterprise Architecture Audit Explore Technical Docs

Air-gapped & private cloud native
Zero raw data egress
KVKK / GDPR / PDPA / BNM ready
Sub-15ms latency SLA