Skip to content

Configuration

Where to find each configuration field for the LoxiLB Inference Gateway. This page is an index — it does not duplicate the field tables.

The full field reference lives elsewhere

The detailed serviceArguments reference for a load-balancer rule is the AI Gateway → Configuration Reference. Start there for any load-balancer create (POST /netlox/v1/config/loadbalancer) body.

How rules are configured

Manage routing rules through the load-balancer REST API on your protected management endpoint:

POST https://gateway.example.com/netlox/v1/config/loadbalancer
{ "serviceArguments": { ... }, "endpoints": [ ... ] }

Fields are set per-rule inside serviceArguments (and per-endpoint inside endpoints[]). See the Configuration Reference for the exhaustive tables and a worked example.

Quick lookup — field → page

Field(s) Concern Documented in
externalIP, port, portMax, protocol, host, path_prefix, path_match_mode Core service / L7 matching Configuration Reference §1
sel (0–10) LB algorithm Configuration Reference §2 · LLM Routing
mode (0–5, 4=fullproxy) NAT / proxy mode Configuration Reference §2 · Running Modes
security (0–2) TLS posture: plain, frontend termination, or frontend termination plus backend re-encryption Configuration Reference §8 · mTLS
cb_enable, vip_qos_policy_id Fullproxy circuit breaker; association to a pre-created QoS policy Configuration Reference §1 · AI Quotas and QoS
model_name, backend_protocol, trace_type AI model routing Configuration Reference §3 · Model Load Balancing
sse_mode, max_stream_duration_sec, backend_keepalive_interval_sec SSE streaming Configuration Reference §3 · SSE & Quota
kvExactMode, kvBlockSize, kvHashAlgo, kvZmqPort, kvEngineType, kvDpRankCount, kvExactApiMode, kvModelProfile, pdBootstrapPort Typed-engine, strict model-profile, and KV-cache routing (vllm, sglang, trtllm, llamacpp) Configuration Reference §4 · Model Profiles and KV-Exact Readiness · KV-Cache Routing
sockMapMode Experimental directional sockmap acceleration for eligible plain HTTP/1.1 fullproxy services Configuration Reference §1 · Sockmap Acceleration
pd_disagg_mode, pd_cache_aware_mode, pd_session_ttl_sec, pd_cache_threshold, pd_balance_abs_threshold P/D disaggregation Configuration Reference §6 · P/D Disaggregation
session_header_name Session affinity (sel: 3) Configuration Reference §7
mtls_frontend, alpn_protocols, tls_ciphers, tls_versions, hsts_* Frontend TLS / mTLS Configuration Reference §8 · Frontend mTLS
mtls_backend.verify_server_cert, backend_ca_cert_id, backend_client_cert_id, backend_tls_server_name, backend_tls_effective (read-only) Backend TLS verification and client certificate (security: 2) Configuration Reference §8 · Backend TLS
endpointIP, weight, targetPort, ep_role, nixl_port, backup Endpoint pool Configuration Reference §9
probetype, probereq, proberesp, probeRetries, inactiveTimeOut, timeout* Health & probes Configuration Reference §10
--aikey-db-*, AIGW_DB_PASSWORD Independent PostgreSQL AI-key and quota store AI Key Store Operations