Skip to content
Inferect
DocsReference

Glossary

Terms used across the platform and this documentation, defined once.

3 min read

Terms

TermDefinition
Data planeThe OpenAI-compatible proxy surface — completions, model discovery, routing explain. API-key auth.
Control planeThe REST API for orgs, keys, credentials, providers, policies, dashboard, billing.
BFF proxyThe same-origin server route that injects auth so the browser never holds a raw token.
BYOKBring Your Own Keys — you supply provider credentials, envelope-encrypted per org, provider, and purpose.
RLSRow-Level Security — Postgres policies enforcing tenant isolation on every tenant table.
ProviderA model backend. 18 total: 5 native adapters, 13 OpenAI-compatible.
Routing decisionThe router's output: chosen provider, strategy, ranking, rejected candidates, confidence, estimated cost/latency/savings.
Scoring modeA routing strategy — latency, quality, balanced, coding, reasoning, vision, long_context, embeddings, or json.
Circuit breakerPer-provider health tracking that pulls unhealthy providers out of routing.
Semantic cacheA pgvector-backed cache keyed by embedding similarity, alongside exact-match.
Shadow modeObserve-only experimentation — mirrors traffic to candidates without touching production output.
EntitlementA feature or limit derived from a subscription's plan. Code checks entitlements, never plan names.
PlanA billing tier — Free, Business ($149/mo), or Enterprise.
Estimated savingsA modelled comparison against the most expensive eligible candidate for a request — never a billing figure.
Usage recordA per-request metering row: provider, model, latency, tokens, retries, fallover, HTTP status.
ADRArchitecture Decision Record — a binding, dated record of an architectural decision.