Glossary

Also known as: API · endpoint contract · API agreement

An agreement that makes a provider and its consumers interpret inputs, outputs, authorization, errors, and state in the same way. It owns the observable boundary rather than a screen or implementation shape.

GET /api/search?q=term · 401 · 403 · 404 · 409 · 500

Also known as: idempotent · duplicate request safety · retry safety

The property that repeated processing of the same intent converges on the same observable business state as one processing. It limits duplicate effects instead of pretending retries do not happen.

Idempotency-Key: order-123 · ON CONFLICT DO NOTHING

Also known as: SSOT · source ownership · one authoritative source

A principle that gives one authoritative source the meaning and write authority for a fact while treating other representations as derived state. It clarifies ownership and synchronization boundaries; it does not mean copies never exist.

notes/ · courses/ · apps/console/sql/

Also known as: DB migration · schema migration · DDL change

A versioned change that moves operational data structures or constraints from one state to the next. Successful SQL syntax is a separate judgment from compatibility and recoverability.

CREATE INDEX CONCURRENTLY · forward fix

Also known as: rate limiting · request quota · call limit

A policy that limits request volume by time, subject, or route to protect resources and external spend. The contract includes the exceeded response and retry guidance, not just a number.

60 requests / minute · HTTP 429

Also known as: caching · cache hit · cache miss

A way to store a reusable copy of a result to reduce latency and source load. A cache hit proves speed, not freshness, authorization, or consistency.

Cache-Control: max-age=300 · cache hit / miss

Also known as: SSE · EventSource · server events

A contract where a server keeps an HTTP connection open and sends text events to a client in one direction. Separate the wake-up signal from business data and define a reconnect cursor.

Content-Type: text/event-stream · Last-Event-ID: 42

Also known as: TURN cost gate · video chat cost gate · WebRTC spend gate

A rule that separates WebRTC feature code from operational entry points and keeps expensive TURN, SFU, and video paths disabled until an explicit configuration and cost approval exist.

WEBRTC_VOICE_ENABLED=false · TURN·SFU·video disabled

Also known as: Fixture · test setup data · repeatable test state

A small bundle of input, state, and doubles prepared so one test can be reproduced under the same conditions. It differs from an evaluation dataset that represents an operational distribution.

seeded PostgreSQL row · temporary upload file

Also known as: S3 · object store · object storage

A system that stores large file bytes under an object key with metadata. Storage success is not the same as business completion or authorization verification.

bucket/object-key · presigned URL

Also known as: capability file · Tauri permission · Tauri capability

An explicit boundary that groups the commands and permissions available to a Tauri window. It grants only the needed capabilities instead of opening a local server or filesystem broadly for convenience.

core:default · block localhost:8000

Also known as: 작업 스케줄러 · 예약 실행기 · Scheduler

A role that creates or dispatches asynchronous work when a configured time, interval, or condition is reached.

Also known as: 객체 관계 매핑 · ORM

A library or technique that maps application objects to relational tables and helps compose queries and transactions.

Also known as: DB Pool · DB 연결 풀 · DB Connection Pool

A resource-management structure that keeps a bounded set of expensive database connections for requests to borrow and return.

Also known as: Execution Plan · SQL Plan · Query Execution Plan

The estimated cost and order of table scans, index access, joins, and sorting that a database chooses to execute a SQL statement.

Seq Scan · Index Scan · Index Only Scan

Also known as: Postgres · PostgreSQL

An open-source server-based relational database that accepts multiple connections and provides SQL, transactions, constraints, indexes, and operational recovery features.

Also known as: SQLite

An embedded, file-based relational database that provides SQL, transactions, and constraints without a separate server process.

Also known as: JSON 웹 토큰 · JWT

A standard representation that carries claims about a subject and usage conditions, protected for integrity with JWS or encrypted with JWE. The name JWT alone does not reveal whether a signature, MAC, or encryption is used.

Also known as: Audit Trail · Security Audit Record · 감사 기록

A tamper-resistant record of who attempted to read or change a resource, when, why, and with what result. High-risk work records durable intent before execution and a separate result afterward.

Also known as: safe default · Fail-closed

A policy that rejects an action by default when validation or a dependency fails, preventing an uncertain result from being accepted as success.

Also known as: 지알피시 · gRPC

An RPC framework that commonly combines Protocol Buffers contracts, generated code, and HTTP/2 transport.

Also known as: 웹소켓 · raw WebSocket · WebSocket

A persistent bidirectional protocol that lets both client and server send messages after one connection is established.

Also known as: 웹 실시간 통신 · WebRTC

A collection of standards and APIs for low-latency audio, video, and data transport between browsers and devices.

Also known as: 아이스 · 연결 경로 탐색 · ICE

The process by which WebRTC peers gather network candidates, test connectivity, and select a communication path.

Also known as: 턴 · TURN

A protocol and server role that relays media or DataChannel packets between peers when a direct path is unavailable.

Also known as: AI cost gate · LLM spend gate · provider cost gate

An execution boundary that blocks usage-priced external AI providers by default and permits calls only after data transfer and budget approval.

default-closed cloud provider · enable after token budget approval

Also known as: 음성 합성 · Speech Synthesis · TTS

An AI capability that converts text into playable speech using a selected voice, speed, and output format.

Also known as: 검색 증강 생성 · RAG

An architecture that retrieves relevant material, adds it to the model context as evidence, and then generates an answer.

Also known as: HNSW Index · HNSW

An ANN index that follows layered neighbor graphs from a broad search toward a more precise search.

Also known as: Postgres Vector Extension · pgvector

A PostgreSQL extension that adds vector types, distance operators, and exact or approximate-nearest-neighbor indexes.

An frontend build tool that provides an ES module development server and production bundles.

An app framework that runs a web UI in the operating-system WebView and gates native capabilities through Rust commands and permissions.

A state library that manages shared React UI state through small stores and selectors.

A server-state library for managing remote data fetching, caching, retries, and invalidation in clients.

An asynchronous local browser database for structured, larger data with transactions and indexes.

An asynchronous Python web framework that builds HTTP APIs and documentation from type hints and validation models.

The servlet-based Spring HTTP application layer for request mapping, validation, and response serialization.

Database access control that evaluates the current principal and row attributes before allowing reads or writes.

A testing tool that automates navigation, input, networking, and accessible UI behavior in real browser engines.

An integration-testing tool that starts real dependencies such as databases and brokers in temporary containers.

A desktop runtime for local models that can expose inference through an OpenAI-compatible HTTP interface.

A browser inference runtime for ONNX models using WebAssembly or WebGPU execution providers.

A user-interface library that synchronizes a component tree declared from state and props with the screen.

A web framework that combines React server and client components, routing, data access, builds, and deployment boundaries.

An in-memory data store for fast access to short-lived caches, sessions, rate limits, and distributed state.

A streaming platform that retains events in ordered partition logs so consumers can process them from independent positions.

The process of locating text in images or scanned documents and converting it into machine-readable characters.

A web server that automates TLS and reverse proxying while separating the public entry point from internal services.