harper-component-model

name	harper-component-model
description	reasoning about how a Harper…

Harper Component Model

Overview

Harper (formerly HarperDB; the product and company now at harper.fast) is an open-source Node.js platform that fuses database, cache, application logic, and messaging into a single in-memory process. Fabric is Harper's distributed deploy network: you develop locally, then deploy the same component to Fabric.

Everything you build is a component. Understanding the component hierarchy is the prerequisite for every other Harper decision — config, resources, schema, and deploy all hang off it.

The three tiers

Harper organizes functionality into three tiers, top to bottom:

1. Applications — the user-facing product. An application is a type of component. It implements business logic (REST/GraphQL endpoints, web UI, real-time) by depending on extensions/plugins. This is what this project is.
2. Plugins (v5) / Extensions (deprecated) — the building blocks an application depends on. A plugin exports a single handleApplication(scope) method and always runs on worker threads. The deprecated Extension API used start, handleFile, handleDirectory, and setupDirectory instead. handleApplication() cannot coexist with Extension API methods — defining both throws. Prefer the Plugin API for any custom building block.
3. Core services — the high-performance database, networking middleware, and the component manager. You configure these; you don't reimplement them.

An application is the component you ship. Extensions/plugins are the capabilities it consumes. Built-in extensions (graphqlSchema, jsResource, rest, static, …) are provided by core; you only enable them in config.yaml. See [[harper-config-yaml]].

Built-in extensions an application typically uses

These ship with Harper and are enabled (not installed) via config.yaml:

• graphqlSchema — define database tables/types from GraphQL schema files. See [[harper-schema-graphql]].
• jsResource — load custom JavaScript resources (resources.js). See [[harper-resources]].
• rest — auto-generate REST endpoints for exported resources/tables.
• static — serve static files (the web/** directory) over HTTP.
• roles — role-based access control from roles.yaml.
• loadEnv — load environment variables from .env.
• dataLoader — seed Harper tables from JSON/YAML.
• fastifyRoutes — custom Fastify route handlers.

Where things live in this project

This project wraps Harper's native model with a fixed layout under harper-app/:

Path	Role	Source or generated
harper-app/config.yaml	Component config — which extensions are active	Source
harper-app/schema.graphql	Table/type definitions	Source
src/** (TypeScript)	Resources, browser modules, shared libs, scripts	Source
harper-app/resources.js	Loaded by jsResource	Generated — never edit; build from TS
harper-app/web/**	Served by static	Generated — never edit; build from TS

The TypeScript under src/ is the source of truth. resources.js and web/** are deploy artifacts produced by bun run build. Never hand-edit them — change the matching TypeScript and rebuild. See [[harper-build-and-deploy]].

Decision guide

• Adding a backend behavior? It's almost always a resource (custom logic) or a schema change (new table/field), enabled through config.yaml. Don't ship a client-side workaround for missing backend behavior — make the Harper change.
• Adding a reusable building block / npm-publishable capability? That's a plugin (pluginModule), not application code.
• Adding static UI? It belongs in web/** (generated from src/ UI code), served by the static extension.
• Not sure if it deploys? A deployable Harper app must keep config.yaml, schema.graphql, resources.js, and web/** at the component root that Fabric packages. If your change touches that surface, build before packaging.

Sources

• Components overview
• Plugin API
• Applications
• Harper platform