Files
Ekklesia/docs/module-development/README.md
T
ekklesia 831a48187c
CI / verify (pull_request) Successful in 19s
CI / infrastructure-smoke (pull_request) Successful in 21s
CI / migration-smoke (pull_request) Failing after 16s
test: add PostgreSQL migration integration coverage
2026-07-21 09:37:15 -05:00

295 lines
12 KiB
Markdown

# Module development
An Ekklesia module is a backend capability package. It owns business behavior and data declarations;
it never owns components, styles, layouts, renderers, navigation, or other UI.
## Required package shape
Each immediate child of `modules/` must be an ESM workspace package with a unique package name, a
strict semantic version, and the module layer marker:
```text
modules/example/
├── drizzle/ # when the module has migration history
│ ├── 0000_initial.sql
│ └── meta/
│ ├── 0000_snapshot.json
│ ├── _ekklesia.json
│ └── _journal.json
├── dist/schema/index.js # when the module owns tables
├── dist/server/index.js
├── drizzle.config.ts # when generating migrations
├── ekklesia.module.json
├── package.json
├── src/schema/index.ts # when the module owns tables
├── src/schema/tables.ts # pure Drizzle declarations used by Drizzle Kit
└── src/server/index.ts
```
The schema and migration paths are optional. A tableless module omits `entrypoints.schema`; a
module with no migration history omits `entrypoints.migrations` and needs no placeholder files.
```json
{
"type": "module",
"ekklesia": {
"layer": "module"
}
}
```
The package version must match the manifest and runtime-definition versions. Internal package
dependencies use `workspace:` ranges and supported public package entrypoints.
## Canonical manifest
`ekklesia.module.json` is inert and authoritative, so Core can validate every installed module
without executing module code:
```json
{
"manifestVersion": 2,
"id": "calendar",
"name": "Calendar",
"version": "0.1.0",
"compatibility": {
"platform": "^0.1.0",
"moduleSdk": "^0.1.0"
},
"dependencies": {
"modules": {
"required": {
"people": "^0.1.0"
}
},
"capabilities": {
"required": {
"people.records": "^2.0.0"
},
"optional": {
"media.items": ">=1.1.0 <2.0.0"
}
}
},
"capabilities": [
{
"id": "calendar.events",
"version": "1.0.0"
}
],
"permissions": ["calendar.events.read"],
"storage": {
"buckets": [
{
"id": "assets",
"description": "Private event assets owned by Calendar."
}
]
},
"entrypoints": {
"server": "./dist/server/index.js",
"schema": "./dist/schema/index.js",
"migrations": "./drizzle"
}
}
```
Module, capability, and permission identifiers are lowercase and namespaced. A provided capability
has its own exact semantic version because its public contract can evolve independently from the
module release. Capability dependencies use semantic-version ranges. Optional means the provider
may be absent; if it is installed, its capability version must be compatible.
Server and optional schema entrypoints must be compiled ESM `.js` or `.mjs` files. The optional
migration entrypoint is a directory. All entrypoints begin with `./`, remain inside the package,
and exist. Production never imports source TypeScript directly. Modules with no tables omit
`entrypoints.schema`; modules with no migration history omit `entrypoints.migrations`. Unknown
manifest properties are rejected, including presentation fields such as components or renderers.
Storage bucket IDs are logical names scoped to the module. Modules may request multiple buckets.
Core will later map them to deployment-specific physical S3 buckets and provide scoped access;
module code must not configure SeaweedFS or another S3 client directly.
## Runtime server definition
The source entrypoint exports one typed runtime definition and does not repeat the static manifest:
```ts
import { defineModuleRuntime } from "@ekklesia/module-sdk";
export default defineModuleRuntime({
id: "calendar",
version: "0.1.0",
register({ capabilities }) {
capabilities.provide("calendar.events", calendarEvents);
},
async start({ capabilities }) {
const people = capabilities.resolve("people.records");
// Acquire runtime resources here.
},
async stop() {
// Release runtime resources here.
}
});
```
ID and version are intentional build-artifact identity markers and must match the manifest. Core
rejects missing, duplicate, or undeclared capability registrations. A module may resolve only its
own capabilities and the capability dependencies declared in its manifest. The runtime context is
deliberately limited to the module ID and capability access.
Entrypoint imports must have no top-level side effects. Modules are trusted backend plugins, and ESM
evaluation cannot be rolled back. Keep registration declarative; open resources in `start` and
close them in `stop`.
## Drizzle schema definition
A module with tables keeps the generator-facing declarations in `src/schema/tables.ts`:
```ts
import { integer, pgTableCreator, text } from "drizzle-orm/pg-core";
const moduleTable = pgTableCreator((localName) => `calendar_${localName}`);
export const events = moduleTable("events", {
id: integer("id").primaryKey(),
title: text("title").notNull()
});
```
The production schema entrypoint in `src/schema/index.ts` wraps those same objects with the Module
SDK ownership marker:
```ts
import { defineModuleSchema } from "@ekklesia/module-sdk";
import { events } from "./tables.js";
export { events } from "./tables.js";
export default defineModuleSchema({
moduleId: "calendar",
tables: { events }
});
```
The prefix callback converts the module ID to a snake-case namespace and prefixes every local table
name. Here the physical PostgreSQL name is `calendar_events`; `care-team` plus `members` becomes
`care_team_members`. Local names must be lowercase snake case, and the complete physical identifier
must not exceed 63 characters. The SDK's `createModuleTable` provides the same convenience when a
generator-facing pure declaration file is not needed; Core validates built metadata independently
of either helper.
Schema definitions use PostgreSQL tables in the platform-managed default/public schema. They must
be side-effect-free: do not connect to a database, read configuration, or perform I/O during import.
The `moduleId` is an ownership marker and must match the static manifest. Export physical base
tables, never query aliases.
Foreign keys may reference self or another exported table from the same module. They may not target
another module or an omitted private table. Every source column must belong to the declaring table,
and all columns in a compound target must belong to the same exported target table. Modules must
never import another module's tables, columns, relations, schema entrypoint, or migration files.
Store opaque identifiers and integrate through public capabilities, events, or APIs instead.
## Drizzle migrations
Module migrations use the stable Drizzle Kit `0.31.10` PostgreSQL format. Pin that exact version in
the module, set `migrations.prefix` to `index`, and generate from the module's own source schema:
```ts
import { defineConfig } from "drizzle-kit";
export default defineConfig({
dialect: "postgresql",
schema: "./src/schema/tables.ts",
out: "./drizzle",
migrations: { prefix: "index" },
breakpoints: true
});
```
Keep `tables.ts` limited to Drizzle declarations so the stable generator can load it without
executing the ESM Module SDK wrapper. `src/schema/index.ts` remains the production compiled schema
entrypoint and wraps those same exported table objects with `defineModuleSchema`; tables are not
declared twice.
Run the module's `migration:generate` script with a lowercase snake-case name, review the generated
SQL and snapshot, and add the matching entry to `drizzle/meta/_ekklesia.json`. Every entry must
explicitly declare either `{"mode":"none"}` or
`{"mode":"approval-required","reason":"..."}`. Do not add an `approved` flag; approval will be
supplied by an external Core-owned provider and bound to module ID, identity, and checksum.
Migration identities are `<module-id>:<generated-tag>`. Core preserves Drizzle journal order inside
each module and dependency-first discovery order between modules. It calculates SHA-256 from the
exact SQL bytes, so changing whitespace or line endings changes the checksum. Never edit an
already-applied migration: Core compares the current exact checksum and metadata with its fixed
PostgreSQL history and refuses changed, missing, unknown, or out-of-prefix entries. Add a new forward
migration instead. Downgrade and uninstall migrations are not supported.
A module may migrate only its own objects. SQL is intentionally treated as opaque: offline
validation checks the artifact and metadata contract but cannot prove SQL ownership or classify
destructive behavior. Review those properties explicitly. Core is the only migration executor. It
follows Drizzle's exact `--> statement-breakpoint` delimiter and never naively splits semicolons or
falls back to non-transactional execution.
## Discovery and lifecycle
From the repository root, run:
```bash
pnpm install
pnpm modules:validate
pnpm migrations:validate
pnpm migrations:plan
pnpm verify
```
`pnpm modules:validate` builds the workspace and then atomically validates every installed module.
It rejects malformed manifests, duplicate IDs or package names, version mismatches, incompatible
platform or SDK ranges, missing required modules or capabilities, incompatible installed
dependencies, cycles, and missing, escaping, or non-JavaScript server or schema entrypoints.
Missing optional dependencies are allowed. Valid modules are ordered dependency-first with stable
identifier sorting.
After complete static discovery succeeds, validation loads optional schema entrypoints
dependency-first and rejects invalid definitions, unnamespaced or duplicate tables, custom
PostgreSQL schemas, and foreign keys outside the owning module. The aggregate is published only when
every schema is valid. This slice does not connect to PostgreSQL or execute migrations.
`pnpm migrations:validate` performs complete static discovery first, then reads every declared
migration directory without importing module code or connecting to PostgreSQL. It validates native
Drizzle journals and snapshot chains, Ekklesia policy metadata, ordering, inventory, paths, and
checksums atomically. Its concise report lists every installed module and migration in deterministic
order.
With `DATABASE_URL` set through the uncommitted root `.env`, `pnpm migrations:plan` performs a
read-only applied-history comparison. The plan can become stale. `pnpm migrations:apply` is a
separate, explicit database mutation: Core acquires its bounded session advisory lock, rereads and
validates history, preflights the whole destructive-approval set, and applies all pending modules in
dependency-first order in one transaction. If a later migration fails, every schema and history
change from that invocation rolls back; migrations committed by an earlier invocation remain.
An `approval-required` migration needs an approval from an external provider matching its module
ID, exact identity, and exact checksum. Module metadata cannot grant approval, and there is no
wildcard, force, or approve-all option. The default developer command supplies no approvals and
therefore refuses destructive pending migrations.
Runner behavior is unit-tested through a narrow database port. The separate
`pnpm test:migrations:integration` command now exercises the production adapter against disposable
PostgreSQL, including DDL rollback, advisory-lock contention, history behavior, approvals, and
sanitized failures. It remains outside the Docker-free `pnpm verify` gate; see the
[testing guide](../development/testing.md).
Core completes discovery before importing any server entrypoint. It then loads and verifies every
runtime definition before running hooks, registers all modules dependency-first into an unpublished
staged capability registry, and starts them in the same order. Shutdown is reverse-order. If a later
module fails to start, Core stops already-started modules in reverse order and reports any cleanup
failures without leaking caught exceptions or machine paths. A failing `start` hook must clean up
its own partial work before rejecting.
## Start from the template
Copy `templates/module` to `modules/<your-module-id>` and follow its README. No central registry edit
is needed: the workspace glob, build, discovery, and runtime loading all find the new package.