feat(species): auto-discover species manifests + scaffold script

Registry
--------
src/domain/species/registry.ts previously hard-coded the species list,
which meant adding a new animal forced an edit in two places. The
registry now uses Vite's `import.meta.glob('./*/index.ts', { eager: true })`
and registers every manifest that default-exports a SpeciesManifest.
Duplicate ids log a warn and the first wins. Existing named imports
keep working — equine/index.ts adds `export default equine` next to the
named export.

The species dropdown is also now sorted alphabetically by labelFr (FR
locale) instead of insertion order, so two species don't shuffle
position based on glob ordering.

Scaffold script
---------------
`pnpm scaffold:species` (scripts/scaffold-species.mjs) is an
interactive Node ESM prompt that asks for the id, FR/EN labels, asset
folder, main SVG filename, detail SVG mappings, and the vertebral
formula, then writes:

  src/domain/species/<id>/index.ts        (with `export default`)
  src/domain/species/<id>/vertebrae.ts
  src/domain/species/<id>/pathologies.ts  (Zod-parses the JSON)
  src/domain/species/<id>/pathologies.json (minimal valid skeleton)
  public/species/<assetFolder>/           (directory placeholder)

The script refuses to overwrite existing files, so it's safe to rerun
after typos. SVG assets are not generated — drop them in the public
folder yourself.

Findings JSON updates
---------------------
No runtime hot-reload mechanism is needed in dev: `pathologies.ts`
imports `pathologies.json` and runs `SpeciesFindingsSchema.parse(data)`
at module load, so a JSON edit triggers Vite HMR + an immediate Zod
validation pass. In prod, rebuild and redeploy.

Docs
----
docs/ADDING-A-SPECIES.md rewritten around the new flow: quick start via
the scaffold script first, manual instructions second, and removed the
stale references to the flat `Pathology[]` model and the no-longer-
required registry edit step.

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Bastien ChanotandClaude committed 2026-05-20 05:46:33 +02:00
1 parent 700567fc4f
commit 70b7f20f6f
5 files changed
+432 -170

No files matched your search

+150 -163
View File
@@ -1,61 +1,113 @@
# Adding a New Species
This guide walks through every step required to add a new species (e.g. dog,
cat, bovine) to Osteo Report. The reference implementation is `equine` —
mirror its structure when adding a new species.
Osteo Report ships with one species (`equine`). Adding a new one is
mostly authoring work — anatomy SVGs and a findings JSON — backed by a
small amount of boilerplate that the project can scaffold for you.
Throughout this document, replace `<id>` with the new species' kebab-case id
(e.g. `canine`, `feline`, `bovine`).
> Throughout this document, replace `<id>` with the new species'
> kebab-case id (e.g. `canine`, `feline`, `bovine`).
---
## 1. Author the anatomy SVGs
## Quick start — `pnpm scaffold:species`
All SVG assets must conform to the conventions in
[`docs/ANATOMY-SVG-CONVENTION.md`](./ANATOMY-SVG-CONVENTION.md). The parser
in `src/anatomy/svgConvention.ts` reads the same attributes, so any drift
breaks the editor canvas.
Every clickable zone element must carry these data attributes:
| Attribute | Required | Notes |
|----------------------|----------|-----------------------------------------------------------------|
| `data-anno-id` | yes | Stable kebab-case id, unique within the SVG. |
| `data-anno-name-fr` | yes | French anatomical label (used in tooltips and PDF text). |
| `data-anno-name-en` | yes | English anatomical label (used as fallback and a11y hint). |
| `data-anno-category` | yes | One of: `articulation`, `charniere`, `spine-zone`, `os`, `os-overlay`, `head`, `cadran-crane`. |
| `data-anno-opens-detail` | no | Id of a detail popup SVG to open when the zone is clicked. The id must match a key in the species manifest's `details` map. |
Reuse the equine SVG patterns:
- **Hitbox + visible twin** — used for `articulation`, `charniere`, `os-overlay`, `head`.
- **Single visible element** — used for `spine-zone`, `os`, `cadran-crane`.
See `public/species/equine/Horse.svg` and `public/species/equine/Horse_head.svg`
for fully worked examples.
## 2. Drop the SVGs into `public/`
Put the assets at:
```
public/species/<id>/<Main>.svg # main side-view body schema
public/species/<id>/<Main>_<detail>.svg # one detail SVG per opens-detail key
```bash
pnpm scaffold:species
```
For example, for a dog with a single head detail popup:
The interactive prompt asks for:
- species id (kebab-case, e.g. `canine`),
- French and English display labels,
- the asset folder under `public/species/` (defaults to the id),
- the main SVG filename,
- zero or more detail-popup SVG paths (mapped by `data-anno-opens-detail` key),
- the vertebral formula (C / T / L / S / Cd).
It then writes:
```
public/species/canine/Dog.svg
public/species/canine/Dog_head.svg
src/domain/species/<id>/
index.ts ← SpeciesManifest with `export default`
vertebrae.ts ← VertebralFormula constant
pathologies.ts ← parses pathologies.json via SpeciesFindingsSchema
pathologies.json ← minimal valid catalog skeleton
public/species/<assetFolder>/ (directory created, empty)
```
The `public/` prefix is implicit at runtime — the manifest references files
by absolute path starting with `/species/...`.
The script refuses to clobber existing files, so it is safe to rerun.
## 3. Create the species manifest
After scaffolding, drop the anatomy SVGs into the `public/species/...`
folder, fill in `pathologies.json`, and the species shows up
automatically in the species picker — see "How auto-discovery works"
below.
Create `src/domain/species/<id>/index.ts` exporting a `SpeciesManifest`:
---
## How auto-discovery works
`src/domain/species/registry.ts` no longer hard-codes species. It uses
Vite's `import.meta.glob`:
```ts
const modules = import.meta.glob<{ default: SpeciesManifest }>(
'./*/index.ts',
{ eager: true },
)
```
Every `src/domain/species/<id>/index.ts` that **default-exports** a
`SpeciesManifest` is picked up at build time and registered by its
`manifest.id`. Duplicates are warned about and the first one wins.
What this means in practice:
- **No registry edit needed.** Drop a properly-structured species
folder and it appears in `AnimalForm`, `getSpecies(id)`, the editor,
and the PDF exporter.
- **The `id` must be unique.** Two folders default-exporting manifests
with the same `id` triggers a console warning and the later module
is ignored.
- **A missing default export is silent.** If you forget
`export default`, the species is invisible. The scaffold script
emits the default export for you.
---
## Manual setup (without the script)
The scaffold script just writes the files below. If you'd rather do it
by hand, the same workflow applies — make sure your `index.ts`
default-exports the manifest so auto-discovery sees it.
### 1. Author the anatomy SVGs
All SVG assets must conform to
[`docs/ANATOMY-SVG-CONVENTION.md`](./ANATOMY-SVG-CONVENTION.md). Every
clickable element must carry:
| Attribute | Required | Notes |
| -------------------------- | -------- | -------------------------------------------------------------------------------------- |
| `data-anno-id` | yes | Stable identifier, unique within the SVG. The renderer treats it as opaque. |
| `data-anno-name-fr` | yes | French anatomical label. |
| `data-anno-name-en` | yes | English anatomical label. |
| `data-anno-category` | yes | Free-form string used for grouping (e.g. `articulations`, `cadrans-du-cr-ne`). |
| `data-anno-opens-detail` | no | Key of a detail popup SVG. Must match a key in the manifest's `details` map. |
| `data-anno-special` | no | Behavior hint. `vertebral-precision` opens the VertebraPicker on click. |
Drop the SVGs into:
```
public/species/<assetFolder>/<Main>.svg
public/species/<assetFolder>/<Main>_<detail>.svg
```
The `public/` prefix is implicit at runtime — the manifest references
files by absolute path starting with `/species/...`.
### 2. Create the species manifest
`src/domain/species/<id>/index.ts`:
```ts
import type { SpeciesManifest } from '@/domain/species/types'
@@ -69,50 +121,23 @@ export const canine: SpeciesManifest = {
head: '/species/canine/Dog_head.svg',
},
vertebralFormula: CANINE_VERTEBRAL_FORMULA,
pathologies: () => import('./pathologies'),
findings: () => import('./pathologies'),
}
export default canine // <- required for auto-discovery
```
Required fields (see `src/domain/species/types.ts`):
`details` keys must match every distinct `data-anno-opens-detail`
attribute used in the main SVG. There is no filename convention — the
mapping is explicit through this object.
- `id` — must match the folder name and the registry key.
- `labelFr` — French display name shown in the species picker.
- `mainSvg` — absolute path to the main anatomy SVG.
- `details` — map of detail id → SVG path. Keys must match every distinct
`data-anno-opens-detail` attribute used in the main SVG.
- `vertebralFormula` — segment counts for the VertebraPicker (see step 5).
- `pathologies` — lazy import returning `{ default: Pathology[] }` (step 6).
### 3. Provide the vertebral formula
## 4. Register the manifest
Edit `src/domain/species/registry.ts` to add the new manifest:
```ts
import type { SpeciesManifest } from './types'
import { equine } from './equine'
import { canine } from './canine'
const SPECIES_REGISTRY: Record<string, SpeciesManifest> = {
[equine.id]: equine,
[canine.id]: canine,
}
```
The species will now appear in the species `<Select>` of `AnimalForm`,
and `getSpecies(id)` will resolve it everywhere it is consumed (editor,
PDF export, anatomy canvas).
## 5. Provide a vertebral formula
Create `src/domain/species/<id>/vertebrae.ts`:
`src/domain/species/<id>/vertebrae.ts`:
```ts
import type { VertebralFormula } from '@/domain/species/types'
/**
* Vertebral formula for the dog (Canis lupus familiaris).
* Cd count capped at 20 for the picker — actual count varies 6–23.
*/
export const CANINE_VERTEBRAL_FORMULA: VertebralFormula = {
C: 7,
T: 13,
@@ -122,114 +147,76 @@ export const CANINE_VERTEBRAL_FORMULA: VertebralFormula = {
}
```
The keys map to the cervical, thoracic, lumbar, sacral, and caudal counts
the `VertebraPicker` will offer when the practitioner clicks a `spine-zone`
on the main SVG. Cap `Cd` at a reasonable maximum for picker usability
when the real count varies between individuals.
Cap `Cd` at a sensible maximum (caudal counts vary heavily across
individuals).
Look up the reference counts in a veterinary anatomy textbook (König &
Liebich for dogs/cats, Dyce/Sack/Wensing for general comparative anatomy).
### 4. Provide the findings catalog
## 6. Provide a pathologies seed file
Two files:
Create `src/domain/species/<id>/pathologies.ts`:
`src/domain/species/<id>/pathologies.json` — authored tree of zones,
categories, sub-categories and leaves. The shape is defined by
`src/domain/schema/findings.ts`. The scaffold script ships a minimal
skeleton; populate the `zones` map by mirroring the equine reference at
`src/domain/species/equine/pathologies.json`.
`src/domain/species/<id>/pathologies.ts` — loader:
```ts
import type { Pathology } from '@/domain/schema/pathology'
import {
SpeciesFindingsSchema,
type SpeciesFindings,
} from '@/domain/schema/findings'
import data from './pathologies.json'
/**
* Canine pathology database.
* Per v1 contract, an empty array is acceptable — the practitioner can
* fall back to free-text pathology input until this is populated.
*/
const CANINE_PATHOLOGIES: Pathology[] = []
const FINDINGS: SpeciesFindings = SpeciesFindingsSchema.parse(data)
export default CANINE_PATHOLOGIES
export default FINDINGS
```
Each entry, when populated, must pass `PathologySchema.parse()` and have
at least one `sources` entry citing a peer-reviewed reference. See
`src/domain/schema/pathology.ts` for the full schema.
The Zod parse runs on first import and **fails fast** if the JSON
violates the schema (missing labels, malformed children, empty
disclaimer, etc.). That is your validation gate — there is no separate
lint step to run.
Empty arrays are allowed for v1 — the editor will simply show the
free-text pathology input only, exactly like equine ships today.
---
## 7. Add the two tests
## Updating the findings JSON
Mirror the equine reference tests:
The JSON is bundled into the app at build time, so updates flow
through normal dev/deploy paths:
### 7a. Registry test
- **In dev (`pnpm dev`)**: edit the JSON, Vite HMR reloads the module,
the editor picks up the new tree on the next interaction. A bad
edit fails loudly because the Zod parse runs at module load.
- **In prod**: rebuild and redeploy. The findings tree is statically
imported and there is no backend serving JSON at runtime.
Add to `tests/unit/species.registry.test.ts` (or create
`tests/unit/species.<id>Registry.test.ts` if you prefer one file per
species). Pattern:
Snapshots in already-exported PDFs are unaffected — observations carry
`labelFrSnapshot` / `labelEnSnapshot` copied at selection time, so
catalog edits cannot retroactively rewrite past reports.
```ts
it('returns the canine manifest', () => {
const c = getSpecies('canine')
expect(c).toBeDefined()
expect(c?.labelFr).toBe('Chien')
expect(c?.mainSvg).toBe('/species/canine/Dog.svg')
expect(c?.details.head).toBe('/species/canine/Dog_head.svg')
expect(c?.vertebralFormula).toEqual({ C: 7, T: 13, L: 7, S: 3, Cd: 20 })
})
```
---
### 7b. Real-SVG parse test
## How findings link back to clicked SVG zones
Mirror `tests/unit/anatomy.horse.test.ts`. Create
`tests/unit/anatomy.<id>.test.ts` that loads the real SVG from
`public/species/<id>/` and asserts:
The catalog keys observations by stable JSON zone ids (e.g.
`sphenoid`). SVG hitboxes carry opaque generator ids
(e.g. `point-mpbymcln`). The bridge is the **FR/EN label** parsed from
`data-anno-name-fr` / `data-anno-name-en`, matched against the JSON's
`zone.name_fr` / `zone.name_en` via the fuzzy matcher in
`src/lib/zoneMatch.ts`.
- `parseSvgBones` does not throw and returns a non-empty map.
- Every required `data-anno-category` you used is present at least once
(especially the `spine-zone` count, the `head` zone with `opensDetail`,
and any `charniere` hinges).
- Detail SVGs (one test per detail id) parse cleanly and contain the
expected `cadran-crane` zones.
Pattern:
```ts
import { readFileSync } from 'node:fs'
import { fileURLToPath } from 'node:url'
import { dirname, resolve } from 'node:path'
import { parseSvgBones } from '@/anatomy/svgConvention'
const HERE = dirname(fileURLToPath(import.meta.url))
const DOG_PATH = resolve(HERE, '../../public/species/canine/Dog.svg')
function loadSvg(path: string): SVGSVGElement {
const raw = readFileSync(path, 'utf8')
return new DOMParser()
.parseFromString(raw, 'image/svg+xml')
.documentElement as unknown as SVGSVGElement
}
describe('real Dog.svg', () => {
const map = parseSvgBones(loadSvg(DOG_PATH))
it('parses without throwing', () => {
expect(map.size).toBeGreaterThan(10)
})
it('contains all 5 spine-zone bones', () => {
const spine = [...map.values()].filter((b) => b.category === 'spine-zone')
expect(spine.length).toBe(5)
})
})
```
Also add a `tests/unit/species.<id>Pathologies.test.ts` that mirrors
`tests/unit/species.equinePathologies.test.ts` — even if the array is empty,
the schema parse loop and the `speciesIds.includes('<id>')` assertion become
guardrails for future contributors.
In practice you do not need to make the SVG ids meaningful; you just
need each zone's label to match its catalog entry under normalization
(lowercase + strip accents + drop non-alphanumerics) or be close
enough that Levenshtein finds it.
---
## CI gate before merging
```
```bash
pnpm lint
pnpm typecheck
pnpm test:run