Merge remote-tracking branch 'origin/edge' into docs/spawn-atlas

Follows the redesign in website #112. Two decisions from the original §6 were
rejected in review and replaced; the docs now describe what was actually built.

**The committed artifact is gone.** A shard's maps change over its life, so a
snapshot in the repo silently drifts from the world players actually see. The
ServUO tree is the single source of truth and the atlas is re-derived on every
server boot, hash-gated so an unchanged tree costs one read pass and no write.

**Nothing may name a facet.** The first implementation carried a lookup table of
the six stock UO facets. A shard may add facets, replace them outright, or rename
them when its maps are updated, and a built-in list mishandles all three
silently. Reconciliation is now by matching against the facet set discovered from
the shard's own data, with an unmatched name keeping its own rather than being
forced into a wrong bucket.

## Changes

- **`website/SPAWN_ATLAS.md`** rewritten: the two ideas that shape the design,
  how to configure the tree path, the boot flow as a decision tree, the
  approve/reject flow, and the code layout. The artwork policy is unchanged and
  still explicit — no art ever ships, operators extract their own from their own
  client files.
- **`link/v3.md` §6.1 (new)** records the two rejected decisions plus the two
  boot-path contracts. The old "what real data changed" list becomes §6.2. §6's
  now-superseded passages — the artifact bullet, the payload budget, the operator
  re-run story — are marked rather than deleted, so the reasoning stays legible.
- **`website/BACKEND_DESIGN.md`** documents `shard_atlas_pending` and the two
  contracts that make it safe: a facet removal is staged for a human, and the
  boot refresh can never block startup.

The two contracts are the part worth reviewing. Losing a facet is
indistinguishable at boot from a half-copied or mid-update tree, so it is staged
rather than applied; and no failure mode of the atlas — missing path, unreadable
mount, malformed file, database error — is allowed to stop the site coming up.

---

- [x] AI-assisted: written with **Claude Code** (Claude Opus 5), reviewed before opening.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01U7CBg11prhLimL9iHSX1bP

link/v3.md

@@ -13,7 +13,7 @@ Each part is marked off here as it lands on `edge`. §9 carries the same state p
 |---|---|---|---|
 | 1 | **A** — visibility framework + actor-leak fix (§3) | ✅ **Done** | website [#109](https://gitea.whitlocktech.com/RunicGateway/website/pulls/109) + [#110](https://gitea.whitlocktech.com/RunicGateway/website/pulls/110), docs [#64](https://gitea.whitlocktech.com/RunicGateway/docs/pulls/64) + [#65](https://gitea.whitlocktech.com/RunicGateway/docs/pulls/65) |
 | 2 | **B/1** — `world.ruleset` (§5) | ✅ **Done** | servuo-plugins [#3](https://gitea.whitlocktech.com/RunicGateway/servuo-plugins/pulls/3), link [#17](https://gitea.whitlocktech.com/RunicGateway/link/pulls/17), website [#111](https://gitea.whitlocktech.com/RunicGateway/website/pulls/111), docs [#66](https://gitea.whitlocktech.com/RunicGateway/docs/pulls/66) |
-| 3 | **C** — spawn atlas (§6) | ⬜ **Next** | — |
+| 3 | **C** — spawn atlas (§6) | 🟡 **Data pipeline done** | website [#112](https://gitea.whitlocktech.com/RunicGateway/website/pulls/112) (parsers + CLI + tables); API/client PR next |
 | 4 | **B/2** — `points.board` (§7) | ⬜ Not started | — |
 | 5 | **B/3** — `vendor.listing` (§8) | ⬜ Not started | — |
 | 6 | **Cutover** — `PROTOCOL_VERSION` 2→3 (§4) | ⬜ Not started | — |
@@ -329,19 +329,31 @@ frame during verification.
 
 **No plugin, no sidecar, no `Bridge.cfg` knob, no new kinds.** Not part of the v3 wire change.
 
-**Decision: committed generated artifact + idempotent DB import**, split in two because the build
-needs the ServUO tree (which the website container does not have) and the import does not. Not
-runtime import (10.5 MB of XML per boot), not a browser-served blob.
+> **Status:** data pipeline landed on `edge` — website [#112](https://gitea.whitlocktech.com/RunicGateway/website/pulls/112)
+> (parsers, build/import CLI, tables, artifact). API + client pages are the second website PR.
+> Part C ships as **two** website PRs, not one: the parsing half is where the correctness risk
+> lives, and burying it under routes and React would have meant reviewing it in a 10k-line diff.
+> Full operator documentation: [`docs/website/SPAWN_ATLAS.md`](../website/SPAWN_ATLAS.md).
+>
+> **§6 below is the original design and is partly superseded.** §6.1 records two decisions that were
+> rejected in review and replaced (the committed artifact, and the fixed facet list); §6.2 records
+> the corrections the real ServUO data forced. Read both before trusting §6.
+
+**Decision (revised at implementation time): the shard's ServUO tree is the single source of truth,
+re-derived on every server boot.** The original plan here was a committed generated artifact plus an
+idempotent import. That was rejected in review for two reasons, recorded in §6.1: a snapshot in the
+repo goes stale as a shard's maps change, and the design leaned on a fixed facet list that no shard
+is obliged to keep. Still not a browser-served blob; still parsed server-side only.
 
 New in `website/server/`:
 
 - `src/utils/spawnAtlasParse.js` — **pure functions, no fs**, so they are unit-testable in CI without
   a ServUO tree: `parseObjects2()`, `parsePoints()`, `parseRegions()`, `parseLocations()`,
   `resolveRegion()`.
-- `scripts/buildSpawnAtlas.js` (`--servuo <path> --out db/data/`) and `scripts/importSpawnAtlas.js`
-  (TRUNCATE + batched INSERT in one transaction); `package.json` scripts `atlas:build`, `atlas:import`.
-- `db/data/spawnAtlas.<facet>.json` ×13 + `spawnAtlas.index.json` (creatures, champions, regions,
-  landmarks, meta with per-source-file hashes).
+- ~~`scripts/buildSpawnAtlas.js` and a committed `db/data/spawnAtlas.*.json` artifact~~ — dropped,
+  see §6.1 R1. Replaced by `src/utils/spawnAtlasSource.js` (the only thing that reads a ServUO tree,
+  shared by the boot path and the CLI) and a `scripts/importSpawnAtlas.js` that is a thin CLI over
+  the model. `package.json` gains `atlas:import` only.
 - `src/model/shardAtlas/{shardAtlas.db.js,shardAtlas.model.js}` following the `shardState` split.
 - `src/router/v1/public/atlas.{router,controller}.js`; `test/spawnAtlas.parse.test.js`.
 
@@ -372,16 +384,92 @@ stays CLI-only.**
 
 Client: `routes/public/Atlas.jsx` (`/site/atlas`) and `AtlasCreature.jsx` (`/site/atlas/:slug`).
 
-**Payload risk** — a monolithic artifact would be 2–3 MB of committed JSON. Shard per facet and drop
-every `<Points>` field the site cannot use (`UniqueId`, all trigger/refractory/proximity/sequential
-fields, sound ids), keeping Name/Map/X/Y/W/H/Range/MaxCount/MinDelay/MaxDelay/TOD*/types — well under
-1 MB. The artifact never reaches the browser; the browser sees only paginated API responses.
+**Payload risk** — *superseded by §6.1 R1; nothing is committed.* The field selection it describes
+still applies at parse time: every `<Points>` field the site cannot use (`UniqueId`, all
+trigger/refractory/proximity/sequential fields, sound ids) is dropped, keeping
+Name/Map/X/Y/W/H/Range/MaxCount/MinDelay/MaxDelay/TOD*/types. Parsed data never reaches the browser;
+the browser sees only paginated API responses.
 
-**Operator re-run story** — spawns changed → `npm run atlas:build -- --servuo <path>` on a machine
-with the tree → commit the regenerated `db/data/spawnAtlas.*.json` → deploy → `npm run atlas:import`
-(or `POST /admin/shard/atlas/import`). `shard_atlas_meta.source` holds per-file hashes, so
-`GET /admin/shard/atlas/status` reports when the DB is behind the artifact. Full detail in
-`docs/website/SPAWN_ATLAS.md`.
+**Operator re-run story** — *revised by §6.1 R1.* Spawns changed → restart, or
+`npm run atlas:import` / `POST /admin/shard/atlas/import` to apply without one. `shard_atlas_meta`
+holds a sha256 per source file, so the server can tell on boot whether anything changed, and
+`GET /admin/shard/atlas/status` reports drift. If the change would remove a facet it is staged for
+approval rather than applied (§6.1 R3). Full detail in `docs/website/SPAWN_ATLAS.md`.
+
+### 6.1 What implementation changed
+
+Two design decisions in §6 were rejected in review and replaced; the rest are corrections the real
+ServUO data forced. Kept as a diff rather than edited in place, because each is a trap the next
+person would otherwise re-enter.
+
+**R1. The committed artifact is gone — the tree is re-parsed on every boot.** §6 proposed building a
+generated artifact, committing it, and importing it. Two problems. A shard's maps change over its
+life, so a snapshot in the repo silently drifts from the world players actually see; and the build/
+import split existed only to work around the website container not having a tree, which is a
+deployment question (mount it) rather than a reason to freeze data. The server now hashes the source
+files on boot and re-derives the atlas when they differ. `scripts/buildSpawnAtlas.js`, the 1.41 MB
+artifact, and the whole encode/decode seam it needed are deleted.
+
+**R2. Nothing may name a facet.** The first implementation carried a lookup table of the six stock
+UO facets to reconcile the spelling drift between sources. A shard may add facets, replace them
+outright, or rename them when its maps are updated, and a built-in list mishandles all three
+silently. Reconciliation is now by *matching* against the facet set discovered from the shard's own
+spawn and region data — exact key, then prefix in either direction — with an unmatched name keeping
+its own rather than being forced into a wrong bucket.
+
+**R3. Two contracts on the boot path.** It never blocks startup: no path, an unreadable mount, a
+malformed file or a database error is caught and logged, and the site comes up serving whatever
+atlas it had. And a refresh that would REMOVE a facet is never applied automatically — facet loss
+is indistinguishable at boot from a half-copied or mid-update tree, so it is staged in
+`shard_atlas_pending` for an admin to approve or reject. Only the decision is stored (source hashes
++ the facet diff, a few KB); approving re-parses, so what lands matches the tree at approval time.
+A rejection is remembered against those hashes so it does not re-prompt every restart.
+
+### 6.2 What the build against real data changed
+
+Six corrections to the design above, from running it against stock ServUO 57.4. Kept as a diff
+rather than edited in place, because each one is a trap the next person would otherwise re-enter.
+
+**1. Six facets, not thirteen.** The design said `spawnAtlas.<facet>.json ×13`, assuming one facet
+per spawn file. There are 13 files but only **6** facets — `Eodon.xml`, `GravewaterLake.xml`,
+`TreasuresOfKotl.xml` and the other named-area files carry TerMur/Trammel points. The facet comes
+from each record's own `<Map>`, never the file name, and the artifact shards 6 ways.
+
+**2. The XML dependency call: hand-rolled, zero deps.** §6 left `fast-xml-parser` vs a ~120-line
+tokenizer open. Resolved as the tokenizer — a deliberate *subset* parser covering only what these
+files use. The server keeps zero XML dependencies at any tier.
+
+**3. Facet names disagree between sources — a silent failure.** `Data/Locations/*.xml` spells them
+`Ter Mur` and `Tokuno Islands`; `<Map>` and `<Facet name>` say `TerMur` and `Tokuno`. Unreconciled,
+the landmark bucket is keyed differently from the points looking it up, so the fallback never fires
+and **every unregioned spawn in Ter Mur and Tokuno reads "Wilderness"** — a plausible-looking atlas
+that is quietly wrong for two facets. All facet names now pass through `normalizeFacet()`.
+
+**4. Spawn type tokens carry XmlSpawner directives.** `<Objects2>` types are not always bare class
+names: `Fairy,{RND,4,8}`, `alchemist/z/-50`, `Agralem/Name/Agralem`, `greatape,true`. Taken literally
+they invent creatures that do not exist *and* split real ones in two, since `Fairy` and
+`Fairy,{RND,4,8}` slug apart. 71 of 845 entries were affected; stripping at the first `/` or `,`
+leaves **800** real creatures. (The design's "~1,500 creature rows" estimate was high; 800 only
+reinforces the plain-`INDEX`-not-`FULLTEXT` call.)
+
+**5. The artifact would have been 1.41 MB, not "well under 1 MB" — and is now moot.** Dropping the
+unused `<Points>` fields as the design directed still left 4.40 MB; three further encodings brought
+it to 1.41 MB, and getting under 1 MB would have meant dropping the spawner `name`. The size budget
+in §6 was simply optimistic for 6,455 points. Superseded by §6.1 R1: there is no artifact, so there
+is no payload to budget and no encode/decode seam to keep in sync.
+
+**6. `DELETE`, not `TRUNCATE`.** The design said "TRUNCATE + batched INSERT in one transaction",
+which does not hold: `TRUNCATE` is DDL in MariaDB and implicitly commits, so a mid-import failure
+would leave the atlas half-loaded. `DELETE` is transactional, and at ~7k rows the cost is
+irrelevant. Point ids are also assigned explicitly rather than by `AUTO_INCREMENT`, because the
+join rows need them and `conn.batch()` reports no usable `insertId`.
+
+**Measured result:** 6,455 points, 800 creatures, 23,927 point/type rows, 387 regions, 558
+landmarks, 25 champion altars. The placement transform resolves **83.2%** of points (3,689 by
+region, 1,690 by landmark, 1,086 Wilderness).
+
+**One thing the design got exactly right:** the point-in-rect transform really is the reason to
+build this. "Where does a lizardman spawn?" answers *Shrines, Isamu-Jima, Yew* across three facets.
 
 ---
 
@@ -577,7 +665,7 @@ inherently up to one full cycle old, and the UI must say so.
 |---|---|---|---|---|
 | 1 | **A** — visibility framework + actor-leak fix | website, docs | none | ✅ Done |
 | 2 | **B/1** — `world.ruleset` (§5) | all four | new kind | ✅ Done |
-| 3 | **C** — spawn atlas (§6) | website, docs | none | ⬜ **Next** |
+| 3 | **C** — spawn atlas (§6) | website, docs | none | 🟡 Pipeline done, API/client next |
 | 4 | **B/2** — `points.board` (§7) | all four | new kind + `char.profile` field | ⬜ |
 | 5 | **B/3** — `vendor.listing` (§8) | all four | new kinds | ⬜ |
 | 6 | **Cutover** — `PROTOCOL_VERSION` 2→3, `edge` → `main` | all four | the bump | ⬜ |

website/BACKEND_DESIGN.md

@@ -388,6 +388,67 @@ feature is ignored (a stale row must not resurrect a removed feature), an invali
 the default rather than failing open, and a rule touching a locked field (`acct` / `webId`) is
 discarded. See §6.5.
 
+### shard_spawn_* / shard_regions / shard_landmarks / shard_champion_spawns / shard_atlas_meta — the spawn atlas (Protocol 3.0)
+
+Static shard **content**, not live shard state. Nothing here comes from the sidecar: the atlas is
+derived from the shard's own ServUO tree, re-read on **every server boot** and hash-gated so an
+unchanged tree costs one read pass and no write. Nothing is precomputed and committed — a shard's
+maps change over its life, and a snapshot in the repo would silently drift from the world players
+actually see. These tables stay populated whether the shard is up or not. Full operator detail in
+[`SPAWN_ATLAS.md`](SPAWN_ATLAS.md); the design is `docs/link/v3.md` §6.
+
+**No facet name appears anywhere in the code.** A shard may add facets, replace them, or rename them
+when its maps are updated; the facet set is discovered from the tree, and the loose spellings in
+`Data/Locations` are matched against it rather than looked up in a table.
+
+| Table | Key columns |
+|---|---|
+| `shard_spawn_creatures` | `slug` PK, `name`, `total`, `points`, `facets` JSON, `art` NULL |
+| `shard_spawn_points` | `id` PK, `facet`, `name`, `x`, `y`, `width`, `height`, `spawn_range`, `max_count`, `min_delay`, `max_delay`, `tod_start/end/mode`, `region`, `landmark`, `label` |
+| `shard_spawn_point_types` | `(point_id, slug)` PK, `max_count` |
+| `shard_regions` | `facet`, `name`, `type`, `priority`, `parent`, `rects` JSON |
+| `shard_landmarks` | `facet`, `name`, `grp`, `x`, `y`, `z` |
+| `shard_champion_spawns` | `slug` PK, `name`, `grp`, `type`, `random_type`, `facet`, `x`, `y`, `z`, `radius`, `label` |
+| `shard_atlas_meta` | Singleton (`id = 1`), `payload` JSON (counts + a sha256 per source file), `imported_at` |
+| `shard_atlas_pending` | Singleton (`id = 1`), `status` (`pending`/`rejected`), `payload` JSON, `detected_at` |
+
+The first seven are **import-owned**: a refresh empties and reloads every one inside a single
+transaction, so a failed reload leaves the previous atlas intact rather than a half-loaded world.
+Nothing else writes to them, and nothing holds a foreign key to them — no FKs at all, consistent with
+every other `shard_*` table.
+
+**`shard_atlas_pending` is the security-relevant one.** A refresh that would REMOVE a facet is never
+applied automatically: facet loss is indistinguishable at boot from a half-copied or mid-update tree,
+so it is staged here for an admin to approve or reject, and **startup is never blocked by it**. Only
+the decision is stored — source hashes plus the facet diff, a few KB — and approving re-parses the
+tree, so a multi-megabyte blob never lands in the database and what gets applied matches the tree at
+approval time. A rejection is remembered against those exact hashes so a declined refresh does not
+re-prompt on every restart. Everything else (new facets, renamed regions, changed spawns) applies
+immediately, since none of it can destroy data an operator would miss.
+
+The boot refresh is **best-effort by contract**: no configured path, an unreadable mount, a malformed
+file or a database error is caught and logged, and the site comes up serving whatever atlas it had.
+The tree path comes from the `spawn_atlas_servuo_path` setting, falling back to `SERVUO_PATH`.
+
+Four column choices worth stating, because each one is a trap:
+
+- **`spawn_range`, not `range`**, and **`grp`, not `group`** — both are reserved words.
+- **`DELETE`, not `TRUNCATE`.** `TRUNCATE` is DDL in MariaDB and implicitly commits, which would
+  defeat the all-or-nothing reload. At ~7k rows the difference does not matter.
+- **Point ids are assigned explicitly**, not left to `AUTO_INCREMENT`: the `shard_spawn_point_types`
+  rows need to know them, and `conn.batch()` reports no usable `insertId` for a multi-row insert.
+- **Plain `INDEX` on `name`, deliberately not `FULLTEXT`.** ~800 creature rows makes a `LIKE` scan
+  free, and FULLTEXT's minimum token length would break searches for names like "orc".
+
+`shard_champion_spawns` is the *configured* altar roster ("there is an Unholy Terror altar in
+Deceit"). The live `champ.update` feed in `shard_champs` is the separate answer to "it is on level 3
+right now". Both exist; they are not the same data.
+
+**`shard_spawn_creatures.art` is always NULL on a fresh import.** The project ships no creature
+artwork: sprites live in the operator's own client `.mul`/`.uop` files and are theirs, not ours to
+redistribute. An operator supplies art via a gitignored map plus images under the (already
+gitignored) `server/uploads/atlas/`. Text-only is the normal, supported state.
+
 ---
 
 ## 4. API contract

website/SPAWN_ATLAS.md

@@ -0,0 +1,261 @@
+# Spawn atlas
+
+**Status:** Data pipeline landed on `edge` (website [#112](https://gitea.whitlocktech.com/RunicGateway/website/pulls/112)); API and client pages follow in a second PR.
+**Design:** [`docs/link/v3.md` §6](../link/v3.md) — Protocol 3.0 Part C.
+
+The spawn atlas is a browsable catalogue of what the shard *contains*: which
+creatures spawn, where, how many, and which champion altars are configured. It
+answers "where do I find a lizardman?" with **"Shrines, Yew, Isamu-Jima"** rather
+than with a list of raw coordinates.
+
+## Two things that shape the whole design
+
+**The shard's ServUO tree is the single source of truth.** Nothing is
+precomputed and committed to the repository. A shard's maps change over its
+lifetime — facets get added, replaced, or renamed — and a snapshot in the repo
+would silently drift from the world players actually see. The atlas is therefore
+re-derived from the tree **on every server boot**.
+
+**Facets are not a fixed list.** Nothing in the codebase names Felucca, Trammel,
+or any other stock facet. The facet set is whatever the shard's own files
+declare, discovered at parse time. A shard running entirely custom maps gets
+exactly the same treatment as a stock one, with no code change.
+
+## What it is not
+
+The atlas is **static shard content, not live shard state.**
+
+- It does **not** come from the sidecar. Nothing here touches the bridge, and
+  there is no event kind, no wire change and no `PROTOCOL_VERSION` bump for it.
+  Part C is website-only.
+- It stays fully populated while the shard is down.
+- Its champion table (`shard_champion_spawns`) is the *configured roster* —
+  "there is an Unholy Terror altar in Deceit". The live `champ.update` feed in
+  `shard_champs` is the separate, sidecar-fed answer to "it is on level 3 right
+  now". Both exist; do not conflate them.
+
+Routes live at `/api/v1/public/atlas`, deliberately **not** under `/shard`,
+because `/shard/*` means sidecar-dependent.
+
+## Configuring the tree
+
+The website needs to be able to *read* the ServUO tree — same host, a bind mount,
+or a shared volume. Two ways to point at it, the setting winning over the
+environment:
+
+| Source | Notes |
+|---|---|
+| `spawn_atlas_servuo_path` setting | Admin-editable; changes take effect on the next refresh without a redeploy |
+| `SERVUO_PATH` env var | The deploy-time default, since the path usually describes a mount the deployment sets up |
+
+With neither set the atlas is simply skipped — the site runs normally without
+one.
+
+## The boot path
+
+On every start the server hashes the source files and compares them against what
+is loaded. Unchanged (the normal case on a restart) costs one read pass, ~120 ms,
+and no database write. A real change costs a ~400 ms parse and a reload.
+
+Two contracts govern it:
+
+**1. It never blocks startup.** No configured path, an unreadable mount, a
+malformed file, a database error — every one is caught and logged, and the site
+comes up serving whatever atlas it already had.
+
+**2. A facet disappearing is never applied automatically.** Losing a facet looks
+exactly like a half-copied or mid-update tree, and boot cannot tell that apart
+from a real map change. That refresh is *staged* for a human instead. Everything
+else — new facets, renamed regions, changed spawns — applies immediately, since
+none of it can destroy something an operator would miss.
+
+```
+boot
+ └─ path configured?            no ──▶ skip
+     └─ tree readable?          no ──▶ warn, carry on
+         └─ hashes changed?     no ──▶ done (nothing parsed)
+             └─ parse
+                 └─ a facet would be removed?
+                      no ──▶ import
+                     yes ──▶ stage for admin review; atlas unchanged
+```
+
+### Approving or rejecting a staged refresh
+
+Only the *decision* is stored, never the parsed world — a few KB of source hashes
+plus the facet diff. Approving **re-parses** the tree, so what lands matches the
+tree at approval time rather than at boot, and a multi-megabyte blob never sits
+in the database.
+
+A rejection is remembered against those exact source hashes, so a declined
+refresh does not re-prompt on every restart. Change the tree and the hashes
+differ, which asks again.
+
+From the admin panel (second PR), or from the CLI:
+
+```bash
+cd website/server
+npm run atlas:import -- --status     # what is loaded, and what is pending
+npm run atlas:import -- --approve    # apply the staged refresh
+npm run atlas:import -- --reject     # keep the current atlas, dismiss it
+```
+
+## The CLI
+
+The server refreshes itself on boot, so this is for applying a map change
+*without* a restart, and for the approve/reject flow above.
+
+```bash
+npm run atlas:import                       # import if the tree differs
+npm run atlas:import -- --servuo <path>    # override the path for this run
+npm run atlas:import -- --force            # reimport even if unchanged
+```
+
+`--servuo` is a per-run override and deliberately does **not** persist — changing
+where the atlas permanently reads from is an admin action, not a side effect of a
+one-off import.
+
+## Sources
+
+| File | Count (stock ServUO 57.4) | Used for |
+|---|---|---|
+| `Spawns/*.xml` | 13 files, ~10.5 MB | Every spawner: location, size, delays, time-of-day, creature types |
+| `Data/Regions.xml` | 129 KB, nested | Named regions and their rectangles |
+| `Data/Locations/*.xml` | 6 files | Landmarks (dungeon levels, town markers) |
+| `Config/ChampionSpawns.xml` | 4.8 KB | Configured champion altars |
+
+**A stock tree has 13 spawn files but only 6 facets.** `Eodon.xml`,
+`GravewaterLake.xml`, `TreasuresOfKotl.xml` and the other named-area files hold
+TerMur/Trammel points. The facet always comes from each record's own `<Map>`,
+never from the file name.
+
+## How a coordinate becomes a place name
+
+This is the transform the atlas exists for, in `resolveRegion()`:
+
+1. The highest-`priority` named region whose rectangle contains the point. Ties
+   break toward the **smallest** rect, so a specific room wins over the
+   dungeon-wide rect enclosing it.
+2. Otherwise the nearest landmark within the landmark radius (200 tiles by
+   default), labelled by its **group** ("Covetous"), not its individual marker
+   ("Level 1").
+3. Otherwise `"Wilderness"`.
+
+The radius cap in step 2 is what keeps step 3 reachable. Without it the nearest
+landmark is always *some* landmark however far away, and open countryside gets
+labelled with a dungeon on the far side of the map.
+
+Against stock ServUO this resolves **83.2%** of points (5,369 of 6,455): 3,681 by
+region, 1,688 by landmark, 1,086 Wilderness.
+
+## Three quirks in the source data
+
+Each of these is silent if unhandled — the atlas still builds, it is just wrong.
+
+**Facet names disagree between sources.** `Data/Locations/*.xml` spells them
+`Ter Mur` and `Tokuno Islands`, while `<Map>` and `<Facet name>` say `TerMur` and
+`Tokuno`. Unreconciled, the landmark bucket is keyed differently from the points
+looking it up, so the fallback never fires and every unregioned spawn on those
+facets reads "Wilderness".
+
+This is reconciled **by matching, not by a lookup table** — there is no list of
+facet names anywhere. `facetKey()` collapses spelling differences (lowercase,
+alphanumerics only), and `resolveFacetName()` matches a loose spelling against
+the canonical set discovered from the shard's own spawn and region data, by exact
+key then by prefix in either direction. A name matching nothing keeps its own
+name: forcing a wrong match would file a real custom facet's landmarks under the
+wrong facet, which is worse than leaving it alone.
+
+**Spawn type tokens carry XmlSpawner directives.** The `<Objects2>` type is not
+always a bare class name:
+
+```
+Fairy,{RND,4,8}      alchemist/z/-50      Agralem/Name/Agralem
+GargishRouser,1      greatape,true        GargishRefugee/hue/34532
+```
+
+Taken literally these invent creatures that do not exist *and* split real ones in
+two, because `Fairy` and `Fairy,{RND,4,8}` slug apart into separate entries. 71 of
+845 were affected. Everything from the first `/` or `,` is stripped, leaving 800
+real creatures.
+
+**Case is inconsistent across files.** The same creature is `Lizardman` in one
+file and `lizardman` in another. Slugging collapses them correctly, but the
+display name is chosen deterministically — most common spelling wins, ties break
+to the more capitalised form, then alphabetically — because otherwise it would
+depend on file read order and change on an unrelated restart.
+
+## Tables
+
+All are **import-owned**: a refresh empties and reloads them in one transaction,
+so a failed reload leaves the previous atlas intact rather than a half-loaded
+world. Nothing else writes to them and nothing holds a foreign key to them — no
+FKs at all, consistent with every other `shard_*` table. Full column listings in
+[`BACKEND_DESIGN.md`](BACKEND_DESIGN.md).
+
+| Table | Rows (stock) | Notes |
+|---|---|---|
+| `shard_spawn_creatures` | 800 | `slug` PK; `total` = sum of each type's own max; nullable `art` |
+| `shard_spawn_points` | 6,455 | `spawn_range`, since `range` is reserved in MariaDB |
+| `shard_spawn_point_types` | 23,927 | The many-to-many; one spawner commonly carries six types |
+| `shard_regions` | 387 | Flattened out of the nesting; `rects` JSON |
+| `shard_landmarks` | 558 | `grp`, since `group` is reserved in SQL |
+| `shard_champion_spawns` | 25 | Configured altars, not the live feed |
+| `shard_atlas_meta` | 1 | Singleton; source hashes, for the change check |
+| `shard_atlas_pending` | 0–1 | Singleton; a staged refresh awaiting admin review |
+
+`shard_spawn_creatures.name` carries a plain `INDEX`, deliberately **not
+`FULLTEXT`**: ~800 rows makes a `LIKE` scan free, and FULLTEXT's minimum token
+length would break searches for names like "orc".
+
+The reload uses `DELETE`, not `TRUNCATE` — `TRUNCATE` is DDL in MariaDB and would
+implicitly commit, defeating the all-or-nothing guarantee. Point ids are assigned
+explicitly rather than left to `AUTO_INCREMENT`, because the join rows need them
+and `conn.batch()` reports no usable `insertId`.
+
+## Artwork — operator-supplied, never shipped
+
+**This project ships no creature art and no extraction tooling, and never will.**
+UO sprites live in the operator's own client `.mul`/`.uop` files. They are the
+operator's, not ours to redistribute.
+
+The atlas is fully functional as text. `shard_spawn_creatures.art` is nullable
+and is NULL on every fresh import; pages render without images, which is the
+normal and supported state, not a degraded one.
+
+An operator who wants art:
+
+1. Extracts it from **their own** client files (UOFiddler, ClassicUO tooling, or
+   any art extractor).
+2. Drops the images under `server/uploads/atlas/`.
+3. Copies `server/db/data/spawnAtlas.art.example.json` to `spawnAtlas.art.json`
+   and maps creature slugs to file names.
+4. Restarts, or runs `npm run atlas:import -- --force`.
+
+Both `spawnAtlas.art.json` and `server/uploads/` are gitignored, so neither the
+map nor the images can be committed by accident.
+
+## Code layout
+
+| File | Role |
+|---|---|
+| `src/utils/spawnAtlasParse.js` | **Pure and fs-free** parsers, so CI covers them with no ServUO tree. Zero dependencies. |
+| `src/utils/spawnAtlasSource.js` | The only thing that reads a ServUO tree; shared by the boot path and the CLI |
+| `src/model/shardAtlas/shardAtlas.db.js` | The one-transaction replace |
+| `src/model/shardAtlas/shardAtlas.model.js` | The refresh decision, staging, approve/reject |
+| `scripts/importSpawnAtlas.js` | Thin CLI over the model |
+
+Parsing notes:
+
+- `Regions.xml`, `Locations/*.xml` and `ChampionSpawns.xml` genuinely nest, and
+  get a small hand-rolled **subset** tokenizer — elements, attributes,
+  self-closing tags, comments, the XML declaration, CDATA, and the five
+  predefined entities plus numeric refs. It is not a general-purpose XML parser
+  and must not be reused as one.
+- The ~10.5 MB of `Spawns/*.xml` never touches that tokenizer. Those records are
+  flat, so they get a streaming regex sweep instead; a DOM would allocate a node
+  per element across ~40 fields on every record to keep 14 of them. **Do not put
+  the Points files through a DOM parser.**
+- `<Objects2>` is `Type:MX=n:SB=…` segments joined by `:OBJ=`. Split on `:OBJ=`
+  *first* — a naive `split(':')` shreds it. A single Trammel point carries six
+  types.