test(8.0): Tier-1 integration via deterministic embedder (suite runnable again)

The integration suite loaded the real WASM model for every test, so a full run
took ~hours and reliably aborted (vitest reporter timeout) — which is why it was
never gated and silently rotted. Brainy already separates the embedder from all
other logic, so running integration with a deterministic embedder keeps every
code path real (HNSW build+search, graph traversal, metadata filtering,
transactions, storage/sharding, VFS, aggregation, import) while making the suite
complete in ~2 min on ANY machine — no GPU, no 16 GB requirement.

- Add isDeterministicEmbedMode() — a single source of truth for the test-embedder
  switch (BRAINY_DETERMINISTIC_EMBEDDINGS, plus the legacy BRAINY_UNIT_TEST alias),
  used by EmbeddingManager (init/embed/embedBatch) and brainy's eager-init guard.
- setup-integration.ts opts into it: Tier 1 = functional end-to-end. Real-model
  semantic quality + the real embedding pipeline move to a Tier-2 suite.

Full integration now runs 482 passing / 583 in ~2 min (was un-completable). The
remaining failures are pre-existing test rot, fixed next. Unit 1414 green.

src/embeddings/EmbeddingManager.ts

@@ -14,6 +14,7 @@
 
 import { Vector, EmbeddingFunction } from '../coreTypes.js'
 import { WASMEmbeddingEngine } from './wasm/index.js'
+import { isDeterministicEmbedMode } from './deterministicEmbedMode.js'
 
 declare global {
   /**
@@ -77,9 +78,7 @@ export class EmbeddingManager {
    */
   async init(): Promise<void> {
     // In unit test mode, skip real model initialization
-    const isTestMode =
-      process.env.BRAINY_UNIT_TEST === 'true' ||
-      globalThis.__BRAINY_UNIT_TEST__
+    const isTestMode = isDeterministicEmbedMode()
 
     if (isTestMode) {
       // Production safeguard
@@ -151,9 +150,7 @@ export class EmbeddingManager {
    */
   async embed(text: string | string[] | Record<string, unknown>): Promise<Vector> {
     // Check for unit test environment
-    const isTestMode =
-      process.env.BRAINY_UNIT_TEST === 'true' ||
-      globalThis.__BRAINY_UNIT_TEST__
+    const isTestMode = isDeterministicEmbedMode()
 
     if (isTestMode) {
       if (process.env.NODE_ENV === 'production') {
@@ -233,9 +230,7 @@ export class EmbeddingManager {
   async embedBatch(texts: string[], options?: { signal?: AbortSignal }): Promise<number[][]> {
     if (texts.length === 0) return []
 
-    const isTestMode =
-      process.env.BRAINY_UNIT_TEST === 'true' ||
-      globalThis.__BRAINY_UNIT_TEST__
+    const isTestMode = isDeterministicEmbedMode()
 
     if (isTestMode) {
       if (process.env.NODE_ENV === 'production') {

src/embeddings/deterministicEmbedMode.ts

@@ -0,0 +1,39 @@
+/**
+ * @module embeddings/deterministicEmbedMode
+ * @description Single source of truth for "use deterministic test embeddings".
+ *
+ * Brainy's embedding model is a real WASM transformer — correct, but slow on CPU
+ * (a forward pass per text). For TEST tiers that exercise database *logic* (HNSW
+ * build + search, graph traversal, metadata filtering, transactions, storage,
+ * VFS, find composition), the embedding's semantic *quality* is irrelevant; only
+ * a deterministic, content-derived vector is needed. Switching the embedder to a
+ * deterministic stand-in keeps every other code path real while turning a
+ * multi-hour suite into a sub-minute, gateable one. Semantic quality and the
+ * real embedding pipeline are covered separately by the real-model test tier.
+ *
+ * This flag is honored ONLY outside production: {@link EmbeddingManager} throws
+ * if it is set while `NODE_ENV=production` (mock vectors in prod is a security
+ * risk). Enabled by env var or global so test setup files can opt in per tier.
+ */
+
+interface DeterministicEmbedGlobals {
+  __BRAINY_DETERMINISTIC_EMBED__?: boolean
+  /** Legacy alias — the original unit-test-only spelling. */
+  __BRAINY_UNIT_TEST__?: boolean
+}
+
+/**
+ * @description Whether the deterministic test embedder should be used instead of
+ * the real WASM model. True when `BRAINY_DETERMINISTIC_EMBEDDINGS=true` or the
+ * legacy `BRAINY_UNIT_TEST=true` env var is set, or the matching global is set.
+ * @returns `true` to use deterministic embeddings, `false` for the real model.
+ */
+export function isDeterministicEmbedMode(): boolean {
+  const g = globalThis as DeterministicEmbedGlobals
+  return (
+    process.env.BRAINY_DETERMINISTIC_EMBEDDINGS === 'true' ||
+    process.env.BRAINY_UNIT_TEST === 'true' ||
+    g.__BRAINY_DETERMINISTIC_EMBED__ === true ||
+    g.__BRAINY_UNIT_TEST__ === true
+  )
+}