@broberg/ai-sdk 0.42.0 → 0.43.0

package/README.md

@@ -6,12 +6,51 @@ cost control on every call**.
 A provider-agnostic facade: your code calls `ai.chat()`, `ai.vision()`,
 `ai.image()` — never a provider SDK directly. Swap providers by changing a tier,
 not your call-sites. Every call returns a `Usage` (tokens, cost, latency,
-transport) and can fan that out to any cost sink.
+transport, **data-residency region**) and can fan that out to any cost sink.
 
 ```bash
 bun add @broberg/ai-sdk   # or: npm i @broberg/ai-sdk
 ```
 
+## Where did the data go? — `usage.region`
+
+Every response carries the **data residency of the endpoint that actually answered**:
+
+```ts
+const { text, usage } = await ai.chat({ prompt, tier: "smart" });
+if (usage.region !== "eu") throw new Error(`personal data would have left the EU (${usage.region})`);
+```
+
+**Only `"eu"` is a positive claim.** `"unknown"` means we cannot say, so
+`region !== "us"` is **not** an EU check — it passes every single OpenRouter call.
+
+`region` is derived from the **host** the request used, not from the provider's name —
+`vertex`, `azure` and `bfl` are EU by default but each takes an override, so a
+name-based table would report `eu` for a call someone had pointed at `us-central1`.
+
+**Three things that decide whether your guard works:**
+
+| | |
+|---|---|
+| **After the call** | `usage.region` — `"eu"` \| `"us"` \| `"cn"` \| `"unknown"` |
+| **Before the call** | `regionOfHost(urlOrHostname)` — the only way to refuse *before* bytes leave |
+| **Never** | `regionOfProvider(name)` — see below |
+
+```ts
+import { regionOfHost } from "@broberg/ai-sdk";
+if (regionOfHost("https://api.mistral.ai/v1") !== "eu") throw new Error("not EU");
+```
+
+**Do not build a residency guard on `regionOfProvider`.** A name cannot answer for
+hosting when the provider takes a `baseUrl`, so `regionOfProvider("mistral")` is
+`"unknown"` — and a guard written as `regionOfProvider(p) === "eu"` therefore rejects
+**Mistral, the only EU route there is.** Fail-closed and useless: it filters out the
+thing it exists to allow.
+
+An allowlist of `(provider, model)` pairs has the same hole from the other side: a
+gateway in front of Mistral is still `"mistral"`/`"mistral-large-latest"`, matches both
+fields, and passes. `regionOfHost` answers `"unknown"` for it, which is the honest answer.
+
 ## Quick start
 
 ```ts

package/dist/index.d.ts

@@ -61,6 +61,24 @@ interface Pronunciation {
     word: string;
     /** Say this instead. Azure `<sub alias>`; ElevenLabs plain substitution. */
     alias?: string;
+    /** Also match INSIDE a hyphenated compound — `AI` in `AI-agenter` (F051.3).
+     *
+     *  Danish puts abbreviations there constantly. cms measured their own articles:
+     *  "AI-agenter" 13 times, "AI-native" 12, 60+ occurrences in all, every one read aloud
+     *  as one mangled word. Christian heard it as "HTLM" and asked why AI was not spelled
+     *  out too — it WAS, just never in a compound.
+     *
+     *  **Per entry, and defaulting to `false`, because the right answer depends on the
+     *  word.** `AI` yes; `mail` no — a rule for `mail` must not fire inside `e-mail`.
+     *  Measured rather than reasoned: `@broberg/speech-dictionary` ships the general
+     *  loosening and pays exactly that price today. A single global answer is wrong for
+     *  one of the two classes whichever way it is set, so the person who wrote the entry
+     *  decides.
+     *
+     *  Default `false` is also the safe direction to be wrong in: flipping it would change
+     *  the audio for every existing consumer IN SOUND, where there is no compile error, no
+     *  red test and no log line to notice it by. */
+    matchInCompounds?: boolean;
     /** IPA phonemes. Azure `<phoneme alphabet="ipa" ph>`. ElevenLabs cannot do this
      *  (it THROWS rather than silently skipping). See the type doc above before reaching
      *  for this on a foreign word. */
@@ -1899,13 +1917,17 @@ declare const ttsInputSchema: z.ZodObject<{
         alias: z.ZodOptional<z.ZodString>;
         ipa: z.ZodOptional<z.ZodString>;
         lang: z.ZodOptional<z.ZodString>;
+        /** F051.3 — also match inside a hyphenated compound ("AI" in "AI-agenter"). */
+        matchInCompounds: z.ZodOptional<z.ZodBoolean>;
     }, "strip", z.ZodTypeAny, {
         word: string;
+        matchInCompounds?: boolean | undefined;
         alias?: string | undefined;
         ipa?: string | undefined;
         lang?: string | undefined;
     }, {
         word: string;
+        matchInCompounds?: boolean | undefined;
         alias?: string | undefined;
         ipa?: string | undefined;
         lang?: string | undefined;
@@ -1944,6 +1966,7 @@ declare const ttsInputSchema: z.ZodObject<{
     lang?: string | undefined;
     pronunciations?: {
         word: string;
+        matchInCompounds?: boolean | undefined;
         alias?: string | undefined;
         ipa?: string | undefined;
         lang?: string | undefined;
@@ -1970,6 +1993,7 @@ declare const ttsInputSchema: z.ZodObject<{
     lang?: string | undefined;
     pronunciations?: {
         word: string;
+        matchInCompounds?: boolean | undefined;
         alias?: string | undefined;
         ipa?: string | undefined;
         lang?: string | undefined;
@@ -2366,8 +2390,8 @@ declare const falStubAdapter: ProviderAdapter;
  *  wires the live adapters. */
 declare const stubProviders: Record<string, ProviderAdapter>;
 
-declare const VERSION: "0.42.0";
-declare const SDK_TAG: "@broberg/ai-sdk@0.42.0";
+declare const VERSION: "0.43.0";
+declare const SDK_TAG: "@broberg/ai-sdk@0.43.0";
 
 /** Built-in defaults. Every entry is overridable via AiConfig.defaults or a
  *  per-call override.

package/dist/index.js

@@ -1557,10 +1557,14 @@ function applyPronunciations(haystack, list, render, escape = (s) => s) {
     if (!byLower.has(k)) byLower.set(k, p);
   }
   const alternation = sorted.map((p) => escapeRegex(escape(p.word))).join("|");
-  const re = new RegExp(`(?<![\\w-])(${alternation})(?![\\w-])`, "gi");
-  return haystack.replace(re, (matched) => {
+  const re = new RegExp(`(?<!\\w)(${alternation})(?!\\w)`, "gi");
+  return haystack.replace(re, (matched, _g1, offset) => {
     const entry = byLower.get(matched.toLowerCase());
-    return entry ? render(entry, matched) : matched;
+    if (!entry) return matched;
+    if (!entry.matchInCompounds) {
+      if (haystack[offset - 1] === "-" || haystack[offset + matched.length] === "-") return matched;
+    }
+    return render(entry, matched);
   });
 }
 
@@ -3061,7 +3065,9 @@ var pronunciationSchema = z.object({
   word: z.string(),
   alias: z.string().optional(),
   ipa: z.string().optional(),
-  lang: z.string().optional()
+  lang: z.string().optional(),
+  /** F051.3 — also match inside a hyphenated compound ("AI" in "AI-agenter"). */
+  matchInCompounds: z.boolean().optional()
 });
 var ttsInputSchema = z.object({
   text: z.string(),
@@ -3110,8 +3116,8 @@ var aiConfigSchema = z.object({
 });
 
 // src/version.ts
-var VERSION = "0.42.0";
-var SDK_TAG = "@broberg/ai-sdk@0.42.0";
+var VERSION = "0.43.0";
+var SDK_TAG = "@broberg/ai-sdk@0.43.0";
 
 // src/cost/sinks/upmetrics.ts
 function upmetricsSink(config) {