@broberg/ai-sdk 0.19.0 → 0.20.0

package/dist/index.d.ts

@@ -245,6 +245,9 @@ interface TranscribeRequest {
     language?: string;
     /** Audio length in seconds — enables per-minute cost (Whisper). Omit → cost 0. */
     durationSec?: number;
+    /** Bias recognition toward these brand/jargon terms (Azure phraseList, F029.3).
+     *  Providers without biasing support (Voxtral/Whisper) ignore it. */
+    phrases?: string[];
     spec: TierSpec;
 }
 interface TranscribeResult {
@@ -1312,6 +1315,8 @@ declare const transcribeInputSchema: z.ZodObject<{
     language: z.ZodOptional<z.ZodString>;
     /** Audio length in seconds — enables Whisper per-minute cost. */
     durationSec: z.ZodOptional<z.ZodNumber>;
+    /** Bias toward brand/jargon terms (Azure phraseList, F029.3); others ignore it. */
+    phrases: z.ZodOptional<z.ZodArray<z.ZodString, "many">>;
 }, "strip", z.ZodTypeAny, {
     audio: string | Uint8Array<ArrayBuffer>;
     language?: string | undefined;
@@ -1329,6 +1334,7 @@ declare const transcribeInputSchema: z.ZodObject<{
     })[] | undefined;
     labels?: Record<string, string> | undefined;
     durationSec?: number | undefined;
+    phrases?: string[] | undefined;
 }, {
     audio: string | Uint8Array<ArrayBuffer>;
     language?: string | undefined;
@@ -1346,6 +1352,7 @@ declare const transcribeInputSchema: z.ZodObject<{
     })[] | undefined;
     labels?: Record<string, string> | undefined;
     durationSec?: number | undefined;
+    phrases?: string[] | undefined;
 }>;
 declare const ocrInputSchema: z.ZodObject<{
     tier: z.ZodOptional<z.ZodEnum<["fast", "smart", "powerful", "cheap", "vision", "video", "embedding"]>>;
@@ -1804,6 +1811,14 @@ declare function openrouterAdapter(config?: {
     title?: string;
 }): ProviderAdapter;
 
+declare function requestyAdapter(config?: {
+    apiKey?: string;
+    baseUrl?: string;
+    eu?: boolean;
+    referer?: string;
+    title?: string;
+}): ProviderAdapter;
+
 declare function mistralAdapter(config?: {
     apiKey?: string;
     baseUrl?: string;
@@ -1863,6 +1878,17 @@ declare function azureAdapter(config?: {
     region?: string;
     fetch?: typeof fetch;
     pricePer1kChars?: number;
+    /** USD per audio-minute for transcribe (overrides AZURE_STT_PRICE_PER_MIN). */
+    sttPricePerMin?: number;
+    /** STT base URL override (e.g. a resource custom domain). */
+    sttBaseUrl?: string;
+    /** Resource name → custom-domain STT host `{resource}.cognitiveservices.azure.com`
+     *  (or env AZURE_SPEECH_RESOURCE). Without it, STT uses the regional host. */
+    resource?: string;
+    /** Fast-transcription api-version (overrides the GA default). */
+    sttApiVersion?: string;
+    /** phraseList biasing weight (0–2) applied when a call passes `phrases`. Default 1.5. */
+    sttBiasingWeight?: number;
 }): ProviderAdapter;
 
 interface FalAdapterConfig {
@@ -1945,8 +1971,8 @@ declare const falStubAdapter: ProviderAdapter;
  *  wires the live adapters. */
 declare const stubProviders: Record<string, ProviderAdapter>;
 
-declare const VERSION: "0.19.0";
-declare const SDK_TAG: "@broberg/ai-sdk@0.19.0";
+declare const VERSION: "0.20.0";
+declare const SDK_TAG: "@broberg/ai-sdk@0.20.0";
 
 /** Built-in defaults. Every entry is overridable via AiConfig.defaults or a
  *  per-call override. Model IDs are current at scaffold time; callers pin their
@@ -2269,4 +2295,4 @@ interface StreamTransportRequest extends TransportRequest {
  */
 declare function streamTransport(req: StreamTransportRequest): AsyncIterable<string>;
 
-export { AZURE_DANISH_VOICES, AZURE_DANISH_VOICE_LIST, type AiClient, type AiConfig, type AzureVoiceInfo, type BatchJob, type BatchRequestItem, type BatchResultItem, type BflAdapterConfig, type BflCredits, type BudgetConfig, BudgetExceededError, BudgetGuard, type BudgetStore, type CallOptions, type Capability, type ChatInput, type ChatRequest, type ChatResult, type ChatStreamEvent, type ClassifyInput, type ClassifyResult, type ContentPart, type Contracts, type CostQuery, type CostSink, type CostSummary, type CostSummaryQuery, type CostTimeseriesQuery, DEFAULT_TIER_MAP, type DesignInput, type DesignResult, type DialogueRequest, type DialogueTurn, type DiscordSinkConfig, ELEVENLABS_DANISH_VOICES, type EmbeddingInput, type EmbeddingRequest, type EmbeddingResult, type ExtractInput, type ExtractResult, type FalAdapterConfig, type HttpResponse, type ImageInput, type ImageRequest, type ImageResult, type LoraWeight, type Message, type MockupInput, type MockupResult, type ModerationInput, type ModerationItem, type ModerationRequest, type ModerationResult, type OcrInput, type OcrPage, type OcrRequest, type OcrResult, type OpenAICompatibleConfig, type PodcastInput, type PodcastResult, type PricingEntry, type ProviderAdapter, type RefreshOptions, type RefreshResult, type RerankInput, type RerankResult, type Role, SDK_TAG, type SqliteBudgetStoreConfig, type SqliteSinkConfig, StreamHttpError, type SubprocessResponse, type Tier, type TierSpec, type Tool, type ToolCall, type TrainStyleInput, type TrainStyleRequest, type TrainStyleResult, type TranscribeInput, type TranscribeRequest, type TranscribeResult, type TranslateInput, type TranslateResult, type Transport, type TransportRequest, type TransportResponse, type TtsInput, type TtsRequest, type UpmetricsCostClientConfig, UpmetricsCostError, type UpmetricsCostRow, type UpmetricsCostSummary, type UpmetricsCostTimeseries, type UpmetricsSinkConfig, type Usage, VERSION, type VideoInput, type VisionInput, aiConfigSchema, anthropicAdapter, anthropicApiAdapter, anthropicSubprocessAdapter, azureAdapter, bflAdapter, bflCredits, chatInputSchema, computeCost, createAI, deepinfraAdapter, defaultProviders, discordSink, elevenlabsAdapter, embeddingInputSchema, falAdapter, falStubAdapter, freshUsage, fromProviderToolCall, geminiAdapter, getCostSummary, getPrice, httpTransport, imageInputSchema, listAzureDanishVoices, makeContracts, makeOpenAICompatibleAdapter, messageSchema, mistralAdapter, multiSink, noopSink, openaiAdapter, openaiStubAdapter, openrouterAdapter, parseClaudeCliJson, parseJsonLoose, refreshAvailability, resetRefreshState, resetRegistry, resolveAzureVoice, resolveTier, resolveVoice, sqliteBudgetStore, sqliteSink, streamTransport, stubProviders, subprocessTransport, tierSpecSchema, toProviderTools, toolSchema, translateInputSchema, upmetricsCostClient, upmetricsSink, usdFromMicro, visionInputSchema };
+export { AZURE_DANISH_VOICES, AZURE_DANISH_VOICE_LIST, type AiClient, type AiConfig, type AzureVoiceInfo, type BatchJob, type BatchRequestItem, type BatchResultItem, type BflAdapterConfig, type BflCredits, type BudgetConfig, BudgetExceededError, BudgetGuard, type BudgetStore, type CallOptions, type Capability, type ChatInput, type ChatRequest, type ChatResult, type ChatStreamEvent, type ClassifyInput, type ClassifyResult, type ContentPart, type Contracts, type CostQuery, type CostSink, type CostSummary, type CostSummaryQuery, type CostTimeseriesQuery, DEFAULT_TIER_MAP, type DesignInput, type DesignResult, type DialogueRequest, type DialogueTurn, type DiscordSinkConfig, ELEVENLABS_DANISH_VOICES, type EmbeddingInput, type EmbeddingRequest, type EmbeddingResult, type ExtractInput, type ExtractResult, type FalAdapterConfig, type HttpResponse, type ImageInput, type ImageRequest, type ImageResult, type LoraWeight, type Message, type MockupInput, type MockupResult, type ModerationInput, type ModerationItem, type ModerationRequest, type ModerationResult, type OcrInput, type OcrPage, type OcrRequest, type OcrResult, type OpenAICompatibleConfig, type PodcastInput, type PodcastResult, type PricingEntry, type ProviderAdapter, type RefreshOptions, type RefreshResult, type RerankInput, type RerankResult, type Role, SDK_TAG, type SqliteBudgetStoreConfig, type SqliteSinkConfig, StreamHttpError, type SubprocessResponse, type Tier, type TierSpec, type Tool, type ToolCall, type TrainStyleInput, type TrainStyleRequest, type TrainStyleResult, type TranscribeInput, type TranscribeRequest, type TranscribeResult, type TranslateInput, type TranslateResult, type Transport, type TransportRequest, type TransportResponse, type TtsInput, type TtsRequest, type UpmetricsCostClientConfig, UpmetricsCostError, type UpmetricsCostRow, type UpmetricsCostSummary, type UpmetricsCostTimeseries, type UpmetricsSinkConfig, type Usage, VERSION, type VideoInput, type VisionInput, aiConfigSchema, anthropicAdapter, anthropicApiAdapter, anthropicSubprocessAdapter, azureAdapter, bflAdapter, bflCredits, chatInputSchema, computeCost, createAI, deepinfraAdapter, defaultProviders, discordSink, elevenlabsAdapter, embeddingInputSchema, falAdapter, falStubAdapter, freshUsage, fromProviderToolCall, geminiAdapter, getCostSummary, getPrice, httpTransport, imageInputSchema, listAzureDanishVoices, makeContracts, makeOpenAICompatibleAdapter, messageSchema, mistralAdapter, multiSink, noopSink, openaiAdapter, openaiStubAdapter, openrouterAdapter, parseClaudeCliJson, parseJsonLoose, refreshAvailability, requestyAdapter, resetRefreshState, resetRegistry, resolveAzureVoice, resolveTier, resolveVoice, sqliteBudgetStore, sqliteSink, streamTransport, stubProviders, subprocessTransport, tierSpecSchema, toProviderTools, toolSchema, translateInputSchema, upmetricsCostClient, upmetricsSink, usdFromMicro, visionInputSchema };

package/dist/index.js

@@ -1056,6 +1056,24 @@ function openrouterAdapter(config = {}) {
   });
 }
 
+// src/providers/requesty.ts
+var US_BASE = "https://router.requesty.ai/v1";
+var EU_BASE = "https://router.eu.requesty.ai/v1";
+function requestyAdapter(config = {}) {
+  return makeOpenAICompatibleAdapter({
+    name: "requesty",
+    baseUrl: config.baseUrl ?? (config.eu ? EU_BASE : US_BASE),
+    apiKey: config.apiKey,
+    extraHeaders: {
+      "HTTP-Referer": config.referer ?? "https://broberg.ai",
+      "X-Title": config.title ?? "@broberg/ai-sdk"
+    },
+    // Requesty returns ground-truth usage.cost (USD) by default — use it over the
+    // local pricing-table estimate (same as OpenRouter, F010).
+    costFromResponseField: true
+  });
+}
+
 // src/providers/mistral.ts
 var MISTRAL_OCR_PRICE_PER_PAGE = 2e-3;
 var VOXTRAL_PRICE_PER_MIN = {
@@ -1305,8 +1323,29 @@ function elevenlabsAdapter(config = {}) {
 
 // src/providers/azure.ts
 var AZURE_TTS_PRICE_PER_1K_CHARS = 0.016;
+var AZURE_STT_PRICE_PER_MIN = 0.0167;
+var DEFAULT_STT_API_VERSION = "2025-10-15";
 var DEFAULT_REGION = "westeurope";
 var DEFAULT_FORMAT = "audio-24khz-48kbitrate-mono-mp3";
+var AZURE_LOCALE_MAP = {
+  da: "da-DK",
+  en: "en-US",
+  de: "de-DE",
+  sv: "sv-SE",
+  nb: "nb-NO",
+  no: "nb-NO",
+  fi: "fi-FI",
+  nl: "nl-NL",
+  fr: "fr-FR",
+  es: "es-ES",
+  it: "it-IT",
+  pt: "pt-PT"
+};
+function toAzureLocale(lang) {
+  if (!lang) return "da-DK";
+  if (lang.includes("-")) return lang;
+  return AZURE_LOCALE_MAP[lang.toLowerCase()] ?? lang;
+}
 var AZURE_DANISH_VOICE_LIST = [
   { name: "christel", voiceId: "da-DK-ChristelNeural", gender: "female", display: "Christel", native: true, defaultRate: 0.85 },
   { name: "seraphina", voiceId: "de-DE-SeraphinaMultilingualNeural", gender: "female", display: "Seraphina", native: false },
@@ -1341,6 +1380,12 @@ function azureAdapter(config = {}) {
   function region() {
     return config.region ?? process.env.AZURE_SPEECH_REGION ?? DEFAULT_REGION;
   }
+  function sttBaseUrl() {
+    if (config.sttBaseUrl) return config.sttBaseUrl.replace(/\/$/, "");
+    const resource = config.resource ?? process.env.AZURE_SPEECH_RESOURCE;
+    if (resource) return `https://${resource}.cognitiveservices.azure.com`;
+    return `https://${region()}.api.cognitive.microsoft.com`;
+  }
   function priceFor(chars, model) {
     const usage = freshUsage({
       provider: "azure",
@@ -1380,7 +1425,40 @@ function azureAdapter(config = {}) {
     const audio = new Uint8Array(await res.arrayBuffer());
     return { audio, mimeType: "audio/mpeg", usage: priceFor(req.text.length, req.spec.model) };
   }
-  return { name: "azure", tts };
+  async function transcribe(req) {
+    const locale = toAzureLocale(req.language);
+    const definition = { locales: [locale] };
+    if (req.phrases && req.phrases.length > 0) {
+      definition.phraseList = { phrases: req.phrases, biasingWeight: config.sttBiasingWeight ?? 1.5 };
+    }
+    const form = new FormData();
+    form.append("audio", new Blob([req.audio]), "audio");
+    form.append("definition", JSON.stringify(definition));
+    const url = `${sttBaseUrl()}/speechtotext/transcriptions:transcribe?api-version=${config.sttApiVersion ?? DEFAULT_STT_API_VERSION}`;
+    const res = await fetchImpl(url, {
+      method: "POST",
+      headers: { "Ocp-Apim-Subscription-Key": key() },
+      body: form
+    });
+    if (!res.ok) {
+      const body = await res.text().catch(() => "");
+      throw new Error(`azure transcribe ${res.status}: ${body.slice(0, 300)}`);
+    }
+    const data = await res.json();
+    const text = data.combinedPhrases?.[0]?.text ?? "";
+    const minutes = data.durationMilliseconds != null ? data.durationMilliseconds / 6e4 : (req.durationSec ?? 0) / 60;
+    const usage = freshUsage({
+      provider: "azure",
+      model: req.spec.model,
+      transport: "http",
+      capability: "transcribe",
+      inputTokens: 0,
+      outputTokens: 0
+    });
+    usage.costUsd = minutes * (config.sttPricePerMin ?? AZURE_STT_PRICE_PER_MIN);
+    return { text, usage };
+  }
+  return { name: "azure", tts, transcribe };
 }
 
 // src/providers/fal.ts
@@ -1696,7 +1774,7 @@ function buildZip(files) {
 }
 
 // src/providers/bfl.ts
-var EU_BASE = "https://api.eu.bfl.ai";
+var EU_BASE2 = "https://api.eu.bfl.ai";
 var BFL_CREDIT_USD = 0.01;
 var BFL_IMAGE_PRICE = 0.06;
 var sleep2 = (ms) => new Promise((r) => setTimeout(r, ms));
@@ -1707,7 +1785,7 @@ async function bflCredits(opts = {}) {
   const apiKey = opts.apiKey ?? process.env.BFL_API_KEY;
   if (!apiKey) throw new Error("bflCredits: BFL_API_KEY not set");
   const doFetch = opts.fetch ?? fetch;
-  const res = await doFetch(`${opts.baseUrl ?? EU_BASE}/v1/credits`, { headers: { "x-key": apiKey } });
+  const res = await doFetch(`${opts.baseUrl ?? EU_BASE2}/v1/credits`, { headers: { "x-key": apiKey } });
   if (!res.ok) {
     throw new Error(`bflCredits ${res.status}: ${(await res.text().catch(() => "")).slice(0, 200)}`);
   }
@@ -1717,7 +1795,7 @@ async function bflCredits(opts = {}) {
 }
 function bflAdapter(config = {}) {
   const doFetch = config.fetch ?? fetch;
-  const base = config.baseUrl ?? EU_BASE;
+  const base = config.baseUrl ?? EU_BASE2;
   const pollIntervalMs = config.pollIntervalMs ?? 1500;
   const timeoutMs = config.timeoutMs ?? 12e4;
   const resolveKey = () => config.apiKey ?? process.env.BFL_API_KEY;
@@ -1808,6 +1886,7 @@ var defaultProviders = {
   gemini: geminiAdapter(),
   deepinfra: deepinfraAdapter(),
   openrouter: openrouterAdapter(),
+  requesty: requestyAdapter(),
   mistral: mistralAdapter(),
   elevenlabs: elevenlabsAdapter(),
   azure: azureAdapter(),
@@ -2173,6 +2252,8 @@ var transcribeInputSchema = z.object({
   language: z.string().optional(),
   /** Audio length in seconds — enables Whisper per-minute cost. */
   durationSec: z.number().positive().optional(),
+  /** Bias toward brand/jargon terms (Azure phraseList, F029.3); others ignore it. */
+  phrases: z.array(z.string()).optional(),
   ...callOptions
 });
 var ocrInputSchema = z.object({
@@ -2689,7 +2770,7 @@ function createAI(config = {}) {
         invoke: async (spec) => {
           const adapter = pickProvider(spec.provider);
           if (!adapter.transcribe) throw new Error(`createAI: provider "${spec.provider}" does not support transcribe`);
-          return adapter.transcribe({ audio, language: input.language, durationSec: input.durationSec, spec });
+          return adapter.transcribe({ audio, language: input.language, durationSec: input.durationSec, phrases: input.phrases, spec });
         }
       });
     },
@@ -2801,8 +2882,8 @@ var stubProviders = {
 };
 
 // src/version.ts
-var VERSION = "0.19.0";
-var SDK_TAG = "@broberg/ai-sdk@0.19.0";
+var VERSION = "0.20.0";
+var SDK_TAG = "@broberg/ai-sdk@0.20.0";
 
 // src/availability/refresh.ts
 var NOT_REFRESHED = { refreshed: false, checked: 0, markedUnavailable: [] };
@@ -3192,6 +3273,7 @@ export {
   parseClaudeCliJson,
   parseJsonLoose,
   refreshAvailability,
+  requestyAdapter,
   resetRefreshState,
   resetRegistry,
   resolveAzureVoice,