@broberg/ai-sdk 0.6.0 → 0.8.0

package/dist/index.d.ts

@@ -13,7 +13,7 @@ interface TierSpec {
     transport: Transport;
 }
 /** High-level capability a call exercises. Mirrors the capability layer (F5). */
-type Capability = "chat" | "vision" | "video" | "translate" | "image" | "embedding" | "transcribe" | "mockup" | "design" | "extract" | "classify" | "rerank";
+type Capability = "chat" | "vision" | "video" | "translate" | "image" | "embedding" | "transcribe" | "ocr" | "moderation" | "podcast" | "tts" | "mockup" | "design" | "extract" | "classify" | "rerank";
 type Role = "system" | "user" | "assistant" | "tool";
 /** A piece of message content. Text everywhere; image parts feed vision. */
 type ContentPart = {
@@ -180,6 +180,55 @@ interface TranscribeResult {
     text: string;
     usage: Usage;
 }
+interface OcrRequest {
+    /** A URL, data-URL, or raw bytes of the document/image. */
+    document: string | Uint8Array;
+    mimeType?: string;
+    spec: TierSpec;
+}
+interface OcrPage {
+    index: number;
+    markdown: string;
+}
+interface OcrResult {
+    pages: OcrPage[];
+    usage: Usage;
+}
+interface ModerationRequest {
+    input: string[];
+    spec: TierSpec;
+}
+interface ModerationItem {
+    /** True if any category tripped. */
+    flagged: boolean;
+    categories: Record<string, boolean>;
+    categoryScores: Record<string, number>;
+}
+interface ModerationResult {
+    results: ModerationItem[];
+    usage: Usage;
+}
+interface DialogueTurn {
+    text: string;
+    voiceId: string;
+}
+interface DialogueRequest {
+    inputs: DialogueTurn[];
+    /** Output container, e.g. "mp3" (default). */
+    format?: string;
+    spec: TierSpec;
+}
+interface PodcastResult {
+    /** Episode audio bytes. */
+    audio: Uint8Array;
+    mimeType: string;
+    usage: Usage;
+}
+interface TtsRequest {
+    text: string;
+    voiceId: string;
+    spec: TierSpec;
+}
 /** The thin contract every provider implements (F4). A provider need only
  *  support the capabilities it offers — `chat` is the baseline; vision/image/
  *  embedding are optional and absence is a typed capability gap. */
@@ -196,6 +245,12 @@ interface ProviderAdapter {
     image?(req: ImageRequest): Promise<ImageResult>;
     embedding?(req: EmbeddingRequest): Promise<EmbeddingResult>;
     transcribe?(req: TranscribeRequest): Promise<TranscribeResult>;
+    ocr?(req: OcrRequest): Promise<OcrResult>;
+    moderate?(req: ModerationRequest): Promise<ModerationResult>;
+    /** Multi-voice dialogue → one audio episode (F020). ElevenLabs. */
+    dialogue?(req: DialogueRequest): Promise<PodcastResult>;
+    /** Single-voice TTS (F020.4) → audio. ElevenLabs. */
+    tts?(req: TtsRequest): Promise<PodcastResult>;
 }
 interface TranslateResult {
     text: string;
@@ -971,6 +1026,283 @@ declare const transcribeInputSchema: z.ZodObject<{
     labels?: Record<string, string> | undefined;
     durationSec?: number | undefined;
 }>;
+declare const ocrInputSchema: z.ZodObject<{
+    tier: z.ZodOptional<z.ZodEnum<["fast", "smart", "powerful", "cheap", "vision", "video", "embedding"]>>;
+    override: z.ZodOptional<z.ZodObject<{
+        provider: z.ZodOptional<z.ZodString>;
+        model: z.ZodOptional<z.ZodString>;
+        transport: z.ZodOptional<z.ZodEnum<["http", "subprocess"]>>;
+    }, "strip", z.ZodTypeAny, {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    }, {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    }>>;
+    fallback: z.ZodOptional<z.ZodArray<z.ZodUnion<[z.ZodEnum<["fast", "smart", "powerful", "cheap", "vision", "video", "embedding"]>, z.ZodObject<{
+        provider: z.ZodString;
+        model: z.ZodString;
+        transport: z.ZodEnum<["http", "subprocess"]>;
+    }, "strip", z.ZodTypeAny, {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    }, {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    }>]>, "many">>;
+    purpose: z.ZodOptional<z.ZodString>;
+    labels: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
+    /** A URL, data-URL, or raw bytes of the document/image. */
+    document: z.ZodUnion<[z.ZodString, z.ZodType<Uint8Array<ArrayBuffer>, z.ZodTypeDef, Uint8Array<ArrayBuffer>>]>;
+    /** image/* → routed as an image; anything else → a document (PDF etc.). */
+    mimeType: z.ZodOptional<z.ZodString>;
+}, "strip", z.ZodTypeAny, {
+    document: string | Uint8Array<ArrayBuffer>;
+    mimeType?: string | undefined;
+    tier?: "fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | undefined;
+    override?: {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    } | undefined;
+    fallback?: ("fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    })[] | undefined;
+    purpose?: string | undefined;
+    labels?: Record<string, string> | undefined;
+}, {
+    document: string | Uint8Array<ArrayBuffer>;
+    mimeType?: string | undefined;
+    tier?: "fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | undefined;
+    override?: {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    } | undefined;
+    fallback?: ("fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    })[] | undefined;
+    purpose?: string | undefined;
+    labels?: Record<string, string> | undefined;
+}>;
+declare const moderationInputSchema: z.ZodObject<{
+    tier: z.ZodOptional<z.ZodEnum<["fast", "smart", "powerful", "cheap", "vision", "video", "embedding"]>>;
+    override: z.ZodOptional<z.ZodObject<{
+        provider: z.ZodOptional<z.ZodString>;
+        model: z.ZodOptional<z.ZodString>;
+        transport: z.ZodOptional<z.ZodEnum<["http", "subprocess"]>>;
+    }, "strip", z.ZodTypeAny, {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    }, {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    }>>;
+    fallback: z.ZodOptional<z.ZodArray<z.ZodUnion<[z.ZodEnum<["fast", "smart", "powerful", "cheap", "vision", "video", "embedding"]>, z.ZodObject<{
+        provider: z.ZodString;
+        model: z.ZodString;
+        transport: z.ZodEnum<["http", "subprocess"]>;
+    }, "strip", z.ZodTypeAny, {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    }, {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    }>]>, "many">>;
+    purpose: z.ZodOptional<z.ZodString>;
+    labels: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
+    input: z.ZodUnion<[z.ZodString, z.ZodArray<z.ZodString, "many">]>;
+}, "strip", z.ZodTypeAny, {
+    input: string | string[];
+    tier?: "fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | undefined;
+    override?: {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    } | undefined;
+    fallback?: ("fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    })[] | undefined;
+    purpose?: string | undefined;
+    labels?: Record<string, string> | undefined;
+}, {
+    input: string | string[];
+    tier?: "fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | undefined;
+    override?: {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    } | undefined;
+    fallback?: ("fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    })[] | undefined;
+    purpose?: string | undefined;
+    labels?: Record<string, string> | undefined;
+}>;
+declare const podcastInputSchema: z.ZodObject<{
+    tier: z.ZodOptional<z.ZodEnum<["fast", "smart", "powerful", "cheap", "vision", "video", "embedding"]>>;
+    override: z.ZodOptional<z.ZodObject<{
+        provider: z.ZodOptional<z.ZodString>;
+        model: z.ZodOptional<z.ZodString>;
+        transport: z.ZodOptional<z.ZodEnum<["http", "subprocess"]>>;
+    }, "strip", z.ZodTypeAny, {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    }, {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    }>>;
+    fallback: z.ZodOptional<z.ZodArray<z.ZodUnion<[z.ZodEnum<["fast", "smart", "powerful", "cheap", "vision", "video", "embedding"]>, z.ZodObject<{
+        provider: z.ZodString;
+        model: z.ZodString;
+        transport: z.ZodEnum<["http", "subprocess"]>;
+    }, "strip", z.ZodTypeAny, {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    }, {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    }>]>, "many">>;
+    purpose: z.ZodOptional<z.ZodString>;
+    labels: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
+    script: z.ZodArray<z.ZodObject<{
+        speaker: z.ZodString;
+        text: z.ZodString;
+    }, "strip", z.ZodTypeAny, {
+        text: string;
+        speaker: string;
+    }, {
+        text: string;
+        speaker: string;
+    }>, "many">;
+    voices: z.ZodRecord<z.ZodString, z.ZodString>;
+    format: z.ZodOptional<z.ZodString>;
+}, "strip", z.ZodTypeAny, {
+    script: {
+        text: string;
+        speaker: string;
+    }[];
+    voices: Record<string, string>;
+    tier?: "fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | undefined;
+    override?: {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    } | undefined;
+    fallback?: ("fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    })[] | undefined;
+    purpose?: string | undefined;
+    labels?: Record<string, string> | undefined;
+    format?: string | undefined;
+}, {
+    script: {
+        text: string;
+        speaker: string;
+    }[];
+    voices: Record<string, string>;
+    tier?: "fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | undefined;
+    override?: {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    } | undefined;
+    fallback?: ("fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    })[] | undefined;
+    purpose?: string | undefined;
+    labels?: Record<string, string> | undefined;
+    format?: string | undefined;
+}>;
+declare const ttsInputSchema: z.ZodObject<{
+    tier: z.ZodOptional<z.ZodEnum<["fast", "smart", "powerful", "cheap", "vision", "video", "embedding"]>>;
+    override: z.ZodOptional<z.ZodObject<{
+        provider: z.ZodOptional<z.ZodString>;
+        model: z.ZodOptional<z.ZodString>;
+        transport: z.ZodOptional<z.ZodEnum<["http", "subprocess"]>>;
+    }, "strip", z.ZodTypeAny, {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    }, {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    }>>;
+    fallback: z.ZodOptional<z.ZodArray<z.ZodUnion<[z.ZodEnum<["fast", "smart", "powerful", "cheap", "vision", "video", "embedding"]>, z.ZodObject<{
+        provider: z.ZodString;
+        model: z.ZodString;
+        transport: z.ZodEnum<["http", "subprocess"]>;
+    }, "strip", z.ZodTypeAny, {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    }, {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    }>]>, "many">>;
+    purpose: z.ZodOptional<z.ZodString>;
+    labels: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
+    text: z.ZodString;
+    voice: z.ZodString;
+}, "strip", z.ZodTypeAny, {
+    text: string;
+    voice: string;
+    tier?: "fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | undefined;
+    override?: {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    } | undefined;
+    fallback?: ("fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    })[] | undefined;
+    purpose?: string | undefined;
+    labels?: Record<string, string> | undefined;
+}, {
+    text: string;
+    voice: string;
+    tier?: "fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | undefined;
+    override?: {
+        provider?: string | undefined;
+        model?: string | undefined;
+        transport?: "http" | "subprocess" | undefined;
+    } | undefined;
+    fallback?: ("fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | {
+        provider: string;
+        model: string;
+        transport: "http" | "subprocess";
+    })[] | undefined;
+    purpose?: string | undefined;
+    labels?: Record<string, string> | undefined;
+}>;
 declare const aiConfigSchema: z.ZodObject<{
     defaults: z.ZodOptional<z.ZodRecord<z.ZodEnum<["fast", "smart", "powerful", "cheap", "vision", "video", "embedding"]>, z.ZodObject<{
         provider: z.ZodString;
@@ -1029,6 +1361,10 @@ type TranslateInput = z.infer<typeof translateInputSchema>;
 type ImageInput = z.infer<typeof imageInputSchema>;
 type EmbeddingInput = z.infer<typeof embeddingInputSchema>;
 type TranscribeInput = z.infer<typeof transcribeInputSchema>;
+type OcrInput = z.infer<typeof ocrInputSchema>;
+type ModerationInput = z.infer<typeof moderationInputSchema>;
+type PodcastInput = z.infer<typeof podcastInputSchema>;
+type TtsInput = z.infer<typeof ttsInputSchema>;
 type AiConfig = z.infer<typeof aiConfigSchema>;
 /** The public facade. Defined here because it depends on the derived inputs. */
 interface AiClient {
@@ -1043,6 +1379,14 @@ interface AiClient {
     image(input: ImageInput): Promise<ImageResult>;
     embedding(input: EmbeddingInput): Promise<EmbeddingResult>;
     transcribe(input: TranscribeInput): Promise<TranscribeResult>;
+    /** OCR (F016.2) — document/image → structured markdown, billed per page. Mistral. */
+    ocr(input: OcrInput): Promise<OcrResult>;
+    /** Moderation (F016.4) — classify text against safety categories. Mistral. */
+    moderate(input: ModerationInput): Promise<ModerationResult>;
+    /** Podcast (F020) — a finished manuscript → one multi-voice audio episode. ElevenLabs. */
+    podcast(input: PodcastInput): Promise<PodcastResult>;
+    /** Single-voice TTS (F020.4) — text → audio. `voice` = curated name or voiceId. ElevenLabs. */
+    tts(input: TtsInput): Promise<PodcastResult>;
     /** Prompt-contract capabilities (F5.5) layered on chat/vision. */
     contracts: Contracts;
 }
@@ -1094,8 +1438,29 @@ declare function openrouterAdapter(config?: {
 declare function mistralAdapter(config?: {
     apiKey?: string;
     baseUrl?: string;
+    fetch?: typeof fetch;
+    pricePerPage?: number;
 }): ProviderAdapter;
 
+/** Curated Danish voices (F020.3) — friendly name → ElevenLabs voiceId. Apps can
+ *  pass these names to `ai.podcast`/`ai.tts` instead of raw IDs. */
+declare const ELEVENLABS_DANISH_VOICES: Record<string, string>;
+/** Resolve a curated voice name to its voiceId; pass a raw voiceId through unchanged. */
+declare function resolveVoice(nameOrId: string): string;
+interface ElevenLabsVoice {
+    voiceId: string;
+    name: string;
+    language?: string;
+}
+declare function elevenlabsAdapter(config?: {
+    apiKey?: string;
+    baseUrl?: string;
+    fetch?: typeof fetch;
+    pricePer1kChars?: number;
+}): ProviderAdapter & {
+    listVoices(): Promise<ElevenLabsVoice[]>;
+};
+
 interface FalAdapterConfig {
     apiKey?: string;
     /** "sync" (default — fal.run, fast models) or "queue" (queue.fal.run, polled). */
@@ -1143,8 +1508,8 @@ declare const falStubAdapter: ProviderAdapter;
  *  wires the live adapters. */
 declare const stubProviders: Record<string, ProviderAdapter>;
 
-declare const VERSION: "0.6.0";
-declare const SDK_TAG: "@broberg/ai-sdk@0.6.0";
+declare const VERSION: "0.8.0";
+declare const SDK_TAG: "@broberg/ai-sdk@0.8.0";
 
 /** 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
@@ -1337,4 +1702,4 @@ interface StreamTransportRequest extends TransportRequest {
  */
 declare function streamTransport(req: StreamTransportRequest): AsyncIterable<string>;
 
-export { type AiClient, type AiConfig, 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 CostSink, type CostSummary, DEFAULT_TIER_MAP, type DesignInput, type DesignResult, type DiscordSinkConfig, type EmbeddingInput, type EmbeddingRequest, type EmbeddingResult, type ExtractInput, type ExtractResult, type FalAdapterConfig, type HttpResponse, type ImageInput, type ImageRequest, type ImageResult, type Message, type MockupInput, type MockupResult, type OpenAICompatibleConfig, type PricingEntry, type ProviderAdapter, type RerankInput, type RerankResult, type Role, SDK_TAG, type SqliteBudgetStoreConfig, type SqliteSinkConfig, StreamHttpError, type SubprocessResponse, type Tier, type TierSpec, type Tool, type ToolCall, type TranscribeInput, type TranscribeRequest, type TranscribeResult, type TranslateInput, type TranslateResult, type Transport, type TransportRequest, type TransportResponse, type UpmetricsSinkConfig, type Usage, VERSION, type VideoInput, type VisionInput, aiConfigSchema, anthropicAdapter, anthropicApiAdapter, anthropicSubprocessAdapter, chatInputSchema, computeCost, createAI, deepinfraAdapter, defaultProviders, discordSink, embeddingInputSchema, falAdapter, falStubAdapter, freshUsage, fromProviderToolCall, geminiAdapter, getCostSummary, getPrice, httpTransport, imageInputSchema, makeContracts, makeOpenAICompatibleAdapter, messageSchema, mistralAdapter, multiSink, noopSink, openaiAdapter, openaiStubAdapter, openrouterAdapter, parseClaudeCliJson, parseJsonLoose, resolveTier, sqliteBudgetStore, sqliteSink, streamTransport, stubProviders, subprocessTransport, tierSpecSchema, toProviderTools, toolSchema, translateInputSchema, upmetricsSink, visionInputSchema };
+export { type AiClient, type AiConfig, 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 CostSink, type CostSummary, 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 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 RerankInput, type RerankResult, type Role, SDK_TAG, type SqliteBudgetStoreConfig, type SqliteSinkConfig, StreamHttpError, type SubprocessResponse, type Tier, type TierSpec, type Tool, type ToolCall, type TranscribeInput, type TranscribeRequest, type TranscribeResult, type TranslateInput, type TranslateResult, type Transport, type TransportRequest, type TransportResponse, type TtsInput, type TtsRequest, type UpmetricsSinkConfig, type Usage, VERSION, type VideoInput, type VisionInput, aiConfigSchema, anthropicAdapter, anthropicApiAdapter, anthropicSubprocessAdapter, chatInputSchema, computeCost, createAI, deepinfraAdapter, defaultProviders, discordSink, elevenlabsAdapter, embeddingInputSchema, falAdapter, falStubAdapter, freshUsage, fromProviderToolCall, geminiAdapter, getCostSummary, getPrice, httpTransport, imageInputSchema, makeContracts, makeOpenAICompatibleAdapter, messageSchema, mistralAdapter, multiSink, noopSink, openaiAdapter, openaiStubAdapter, openrouterAdapter, parseClaudeCliJson, parseJsonLoose, resolveTier, resolveVoice, sqliteBudgetStore, sqliteSink, streamTransport, stubProviders, subprocessTransport, tierSpecSchema, toProviderTools, toolSchema, translateInputSchema, upmetricsSink, visionInputSchema };