@broberg/ai-sdk 0.6.0 → 0.7.0

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" | "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 +180,34 @@ 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;
+}
 /** 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 +224,8 @@ 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>;
 }
 interface TranslateResult {
     text: string;
@@ -971,6 +1001,135 @@ 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 aiConfigSchema: z.ZodObject<{
     defaults: z.ZodOptional<z.ZodRecord<z.ZodEnum<["fast", "smart", "powerful", "cheap", "vision", "video", "embedding"]>, z.ZodObject<{
         provider: z.ZodString;
@@ -1029,6 +1188,8 @@ 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 AiConfig = z.infer<typeof aiConfigSchema>;
 /** The public facade. Defined here because it depends on the derived inputs. */
 interface AiClient {
@@ -1043,6 +1204,10 @@ 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>;
     /** Prompt-contract capabilities (F5.5) layered on chat/vision. */
     contracts: Contracts;
 }
@@ -1094,6 +1259,8 @@ declare function openrouterAdapter(config?: {
 declare function mistralAdapter(config?: {
     apiKey?: string;
     baseUrl?: string;
+    fetch?: typeof fetch;
+    pricePerPage?: number;
 }): ProviderAdapter;
 
 interface FalAdapterConfig {
@@ -1143,8 +1310,8 @@ 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: "0.7.0";
+declare const SDK_TAG: "@broberg/ai-sdk@0.7.0";
 
 /** 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 +1504,4 @@ 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 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 ModerationInput, type ModerationItem, type ModerationRequest, type ModerationResult, type OcrInput, type OcrPage, type OcrRequest, type OcrResult, 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 };

package/dist/index.js

@@ -278,7 +278,9 @@ var PRICING = {
   "mistral:magistral-small-latest": { inputPer1M: 0.5, outputPer1M: 1.5, version: MS },
   "mistral:devstral-latest": { inputPer1M: 0.4, outputPer1M: 2, version: MS },
   "mistral:codestral-latest": { inputPer1M: 0.3, outputPer1M: 0.9, version: MS },
-  "mistral:open-mistral-nemo": { inputPer1M: 0.15, outputPer1M: 0.15, version: MS }
+  "mistral:open-mistral-nemo": { inputPer1M: 0.15, outputPer1M: 0.15, version: MS },
+  // Moderation (F016.4) — per input token; output 0. (OCR is per-page in the adapter.)
+  "mistral:mistral-moderation-latest": { inputPer1M: 0.1, outputPer1M: 0, version: MS }
 };
 function getPrice(provider, model) {
   const exact = PRICING[`${provider}:${model}`];
@@ -1025,12 +1027,77 @@ function openrouterAdapter(config = {}) {
 }
 
 // src/providers/mistral.ts
+var MISTRAL_OCR_PRICE_PER_PAGE = 2e-3;
 function mistralAdapter(config = {}) {
-  return makeOpenAICompatibleAdapter({
-    name: "mistral",
-    baseUrl: config.baseUrl ?? "https://api.mistral.ai/v1",
-    apiKey: config.apiKey
-  });
+  const baseUrl = config.baseUrl ?? "https://api.mistral.ai/v1";
+  const base = makeOpenAICompatibleAdapter({ name: "mistral", baseUrl, apiKey: config.apiKey });
+  function key() {
+    const k = config.apiKey ?? process.env.MISTRAL_API_KEY;
+    if (!k) throw new Error("mistral adapter: API key not set (env MISTRAL_API_KEY)");
+    return k;
+  }
+  const fetchImpl = config.fetch ?? fetch;
+  async function ocr(req) {
+    const isImage = (req.mimeType ?? "").startsWith("image/");
+    const url = typeof req.document === "string" ? req.document : `data:${req.mimeType ?? "application/pdf"};base64,${Buffer.from(req.document).toString("base64")}`;
+    const document = isImage ? { type: "image_url", image_url: url } : { type: "document_url", document_url: url };
+    const res = await fetchImpl(`${baseUrl}/ocr`, {
+      method: "POST",
+      headers: { "content-type": "application/json", authorization: `Bearer ${key()}` },
+      body: JSON.stringify({ model: req.spec.model, document })
+    });
+    if (!res.ok) {
+      const body = await res.text().catch(() => "");
+      throw new Error(`mistral ocr ${res.status}: ${body.slice(0, 300)}`);
+    }
+    const data = await res.json();
+    const pages = (data.pages ?? []).map((p, i) => ({
+      index: p.index ?? i,
+      markdown: p.markdown ?? ""
+    }));
+    const pagesProcessed = data.usage_info?.pages_processed ?? pages.length;
+    const usage = freshUsage({
+      provider: "mistral",
+      model: req.spec.model,
+      transport: "http",
+      capability: "ocr",
+      inputTokens: 0,
+      outputTokens: 0
+    });
+    usage.costUsd = pagesProcessed * (config.pricePerPage ?? MISTRAL_OCR_PRICE_PER_PAGE);
+    return { pages, usage };
+  }
+  async function moderate(req) {
+    const res = await fetchImpl(`${baseUrl}/moderations`, {
+      method: "POST",
+      headers: { "content-type": "application/json", authorization: `Bearer ${key()}` },
+      body: JSON.stringify({ model: req.spec.model, input: req.input })
+    });
+    if (!res.ok) {
+      const body = await res.text().catch(() => "");
+      throw new Error(`mistral moderation ${res.status}: ${body.slice(0, 300)}`);
+    }
+    const data = await res.json();
+    const results = (data.results ?? []).map((r) => {
+      const categories = r.categories ?? {};
+      return {
+        flagged: Object.values(categories).some(Boolean),
+        categories,
+        categoryScores: r.category_scores ?? {}
+      };
+    });
+    const estIn = req.input.reduce((n, s) => n + Math.ceil(s.length / 4), 0);
+    const usage = freshUsage({
+      provider: "mistral",
+      model: req.spec.model,
+      transport: "http",
+      capability: "moderation",
+      inputTokens: data.usage?.prompt_tokens ?? estIn,
+      outputTokens: 0
+    });
+    return { results, usage };
+  }
+  return { ...base, ocr, moderate };
 }
 
 // src/providers/fal.ts
@@ -1432,6 +1499,17 @@ var transcribeInputSchema = z.object({
   durationSec: z.number().positive().optional(),
   ...callOptions
 });
+var ocrInputSchema = z.object({
+  /** A URL, data-URL, or raw bytes of the document/image. */
+  document: z.union([z.string(), z.instanceof(Uint8Array)]),
+  /** image/* → routed as an image; anything else → a document (PDF etc.). */
+  mimeType: z.string().optional(),
+  ...callOptions
+});
+var moderationInputSchema = z.object({
+  input: z.union([z.string(), z.array(z.string())]),
+  ...callOptions
+});
 var budgetSchema = z.object({
   perCallUsd: z.number().positive().optional(),
   rollingUsd: z.number().positive().optional()
@@ -1451,6 +1529,8 @@ var DEFAULT_IMAGE_SPEC = {
   model: "fal-ai/flux/schnell",
   transport: "http"
 };
+var DEFAULT_OCR_SPEC = { provider: "mistral", model: "mistral-ocr-latest", transport: "http" };
+var DEFAULT_MODERATION_SPEC = { provider: "mistral", model: "mistral-moderation-latest", transport: "http" };
 function createAI(config = {}) {
   const cfg = aiConfigSchema.parse(config);
   const providers = cfg.providers ?? defaultProviders;
@@ -1694,6 +1774,42 @@ function createAI(config = {}) {
         }
       });
     },
+    async ocr(input) {
+      input = ocrInputSchema.parse(input);
+      return runCapability({
+        primary: { ...DEFAULT_OCR_SPEC, ...input.override },
+        fallback: input.fallback,
+        capability: "ocr",
+        purpose: input.purpose,
+        labels: input.labels,
+        estIn: 0,
+        // OCR cost is per-page, not token-based
+        estOut: 0,
+        invoke: async (spec) => {
+          const adapter = pickProvider(spec.provider);
+          if (!adapter.ocr) throw new Error(`createAI: provider "${spec.provider}" does not support ocr`);
+          return adapter.ocr({ document: input.document, mimeType: input.mimeType, spec });
+        }
+      });
+    },
+    async moderate(input) {
+      input = moderationInputSchema.parse(input);
+      const items = Array.isArray(input.input) ? input.input : [input.input];
+      return runCapability({
+        primary: { ...DEFAULT_MODERATION_SPEC, ...input.override },
+        fallback: input.fallback,
+        capability: "moderation",
+        purpose: input.purpose,
+        labels: input.labels,
+        estIn: items.reduce((n, s) => n + estTokens(s), 0),
+        estOut: 0,
+        invoke: async (spec) => {
+          const adapter = pickProvider(spec.provider);
+          if (!adapter.moderate) throw new Error(`createAI: provider "${spec.provider}" does not support moderation`);
+          return adapter.moderate({ input: items, spec });
+        }
+      });
+    },
     async embedding(input) {
       input = embeddingInputSchema.parse(input);
       const tier = input.tier ?? EMBEDDING_DEFAULT_TIER;
@@ -1820,8 +1936,8 @@ var stubProviders = {
 };
 
 // src/version.ts
-var VERSION = "0.6.0";
-var SDK_TAG = "@broberg/ai-sdk@0.6.0";
+var VERSION = "0.7.0";
+var SDK_TAG = "@broberg/ai-sdk@0.7.0";
 
 // src/cost/budget-store.ts
 function sqliteBudgetStore(config) {