@broberg/ai-sdk 0.5.1 → 0.7.0

package/dist/index.js

@@ -5,6 +5,8 @@ var DEFAULT_TIER_MAP = {
   powerful: { provider: "anthropic", model: "claude-opus-4-8", transport: "http" },
   cheap: { provider: "anthropic", model: "claude-haiku-4-5", transport: "subprocess" },
   vision: { provider: "anthropic", model: "claude-sonnet-4-6", transport: "http" },
+  // Native video understanding — Gemini leads; flash-lite is the cheap default (F019).
+  video: { provider: "gemini", model: "gemini-2.5-flash-lite", transport: "http" },
   embedding: { provider: "openai", model: "text-embedding-3-small", transport: "http" }
 };
 function resolveTier(tier, override, configMap) {
@@ -205,6 +207,7 @@ function parseArgs(raw) {
 
 // src/cost/pricing.ts
 var V = "2026-06-02";
+var MS = "2026-06-04-mistral.ai";
 var PRICING = {
   // Anthropic (direct API). DEFAULT_TIER_MAP: fast/cheap=haiku, smart/vision=sonnet, powerful=opus.
   "anthropic:claude-haiku-4-5": {
@@ -236,19 +239,48 @@ var PRICING = {
   // Whisper is priced per minute, not per token — not representable here; transcribe
   // (F5.6) computes its own cost. Listed as 0 so token-based compute never charges it.
   "openai:whisper-1": { inputPer1M: 0, outputPer1M: 0, version: V },
-  // OpenRouter (meta-router — model slugs include the upstream vendor).
-  "openrouter:anthropic/claude-sonnet-4-6": { inputPer1M: 3, outputPer1M: 15, version: V },
-  "openrouter:anthropic/claude-haiku-4-5": { inputPer1M: 0.8, outputPer1M: 4, version: V },
+  // OpenRouter (meta-router — model slugs include the upstream vendor). Slugs use
+  // dots (claude-sonnet-4.6) to match OpenRouter's live ids; the dashed forms
+  // never matched a real call. Caught by the F014 catalogue research.
+  "openrouter:anthropic/claude-sonnet-4.6": { inputPer1M: 3, outputPer1M: 15, version: V },
+  // OpenRouter ground-truth $1/$5 — a markup over Anthropic-direct's $0.8/$4
+  // (the `anthropic:` entry above). Was masked while the slug used dashes.
+  "openrouter:anthropic/claude-haiku-4.5": { inputPer1M: 1, outputPer1M: 5, version: "2026-06-04" },
   "openrouter:google/gemini-2.5-flash": { inputPer1M: 0.3, outputPer1M: 2.5, version: V },
+  // Ground-truth from OpenRouter /api/v1/models (was a 0.3 estimate; now 0.279).
   "openrouter:minimax/minimax-m2.7": {
-    inputPer1M: 0.3,
+    inputPer1M: 0.279,
     outputPer1M: 1.2,
-    version: `${V}-estimate`
+    version: "2026-06-04"
   },
   // Google Gemini (direct). Provider key is "gemini" — matches the adapter's
   // usage.provider + the override.provider callers pass. (Image-gen models are
   // priced per-image in the adapter, not here.)
-  "gemini:gemini-2.5-flash": { inputPer1M: 0.3, outputPer1M: 2.5, version: V }
+  "gemini:gemini-2.5-flash": { inputPer1M: 0.3, outputPer1M: 2.5, version: V },
+  // flash-lite is the default `video` tier (F019) — cheap native video understanding.
+  "gemini:gemini-2.5-flash-lite": { inputPer1M: 0.1, outputPer1M: 0.4, version: "2026-06-04-or-xref" },
+  // Mistral (direct, La Plateforme). Official prices from mistral.ai/pricing
+  // (2026-06-04, per Christian's CD report). EU/Paris-hosted — the designated
+  // GDPR-safe provider for client/personal-data workloads (see F015). NB:
+  // medium-3.5 is the premium "Vibe" coding tier ($1.5/$7.5); Large 3 ($0.5/$1.5)
+  // is the cheaper frontier general-purpose model despite the higher number.
+  "mistral:mistral-large-latest": { inputPer1M: 0.5, outputPer1M: 1.5, version: MS },
+  "mistral:mistral-large-2512": { inputPer1M: 0.5, outputPer1M: 1.5, version: MS },
+  "mistral:mistral-medium-latest": { inputPer1M: 1.5, outputPer1M: 7.5, version: MS },
+  "mistral:mistral-medium-3.5": { inputPer1M: 1.5, outputPer1M: 7.5, version: MS },
+  "mistral:mistral-medium-3": { inputPer1M: 0.4, outputPer1M: 2, version: "2026-06-04-or-xref" },
+  "mistral:mistral-small-latest": { inputPer1M: 0.1, outputPer1M: 0.3, version: MS },
+  "mistral:mistral-small-2603": { inputPer1M: 0.1, outputPer1M: 0.3, version: MS },
+  "mistral:ministral-3b-latest": { inputPer1M: 0.1, outputPer1M: 0.1, version: MS },
+  "mistral:ministral-8b-latest": { inputPer1M: 0.15, outputPer1M: 0.15, version: MS },
+  "mistral:ministral-14b-latest": { inputPer1M: 0.2, outputPer1M: 0.2, version: MS },
+  "mistral:magistral-medium-latest": { inputPer1M: 2, outputPer1M: 5, version: MS },
+  "mistral:magistral-small-latest": { inputPer1M: 0.5, outputPer1M: 1.5, version: MS },
+  "mistral:devstral-latest": { inputPer1M: 0.4, outputPer1M: 2, version: MS },
+  "mistral:codestral-latest": { inputPer1M: 0.3, outputPer1M: 0.9, version: MS },
+  "mistral:open-mistral-nemo": { inputPer1M: 0.15, outputPer1M: 0.15, version: MS },
+  // Moderation (F016.4) — per input token; output 0. (OCR is per-page in the adapter.)
+  "mistral:mistral-moderation-latest": { inputPer1M: 0.1, outputPer1M: 0, version: MS }
 };
 function getPrice(provider, model) {
   const exact = PRICING[`${provider}:${model}`];
@@ -302,6 +334,9 @@ function contentBlocks(content) {
   if (typeof content === "string") return content;
   return content.map((p) => {
     if (p.type === "text") return { type: "text", text: p.text };
+    if (p.type === "video") {
+      throw new Error("anthropic adapter: video input is not supported \u2014 use a video provider (e.g. gemini)");
+    }
     if (typeof p.image === "string" && /^https?:\/\//.test(p.image)) {
       return { type: "image", source: { type: "url", url: p.image } };
     }
@@ -552,6 +587,10 @@ function toOpenAIMessage(m) {
   }
   const content = m.content.map((p) => {
     if (p.type === "text") return { type: "text", text: p.text };
+    if (p.type === "video") {
+      const url2 = typeof p.video === "string" ? p.video : `data:${p.mimeType ?? "video/mp4"};base64,${Buffer.from(p.video).toString("base64")}`;
+      return { type: "video_url", video_url: { url: url2 } };
+    }
     const url = typeof p.image === "string" ? p.image : `data:${p.mimeType ?? "image/png"};base64,${Buffer.from(p.image).toString("base64")}`;
     return { type: "image_url", image_url: { url } };
   });
@@ -786,6 +825,10 @@ function partsFrom(content) {
   if (typeof content === "string") return [{ text: content }];
   return content.map((p) => {
     if (p.type === "text") return { text: p.text };
+    if (p.type === "video") {
+      const data2 = typeof p.video === "string" ? p.video.replace(/^data:[^;]+;base64,/, "") : Buffer.from(p.video).toString("base64");
+      return { inlineData: { mimeType: p.mimeType ?? "video/mp4", data: data2 } };
+    }
     const data = typeof p.image === "string" ? p.image.replace(/^data:[^;]+;base64,/, "") : Buffer.from(p.image).toString("base64");
     return { inlineData: { mimeType: p.mimeType ?? "image/png", data } };
   });
@@ -983,6 +1026,80 @@ function openrouterAdapter(config = {}) {
   });
 }
 
+// src/providers/mistral.ts
+var MISTRAL_OCR_PRICE_PER_PAGE = 2e-3;
+function mistralAdapter(config = {}) {
+  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
 var FAL_IMAGE_PRICE_ESTIMATE = {
   "fal-ai/flux/schnell": 3e-3,
@@ -1070,6 +1187,7 @@ var defaultProviders = {
   gemini: geminiAdapter(),
   deepinfra: deepinfraAdapter(),
   openrouter: openrouterAdapter(),
+  mistral: mistralAdapter(),
   fal: falAdapter()
 };
 
@@ -1144,6 +1262,20 @@ function buildVisionMessages(input) {
   ];
 }
 
+// src/capabilities/video.ts
+var VIDEO_DEFAULT_TIER = "video";
+function buildVideoMessages(input) {
+  return [
+    {
+      role: "user",
+      content: [
+        { type: "video", video: input.video, mimeType: input.mimeType },
+        { type: "text", text: input.prompt }
+      ]
+    }
+  ];
+}
+
 // src/capabilities/translate.ts
 var TRANSLATE_DEFAULT_TIER = "fast";
 var TRANSLATE_SYSTEM = "You are a translation engine. Translate the user's text only. Return the translation and nothing else \u2014 no preamble, no quotes.";
@@ -1279,6 +1411,7 @@ var tierSchema = z.enum([
   "powerful",
   "cheap",
   "vision",
+  "video",
   "embedding"
 ]);
 var tierSpecSchema = z.object({
@@ -1336,6 +1469,12 @@ var visionInputSchema = z.object({
   mimeType: z.string().optional(),
   ...callOptions
 });
+var videoInputSchema = z.object({
+  video: z.union([z.string(), z.instanceof(Uint8Array)]),
+  prompt: z.string(),
+  mimeType: z.string().optional(),
+  ...callOptions
+});
 var translateInputSchema = z.object({
   text: z.string(),
   to: z.string(),
@@ -1360,6 +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()
@@ -1379,6 +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;
@@ -1561,6 +1713,27 @@ function createAI(config = {}) {
         }
       });
     },
+    async video(input) {
+      input = videoInputSchema.parse(input);
+      const tier = input.tier ?? VIDEO_DEFAULT_TIER;
+      const messages = buildVideoMessages(input);
+      return runCapability({
+        primary: resolveTier(tier, input.override, cfg.defaults),
+        fallback: input.fallback,
+        capability: "video",
+        tier,
+        purpose: input.purpose,
+        labels: input.labels,
+        estIn: estTokens(input.prompt) + 4e3,
+        // prompt + video tokens (native video ≈ frames)
+        estOut: 512,
+        invoke: async (spec) => {
+          const adapter = pickProvider(spec.provider);
+          if (!adapter.vision) throw new Error(`createAI: provider "${spec.provider}" does not support video`);
+          return adapter.vision({ messages, spec });
+        }
+      });
+    },
     async translate(input) {
       input = translateInputSchema.parse(input);
       const tier = input.tier ?? TRANSLATE_DEFAULT_TIER;
@@ -1601,6 +1774,42 @@ function createAI(config = {}) {
         }
       });
     },
+    async ocr(input) {
+      input = ocrInputSchema.parse(input);
+      return runCapability({
+        primary: { ...DEFAULT_OCR_SPEC, ...input.override },
+        fallback: input.fallback,
+        capability: "ocr",
+        purpose: input.purpose,
+        labels: input.labels,
+        estIn: 0,
+        // OCR cost is per-page, not token-based
+        estOut: 0,
+        invoke: async (spec) => {
+          const adapter = pickProvider(spec.provider);
+          if (!adapter.ocr) throw new Error(`createAI: provider "${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;
@@ -1727,8 +1936,8 @@ var stubProviders = {
 };
 
 // src/version.ts
-var VERSION = "0.5.1";
-var SDK_TAG = "@broberg/ai-sdk@0.5.1";
+var VERSION = "0.7.0";
+var SDK_TAG = "@broberg/ai-sdk@0.7.0";
 
 // src/cost/budget-store.ts
 function sqliteBudgetStore(config) {
@@ -1986,6 +2195,7 @@ export {
   makeContracts,
   makeOpenAICompatibleAdapter,
   messageSchema,
+  mistralAdapter,
   multiSink,
   noopSink,
   openaiAdapter,