@nextblock-cms/cortex 0.15.10 → 0.16.2

package/lib/ai-vision.es.js

@@ -0,0 +1,227 @@
+import { generateText as _ } from "ai";
+import { runWithCortexAiModelFallback as w, omitUnsupportedCortexAiModelOptions as I, getHttpStatusCode as y, isOpenRouterRecoverableRoutingError as R } from "./ai-model-registry.es.js";
+import { truncateOnWordBoundary as C } from "./ai-seo-metadata.es.js";
+const S = "Cortex AI vision generation can only be imported from server-side code.";
+function L() {
+  if (!(typeof window > "u"))
+    throw new Error(S);
+}
+const p = [
+  "google/gemma-4-31b-it:free",
+  "google/gemma-4-26b-a4b-it:free",
+  "google/gemini-2.5-flash-lite",
+  "qwen/qwen3-vl-32b-instruct",
+  "openai/gpt-4o-mini",
+  "google/gemini-2.5-flash"
+], O = [
+  // Dedicated vision-language builds: "qwen3-vl-32b-instruct", "internvl3-78b".
+  /(?:^|\/)[^/]*-vl(?:[-:]|$)/i,
+  // Explicitly named vision variants: "llama-3.2-11b-vision-instruct".
+  /vision/i,
+  // Whole families that are multimodal end to end.
+  /(?:^|\/)gemini-/i,
+  /(?:^|\/)gemma-[3-9]/i,
+  /(?:^|\/)gpt-(?:4o|4\.1|5)/i,
+  /(?:^|\/)claude-/i,
+  /(?:^|\/)llama-4/i,
+  /(?:^|\/)pixtral/i,
+  /(?:^|\/)mistral-(?:small|medium|large)-3/i,
+  /(?:^|\/)nova-(?:lite|pro|premier)/i
+], f = 125, m = { max: 1e3, min: 20 }, M = 6e4, g = /* @__PURE__ */ new Set([
+  '"',
+  "'",
+  "«",
+  // «
+  "»",
+  // »
+  "‘",
+  // ‘
+  "’",
+  // ’
+  "‚",
+  // ‚
+  "“",
+  // “
+  "”",
+  // ”
+  "„",
+  // „
+  "`"
+]), v = /^(?:here(?:'s|’s| is)\s+)?(?:the\s+)?alt(?:[-\s]?text)?\s*[:–—-]\s*/i, N = /^(?:this\s+is\s+)?(?:an?|the)?\s*(?:image|photo|photograph|picture|pic|graphic|illustration|rendering|render)\s+(?:of|showing|depicting|that\s+shows)\s+/i;
+function U(e) {
+  const o = /* @__PURE__ */ new Set(), t = [];
+  for (const n of e) {
+    const r = n?.trim();
+    !r || o.has(r) || (o.add(r), t.push(r));
+  }
+  return t;
+}
+function X(e) {
+  const o = e?.trim();
+  return o ? p.includes(o) ? !0 : O.some((t) => t.test(o)) : !1;
+}
+function P(e) {
+  const o = e.requestedModelId?.trim() || null, t = e.selectedModel?.modelId?.trim() || null, n = e.credentialSource !== "env" && t && X(t) ? t : null;
+  return {
+    modelIds: U([
+      o,
+      n,
+      ...p
+    ])
+  };
+}
+function h(e) {
+  let o = e.trim();
+  for (; o.length >= 2 && g.has(o[0]) && g.has(o[o.length - 1]); )
+    o = o.slice(1, -1).trim();
+  return o;
+}
+function q(e) {
+  return !e.endsWith(".") || e.endsWith("...") || e.endsWith("…") ? e : (e.match(/[.!?](?=\s|$)/g) || []).length === 1 ? e.slice(0, -1).trimEnd() : e;
+}
+function D(e, o = f) {
+  let t = e.replace(/\s+/g, " ").trim(), n = !1;
+  for (let r = 0; r < 2; r++) {
+    t = h(t);
+    const a = t.replace(v, "");
+    a !== t && (t = a.trim(), n = !0), t = h(t);
+    const i = t.replace(N, "");
+    i !== t && (t = i.trim(), n = !0);
+  }
+  return n && t && (t = t.charAt(0).toUpperCase() + t.slice(1)), C(q(t), o);
+}
+function k(e) {
+  const o = e.trim();
+  if (!o)
+    throw new Error("Cortex AI alt text generation requires an image URL.");
+  let t;
+  try {
+    t = new URL(o);
+  } catch {
+    throw new Error(
+      `Cortex AI alt text generation requires an absolute http(s) image URL, received "${o}". A storage key or site-relative path must be resolved to a publicly fetchable URL first, because the model provider downloads the image server-side.`
+    );
+  }
+  if (t.protocol !== "http:" && t.protocol !== "https:")
+    throw new Error(
+      `Cortex AI alt text generation requires an http(s) image URL, received the "${t.protocol}" scheme.`
+    );
+  return t;
+}
+function B(e) {
+  return [
+    "You are NextBlock Cortex AI, writing the alt attribute for an image on a website.",
+    "You are writing for a person using a screen reader who cannot see the image at all.",
+    "Describe only what is visibly present in the image: subjects, their appearance, what they are doing, the setting, and any text that appears in the image.",
+    "Never state or guess anything you cannot see — no names, no locations, no dates, no brands, no emotions, and no backstory unless they are legible in the image itself.",
+    `Write one short factual description of at most ${e} characters.`,
+    'Do not begin with "image of", "photo of", "picture of", "graphic of", or any similar phrase; the screen reader already announces that this is an image.',
+    'Do not add a label such as "Alt text:" and do not wrap the description in quotation marks.',
+    "Do not keyword-stuff, do not list search terms, and do not repeat a word for emphasis.",
+    "Return plain text only: one line, no markdown, no HTML, no JSON, no commentary, no alternatives to choose between.",
+    "If surrounding page context is supplied, use it ONLY to disambiguate what you can already see — for example to choose the right word for an object or to know which product is pictured. Never use it to add detail that is not visible in the image."
+  ].join(" ");
+}
+function F(e) {
+  return [
+    `Write the alt attribute for this image, at most ${e.maxLength} characters.`,
+    e.context ? `Surrounding page context, for disambiguation only: ${e.context}` : null
+  ].filter(Boolean).join(`
+
+`);
+}
+function V(e) {
+  const o = y(e);
+  if (o === 401 || o === 402 || o === 403)
+    return !1;
+  if (R(e) || o && o >= 500)
+    return !0;
+  const t = e instanceof Error ? e.message : String(e);
+  return /NoContentGenerated|No content generated|Provider returned error|empty alt text|image|modality|multimodal|not support|unsupported|aborted|abort|timeout|timed out/i.test(
+    t
+  );
+}
+async function H(e) {
+  L();
+  const o = k(e.imageUrl), t = Number.isFinite(e.maxLength) ? Math.round(e.maxLength) : f, n = Math.min(
+    m.max,
+    Math.max(m.min, t)
+  ), r = e.context?.replace(/\s+/g, " ").trim() || null, { createCortexAiOpenRouterClient: a } = await import("./ai-client.es.js"), i = await a({
+    apiKey: e.apiKey || void 0
+  }), T = P({
+    credentialSource: i.credentialSource,
+    requestedModelId: e.modelId,
+    selectedModel: i.modelSelection
+  }), b = [
+    {
+      content: [
+        {
+          text: F({ context: r, maxLength: n }),
+          type: "text"
+        },
+        {
+          // `mediaType` is intentionally omitted. The SDK downloads the image and
+          // derives the media type from the response, whereas guessing it from the
+          // URL is unreliable for the CDN and signed-URL shapes this CMS produces —
+          // query strings, no extension, or an extension that disagrees with what
+          // the storage bucket actually serves. A wrong media type is worse than a
+          // missing one, because the provider will reject the part outright.
+          image: o,
+          type: "image"
+        }
+      ],
+      role: "user"
+    }
+  ], l = await w({
+    modelIds: T.modelIds,
+    shouldRetry: V,
+    execute: async (c) => {
+      const s = new AbortController(), x = setTimeout(
+        () => s.abort(),
+        M
+      ), d = () => s.abort();
+      e.abortSignal?.addEventListener("abort", d, { once: !0 }), e.abortSignal?.aborted && s.abort();
+      try {
+        const A = I(
+          {
+            abortSignal: s.signal,
+            // Alt text is one short line, but several models in the chain emit
+            // reasoning tokens against the same budget, so leave headroom rather
+            // than truncating the answer to nothing on those.
+            maxOutputTokens: 400,
+            maxRetries: 0,
+            messages: b,
+            system: B(n),
+            temperature: 0.2
+          },
+          {
+            modelId: c,
+            modelSelection: i.modelSelection
+          }
+        ), E = await _({
+          ...A,
+          model: i.model(c)
+        }), u = D(E.text, n);
+        if (!u)
+          throw new Error("Cortex AI returned empty alt text.");
+        return u;
+      } finally {
+        clearTimeout(x), e.abortSignal?.removeEventListener("abort", d);
+      }
+    }
+  });
+  return {
+    altText: l.result,
+    attempts: l.attempts.length,
+    credentialSource: i.credentialSource,
+    modelId: l.modelId
+  };
+}
+export {
+  f as CORTEX_AI_DEFAULT_ALT_TEXT_MAX_LENGTH,
+  p as CORTEX_AI_VISION_MODEL_FALLBACK_REGISTRY,
+  P as buildCortexAiVisionRoutingPolicy,
+  H as generateCortexAiAltText,
+  X as isKnownVisionCapableCortexAiModelId,
+  D as normalizeGeneratedAltText
+};

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nextblock-cms/cortex",
-  "version": "0.15.10",
+  "version": "0.16.2",
   "license": "AGPL-3.0-or-later",
   "publishConfig": {
     "access": "public"
@@ -28,9 +28,9 @@
   },
   "dependencies": {
     "@ai-sdk/openai-compatible": "^2.0.42",
-    "@nextblock-cms/db": "^0.15.10",
-    "@nextblock-cms/ecommerce": "npm:@nextblock-cms/ecom@^0.15.10",
-    "@nextblock-cms/utils": "^0.15.10",
+    "@nextblock-cms/db": "^0.16.2",
+    "@nextblock-cms/ecommerce": "npm:@nextblock-cms/ecom@^0.16.2",
+    "@nextblock-cms/utils": "^0.16.2",
     "ai": "^6.0.170",
     "zod": "^4.3.6"
   },