@broberg/ai-sdk 0.10.4 → 0.10.5

package/dist/index.d.ts

@@ -715,9 +715,13 @@ declare const visionInputSchema: z.ZodObject<{
     image: z.ZodUnion<[z.ZodString, z.ZodType<Uint8Array<ArrayBuffer>, z.ZodTypeDef, Uint8Array<ArrayBuffer>>]>;
     prompt: z.ZodString;
     mimeType: z.ZodOptional<z.ZodString>;
+    /** System instruction — drives instruction-following far better than stuffing
+     *  rules into `prompt` for instruction-heavy vision tasks (e.g. a JSON critic). */
+    system: z.ZodOptional<z.ZodString>;
 }, "strip", z.ZodTypeAny, {
     image: string | Uint8Array<ArrayBuffer>;
     prompt: string;
+    system?: string | undefined;
     mimeType?: string | undefined;
     purpose?: string | undefined;
     tier?: "fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | undefined;
@@ -735,6 +739,7 @@ declare const visionInputSchema: z.ZodObject<{
 }, {
     image: string | Uint8Array<ArrayBuffer>;
     prompt: string;
+    system?: string | undefined;
     mimeType?: string | undefined;
     purpose?: string | undefined;
     tier?: "fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | undefined;
@@ -783,9 +788,12 @@ declare const videoInputSchema: z.ZodObject<{
     video: z.ZodUnion<[z.ZodString, z.ZodType<Uint8Array<ArrayBuffer>, z.ZodTypeDef, Uint8Array<ArrayBuffer>>]>;
     prompt: z.ZodString;
     mimeType: z.ZodOptional<z.ZodString>;
+    /** System instruction — same instruction-following benefit as on vision. */
+    system: z.ZodOptional<z.ZodString>;
 }, "strip", z.ZodTypeAny, {
     video: string | Uint8Array<ArrayBuffer>;
     prompt: string;
+    system?: string | undefined;
     mimeType?: string | undefined;
     purpose?: string | undefined;
     tier?: "fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | undefined;
@@ -803,6 +811,7 @@ declare const videoInputSchema: z.ZodObject<{
 }, {
     video: string | Uint8Array<ArrayBuffer>;
     prompt: string;
+    system?: string | undefined;
     mimeType?: string | undefined;
     purpose?: string | undefined;
     tier?: "fast" | "smart" | "powerful" | "cheap" | "vision" | "video" | "embedding" | undefined;
@@ -1687,8 +1696,8 @@ declare const falStubAdapter: ProviderAdapter;
  *  wires the live adapters. */
 declare const stubProviders: Record<string, ProviderAdapter>;
 
-declare const VERSION: "0.10.4";
-declare const SDK_TAG: "@broberg/ai-sdk@0.10.4";
+declare const VERSION: "0.10.5";
+declare const SDK_TAG: "@broberg/ai-sdk@0.10.5";
 
 /** 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

package/dist/index.js

@@ -638,7 +638,7 @@ function makeOpenAICompatibleAdapter(config) {
     }
     const data = res.json;
     const msg = data.choices?.[0]?.message;
-    const text = msg?.content ?? "";
+    const text = contentToText(msg?.content);
     const toolCalls = msg?.tool_calls?.map(
       (tc) => fromProviderToolCall(tc, "openai")
     );
@@ -745,6 +745,17 @@ function makeOpenAICompatibleAdapter(config) {
     vision: chat
   };
 }
+function contentToText(content) {
+  if (typeof content === "string") return content;
+  if (Array.isArray(content)) {
+    return content.map((p) => {
+      if (typeof p === "string") return p;
+      const t = p?.text;
+      return typeof t === "string" ? t : "";
+    }).join("");
+  }
+  return "";
+}
 function mapFinishReason(reason) {
   switch (reason) {
     case "tool_calls":
@@ -1618,29 +1629,31 @@ var BudgetGuard = class {
 // src/capabilities/vision.ts
 var VISION_DEFAULT_TIER = "vision";
 function buildVisionMessages(input) {
-  return [
-    {
-      role: "user",
-      content: [
-        { type: "text", text: input.prompt },
-        { type: "image", image: input.image, mimeType: input.mimeType }
-      ]
-    }
-  ];
+  const messages = [];
+  if (input.system) messages.push({ role: "system", content: input.system });
+  messages.push({
+    role: "user",
+    content: [
+      { type: "text", text: input.prompt },
+      { type: "image", image: input.image, mimeType: input.mimeType }
+    ]
+  });
+  return messages;
 }
 
 // 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 }
-      ]
-    }
-  ];
+  const messages = [];
+  if (input.system) messages.push({ role: "system", content: input.system });
+  messages.push({
+    role: "user",
+    content: [
+      { type: "video", video: input.video, mimeType: input.mimeType },
+      { type: "text", text: input.prompt }
+    ]
+  });
+  return messages;
 }
 
 // src/capabilities/translate.ts
@@ -1834,12 +1847,17 @@ var visionInputSchema = z.object({
   image: z.union([z.string(), z.instanceof(Uint8Array)]),
   prompt: z.string(),
   mimeType: z.string().optional(),
+  /** System instruction — drives instruction-following far better than stuffing
+   *  rules into `prompt` for instruction-heavy vision tasks (e.g. a JSON critic). */
+  system: z.string().optional(),
   ...callOptions
 });
 var videoInputSchema = z.object({
   video: z.union([z.string(), z.instanceof(Uint8Array)]),
   prompt: z.string(),
   mimeType: z.string().optional(),
+  /** System instruction — same instruction-following benefit as on vision. */
+  system: z.string().optional(),
   ...callOptions
 });
 var translateInputSchema = z.object({
@@ -2452,8 +2470,8 @@ var stubProviders = {
 };
 
 // src/version.ts
-var VERSION = "0.10.4";
-var SDK_TAG = "@broberg/ai-sdk@0.10.4";
+var VERSION = "0.10.5";
+var SDK_TAG = "@broberg/ai-sdk@0.10.5";
 
 // src/cost/budget-store.ts
 function sqliteBudgetStore(config) {