@providerprotocol/ai 0.0.24 → 0.0.26

package/dist/google/index.d.ts

@@ -42,6 +42,17 @@ interface GoogleLLMParams {
     responseMimeType?: 'text/plain' | 'application/json';
     /** Response schema for structured output */
     responseSchema?: Record<string, unknown>;
+    /**
+     * Modalities to generate in the response.
+     *
+     * Use `['IMAGE']` or `['TEXT', 'IMAGE']` with Gemini image generation models
+     * (e.g., gemini-2.5-flash-image aka Nano Banana).
+     */
+    responseModalities?: GoogleResponseModality[];
+    /**
+     * Image generation configuration for Gemini image response modalities.
+     */
+    imageConfig?: GoogleImageConfig;
     /**
      * Presence penalty for new topics
      * Positive values encourage discussing new topics
@@ -124,6 +135,28 @@ interface GoogleLLMParams {
      */
     toolConfig?: GoogleToolConfig;
 }
+/**
+ * Output modality enum values for Gemini responseModalities.
+ *
+ * The API supports TEXT, IMAGE, and AUDIO response types. Some SDK examples
+ * use Title Case values, so both are accepted here.
+ */
+type GoogleResponseModality = 'TEXT' | 'IMAGE' | 'AUDIO' | 'Text' | 'Image' | 'Audio';
+/**
+ * Image generation configuration for Gemini response modalities.
+ */
+interface GoogleImageConfig {
+    /**
+     * Preferred aspect ratio for generated images.
+     * Example: "1:1", "9:16", "16:9".
+     */
+    aspectRatio?: string;
+    /**
+     * Preferred output size for generated images.
+     * Example: "1024x1024".
+     */
+    imageSize?: string;
+}
 /**
  * Configuration for extended thinking/reasoning in Gemini 2.5+ and 3+ models.
  *
@@ -980,4 +1013,4 @@ declare const google: Provider<unknown> & {
     };
 };
 
-export { type CacheCreateOptions, type CacheListOptions, type GoogleBuiltInTool, type GoogleCacheCreateRequest, type GoogleCacheListResponse, type GoogleCacheResponse, type GoogleCacheUpdateRequest, type GoogleCodeExecutionResult, type GoogleCodeExecutionTool, type GoogleEmbedParams, type GoogleFileSearchTool, type GoogleGroundingMetadata, type GoogleHeaders, type GoogleLLMParams, type GoogleMapsTool, type GoogleSearchTool, type GoogleTaskType, type GoogleToolConfig, type GoogleUrlContextTool, cache, google, tools };
+export { type CacheCreateOptions, type CacheListOptions, type GoogleBuiltInTool, type GoogleCacheCreateRequest, type GoogleCacheListResponse, type GoogleCacheResponse, type GoogleCacheUpdateRequest, type GoogleCodeExecutionResult, type GoogleCodeExecutionTool, type GoogleEmbedParams, type GoogleFileSearchTool, type GoogleGroundingMetadata, type GoogleHeaders, type GoogleImageConfig, type GoogleLLMParams, type GoogleMapsTool, type GoogleResponseModality, type GoogleSearchTool, type GoogleTaskType, type GoogleToolConfig, type GoogleUrlContextTool, cache, google, tools };

package/dist/google/index.js

@@ -281,6 +281,16 @@ function transformResponse(data) {
         args: fc.functionCall.args,
         thoughtSignature: fc.thoughtSignature
       });
+    } else if ("inlineData" in part) {
+      const imagePart = part;
+      const dataString = imagePart.inlineData.data;
+      if (dataString) {
+        content.push({
+          type: "image",
+          mimeType: imagePart.inlineData.mimeType ?? "image/png",
+          source: { type: "base64", data: dataString }
+        });
+      }
     } else if ("codeExecutionResult" in part) {
       const codeResult = part;
       if (codeResult.codeExecutionResult.output) {
@@ -326,6 +336,7 @@ function createStreamState() {
     reasoning: "",
     thoughtSignature: void 0,
     toolCalls: [],
+    images: [],
     finishReason: null,
     inputTokens: 0,
     outputTokens: 0,
@@ -387,6 +398,20 @@ function transformStreamChunk(chunk, state) {
           argumentsJson: JSON.stringify(fc.functionCall.args)
         }
       });
+    } else if ("inlineData" in part) {
+      const imagePart = part;
+      const dataString = imagePart.inlineData.data;
+      if (dataString) {
+        state.images.push({
+          data: dataString,
+          mimeType: imagePart.inlineData.mimeType ?? "image/png"
+        });
+        events.push({
+          type: StreamEventType.ImageDelta,
+          index: state.images.length - 1,
+          delta: { data: decodeBase64(dataString) }
+        });
+      }
     } else if ("codeExecutionResult" in part) {
       const codeResult = part;
       if (codeResult.codeExecutionResult.output) {
@@ -424,6 +449,13 @@ function buildResponseFromState(state) {
     } catch {
     }
   }
+  for (const imageData of state.images) {
+    content.push({
+      type: "image",
+      mimeType: imageData.mimeType,
+      source: { type: "base64", data: imageData.data }
+    });
+  }
   for (const tc of state.toolCalls) {
     const toolCallId = tc.id || createGoogleToolCallId(tc.name, toolCalls.length);
     toolCalls.push({
@@ -479,6 +511,14 @@ function normalizeStopReason(reason) {
       return "end_turn";
   }
 }
+function decodeBase64(base64) {
+  const binaryString = atob(base64);
+  const bytes = new Uint8Array(binaryString.length);
+  for (let i = 0; i < binaryString.length; i += 1) {
+    bytes[i] = binaryString.charCodeAt(i);
+  }
+  return bytes;
+}
 
 // src/providers/google/llm.ts
 var GOOGLE_API_BASE = "https://generativelanguage.googleapis.com/v1beta";
@@ -488,7 +528,8 @@ var GOOGLE_CAPABILITIES = {
   structuredOutput: true,
   imageInput: true,
   videoInput: true,
-  audioInput: true
+  audioInput: true,
+  imageOutput: true
 };
 function buildUrl(modelId, action) {
   return `${GOOGLE_API_BASE}/models/${modelId}:${action}`;