@posthog/ai 7.2.1 → 7.2.2

package/dist/index.d.ts

@@ -206,6 +206,7 @@ declare class WrappedModels {
     constructor(client: GoogleGenAI, phClient: PostHog);
     generateContent(params: GenerateContentParameters & MonitoringParams): Promise<GenerateContentResponse>;
     generateContentStream(params: GenerateContentParameters & MonitoringParams): AsyncGenerator<GenerateContentResponse, void, unknown>;
+    private formatPartsAsContentBlocks;
     private formatInput;
     private extractSystemInstruction;
     private formatInputForPostHog;

package/dist/index.mjs

@@ -6,7 +6,7 @@ import { wrapLanguageModel } from 'ai';
 import AnthropicOriginal from '@anthropic-ai/sdk';
 import { GoogleGenAI } from '@google/genai';
 
-var version = "7.2.1";
+var version = "7.2.2";
 
 // Type guards for safer type checking
 const isString = value => {
@@ -113,6 +113,13 @@ const formatResponseOpenAI = response => {
             });
           }
         }
+        // Handle audio output (gpt-4o-audio-preview)
+        if (choice.message.audio) {
+          content.push({
+            type: 'audio',
+            ...choice.message.audio
+          });
+        }
       }
       if (content.length > 0) {
         output.push({
@@ -194,6 +201,19 @@ const formatResponseGemini = response => {
                 arguments: part.functionCall.args
               }
             });
+          } else if (part.inlineData) {
+            // Handle audio/media inline data
+            const mimeType = part.inlineData.mimeType || 'audio/pcm';
+            let data = part.inlineData.data;
+            // Handle binary data (Buffer/Uint8Array -> base64)
+            if (data instanceof Uint8Array || Buffer.isBuffer(data)) {
+              data = Buffer.from(data).toString('base64');
+            }
+            content.push({
+              type: 'audio',
+              mime_type: mimeType,
+              data: data
+            });
           }
         }
         if (content.length > 0) {
@@ -598,6 +618,13 @@ function formatOpenAIResponsesInput(input, instructions) {
 
 const REDACTED_IMAGE_PLACEHOLDER = '[base64 image redacted]';
 // ============================================
+// Multimodal Feature Toggle
+// ============================================
+const isMultimodalEnabled = () => {
+  const val = process.env._INTERNAL_LLMA_MULTIMODAL || '';
+  return val.toLowerCase() === 'true' || val === '1' || val.toLowerCase() === 'yes';
+};
+// ============================================
 // Base64 Detection Helpers
 // ============================================
 const isBase64DataUrl = str => {
@@ -622,6 +649,7 @@ const isRawBase64 = str => {
   return str.length > 20 && /^[A-Za-z0-9+/]+=*$/.test(str);
 };
 function redactBase64DataUrl(str) {
+  if (isMultimodalEnabled()) return str;
   if (!isString(str)) return str;
   // Check for data URL format
   if (isBase64DataUrl(str)) {
@@ -672,12 +700,21 @@ const sanitizeOpenAIImage = item => {
       }
     };
   }
+  // Handle audio format
+  if (item.type === 'audio' && 'data' in item) {
+    if (isMultimodalEnabled()) return item;
+    return {
+      ...item,
+      data: REDACTED_IMAGE_PLACEHOLDER
+    };
+  }
   return item;
 };
 const sanitizeAnthropicImage = item => {
+  if (isMultimodalEnabled()) return item;
   if (!isObject(item)) return item;
-  // Handle Anthropic's image format
-  if (item.type === 'image' && 'source' in item && isObject(item.source) && item.source.type === 'base64' && 'data' in item.source) {
+  // Handle Anthropic's image and document formats (same structure, different type field)
+  if ((item.type === 'image' || item.type === 'document') && 'source' in item && isObject(item.source) && item.source.type === 'base64' && 'data' in item.source) {
     return {
       ...item,
       source: {
@@ -689,8 +726,9 @@ const sanitizeAnthropicImage = item => {
   return item;
 };
 const sanitizeGeminiPart = part => {
+  if (isMultimodalEnabled()) return part;
   if (!isObject(part)) return part;
-  // Handle Gemini's inline data format
+  // Handle Gemini's inline data format (images, audio, PDFs all use inlineData)
   if ('inlineData' in part && isObject(part.inlineData) && 'data' in part.inlineData) {
     return {
       ...part,
@@ -735,6 +773,7 @@ const sanitizeLangChainImage = item => {
   }
   // Anthropic style
   if (item.type === 'image' && 'source' in item && isObject(item.source) && 'data' in item.source) {
+    if (isMultimodalEnabled()) return item;
     return {
       ...item,
       source: {
@@ -2790,6 +2829,39 @@ class WrappedModels {
       throw error;
     }
   }
+  formatPartsAsContentBlocks(parts) {
+    const blocks = [];
+    for (const part of parts) {
+      // Handle dict/object with text field
+      if (part && typeof part === 'object' && 'text' in part && part.text) {
+        blocks.push({
+          type: 'text',
+          text: String(part.text)
+        });
+      }
+      // Handle string parts
+      else if (typeof part === 'string') {
+        blocks.push({
+          type: 'text',
+          text: part
+        });
+      }
+      // Handle inlineData (images, audio, PDFs)
+      else if (part && typeof part === 'object' && 'inlineData' in part) {
+        const inlineData = part.inlineData;
+        const mimeType = inlineData.mimeType || inlineData.mime_type || '';
+        const contentType = mimeType.startsWith('image/') ? 'image' : 'document';
+        blocks.push({
+          type: contentType,
+          inline_data: {
+            data: inlineData.data,
+            mime_type: mimeType
+          }
+        });
+      }
+    }
+    return blocks;
+  }
   formatInput(contents) {
     if (typeof contents === 'string') {
       return [{
@@ -2814,20 +2886,24 @@ class WrappedModels {
             };
           }
           if ('content' in obj && obj.content) {
+            // If content is a list, format it as content blocks
+            if (Array.isArray(obj.content)) {
+              const contentBlocks = this.formatPartsAsContentBlocks(obj.content);
+              return {
+                role: isString(obj.role) ? obj.role : 'user',
+                content: contentBlocks
+              };
+            }
             return {
               role: isString(obj.role) ? obj.role : 'user',
               content: obj.content
             };
           }
           if ('parts' in obj && Array.isArray(obj.parts)) {
+            const contentBlocks = this.formatPartsAsContentBlocks(obj.parts);
             return {
               role: isString(obj.role) ? obj.role : 'user',
-              content: obj.parts.map(part => {
-                if (part && typeof part === 'object' && 'text' in part) {
-                  return part.text;
-                }
-                return part;
-              })
+              content: contentBlocks
             };
           }
         }
@@ -3405,7 +3481,7 @@ var BaseCallbackHandler = class extends BaseCallbackHandlerMethodsClass {
 	}
 	static fromMethods(methods) {
 		class Handler extends BaseCallbackHandler {
-			name = uuid.v4();
+			name = uuid.v7();
 			constructor() {
 				super();
 				Object.assign(this, methods);