Update app.py

app.py

@@ -42,13 +42,18 @@ if not HF_API_TOKEN:
 
 # --- Global Vector Store and Embeddings ---
 try:
+    # Make sure to import HuggingFaceEmbeddings from the new package
+    # if you followed the previous advice to resolve the deprecation warning
+    from langchain_huggingface import HuggingFaceEmbeddings # Or keep langchain_community if you haven't migrated yet
+
     embeddings = HuggingFaceEmbeddings(model_name=HF_EMBEDDING_MODEL_ID)
     logger.info(f"Initialized HuggingFaceEmbeddings with model: {HF_EMBEDDING_MODEL_ID}")
 except Exception as e:
     logger.error(f"Failed to initialize HuggingFaceEmbeddings: {e}. Please ensure the model_id is correct and dependencies are installed.")
     embeddings = None
 
-vectorstore = DocArrayInMemorySearch(embedding_function=embeddings) if embeddings else None
+# Initialize DocArrayInMemorySearch WITHOUT the embedding_function argument here
+vectorstore = DocArrayInMemorySearch() if embeddings else None # <--- FIXED THIS LINE
 text_splitter = RecursiveCharacterTextSplitter(
     chunk_size=1000,
     chunk_overlap=200,
@@ -77,8 +82,8 @@ def add_document_to_vector_store(content: str, source: str, metadata: dict = Non
     Adds content to the global vector store.
     Chunks the content and creates LangChain Documents.
     """
-    if vectorstore is None:
-        logger.warning("Vector store not initialized. Cannot add document.")
+    if vectorstore is None or embeddings is None: # Added check for embeddings too
+        logger.warning("Vector store or embeddings not initialized. Cannot add document.")
         return
 
     try:
@@ -90,11 +95,26 @@ def add_document_to_vector_store(content: str, source: str, metadata: dict = Non
                 doc_metadata.update(metadata)
             docs.append(Document(page_content=chunk, metadata=doc_metadata))
         
-        vectorstore.add_documents(docs)
+        # Pass the embeddings function here when adding documents
+        vectorstore.add_documents(docs, embedding=embeddings) # <--- IMPORTANT: Pass embeddings here
         logger.info(f"Added {len(docs)} chunks from '{source}' to the vector store.")
     except Exception as e:
         logger.error(f"Error adding document from '{source}' to vector store: {e}")
 
+# ... (inside ResearchAgent's __call__ method) ...
+    def __call__(self, question: str) -> str:
+        logger.info(f"Received question: {question[:200]}...")
+        try:
+            global vectorstore
+            if embeddings:
+                # Re-initialize vectorstore without embedding_function
+                vectorstore = DocArrayInMemorySearch() # <--- FIXED THIS LINE as well
+                logger.info("DocArrayInMemorySearch re-initialized for new session.")
+            else:
+                logger.warning("Embeddings not initialized, cannot re-initialize DocArrayInMemorySearch.")
+                return "Error: Embedding model not loaded, cannot process request."
+                
+
 # --- Enhanced Tools ---
 class WikiSearchTool(Tool):
     """Enhanced Wikipedia search with better formatting and error handling"""