Update app.py

app.py

@@ -52,7 +52,7 @@ def chunk_text(text, chunk_size=500):
     return chunks
 
 def get_document_embeddings(documents):
-    """Compute embeddings for documents, using cache if available."""
+    """Compute embeddings for documents, using cache if available, and return a stacked tensor."""
     embeddings = []
     for doc in documents:
         if doc in embedding_cache:
@@ -61,7 +61,8 @@ def get_document_embeddings(documents):
             emb = retriever_model.encode(doc, convert_to_tensor=True)
             embedding_cache[doc] = emb
             embeddings.append(emb)
-    return embeddings
+    # Stack the list of tensors into a single tensor of shape (n_docs, embedding_dim)
+    return torch.stack(embeddings)
 
 def rag_pipeline(question, pdf_files):
     """Optimized RAG pipeline with caching, chunking, and improved retrieval."""
@@ -130,4 +131,4 @@ with gr.Blocks() as demo:
     
     submit_button.click(fn=rag_pipeline, inputs=[question_input, pdf_input], outputs=response_output)
 
-demo.launch(debug = True)
+demo.launch(__debug__=True)