Support for multiple files in RAG virtual assistant

demos/virtual_ai_assistant_demo/main.py

@@ -124,39 +124,45 @@ def load_chat_models(chat_model_name: str, embedding_model_name: str, personalit
                                                     memory=ChatMemoryBuffer.from_defaults())
 
 
-def load_file(file_path: Path) -> Document:
-    ext = file_path.suffix
-    if ext == ".pdf":
-        # Using PyMuPDF (fitz) to read PDF content
-        text = ""
-        with fitz.open(file_path) as pdf:
-            for page in pdf:
-                text += page.get_text("text") + "\n"  # Extract text from each page
-            return Document(text=text, metadata={"file_name": file_path.name})
-
-    elif ext == ".txt":
-        # Reading text files as usual
-        with open(file_path) as f:
-            content = f.read()
-            return Document(text=content, metadata={"file_name": file_path.name})
-
-    else:
-        raise ValueError(f"{ext} file is not supported for now")
-
-
-def load_context(file_path: str) -> None:
+def load_files(file_paths: List[str]) -> list[Document]:
+    documents = []
+    for file_path in map(lambda x: Path(x), file_paths):
+        ext = file_path.suffix
+        if ext == ".pdf":
+            # Using PyMuPDF (fitz) to read PDF content
+            text = ""
+            with fitz.open(file_path) as pdf:
+                for page in pdf:
+                    text += page.get_text("text") + "\n"  # Extract text from each page
+                # remove non-breaking space
+                text.replace("\xa0", " ")
+                documents.append(Document(text=text, metadata={"file_name": file_path.name}))
+
+        elif ext == ".txt":
+            # Reading text files as usual
+            with open(file_path) as f:
+                content = f.read()
+                documents.append(Document(text=content, metadata={"file_name": file_path.name}))
+
+        else:
+            log.warning(f"{ext} file is not supported for now. Skipping {file_path.name}")
+
+    return documents
+
+
+def load_context(file_paths: List[str]) -> None:
     global ov_chat_engine
 
     # limit chat history to 3000 tokens
     memory = ChatMemoryBuffer.from_defaults()
 
-    if not file_path:
+    if not file_paths:
         ov_chat_engine = SimpleChatEngine.from_defaults(llm=ov_llm, system_prompt=chatbot_config["system_configuration"], memory=memory)
         return
 
-    document = load_file(Path(file_path))
+    documents = load_files(file_paths)
     Settings.embed_model = ov_embedding
-    index = VectorStoreIndex.from_documents([document])
+    index = VectorStoreIndex.from_documents(documents)
     ov_chat_engine = index.as_chat_engine(llm=ov_llm, chat_mode=ChatMode.CONTEXT, system_prompt=chatbot_config["system_configuration"],
                                           memory=memory)
 
@@ -201,7 +207,7 @@ def create_UI(initial_message: str, action_name: str) -> gr.Blocks:
         gr.Markdown(chatbot_config["instructions"])
 
         with gr.Row():
-            file_uploader_ui = gr.File(label="Additional context", file_types=[".pdf", ".txt"], scale=1)
+            file_uploader_ui = gr.Files(label="Additional context", file_types=[".pdf", ".txt"], scale=1)
             with gr.Column(scale=4):
                 chatbot_ui = gr.Chatbot(value=[[None, initial_message]], label="Chatbot")
                 with gr.Row():