Added baseline UI using textual

pyproject.toml

@@ -13,7 +13,8 @@ dependencies = [
     "sentence-transformers>=3.3.1",
     "transformers>=4.48.0",
     "accelerate>=1.2.1",
-    "gnews>=0.3.9"
+    "gnews>=0.3.9",
+    "textual>=1.0.0"
 ]
 
 [project.optional-dependencies]

src/news_ask_ai/__init__.py

@@ -1,25 +1,10 @@
-from news_ask_ai.ask.news_ask_engine import NewsAskEngine
-
+from news_ask_ai.ask.news_ask_ui import NewsAskUI
 from news_ask_ai.utils.logger import setup_logger
 
 logger = setup_logger()
 
 
 def main() -> None:
     logger.info("Initializing RAG engine...")
-    search_engine = NewsAskEngine(collection_name="news-collection")
-
-    print("Indicate the topic of the news you want to ask about")
-    topic_input = input("\nYour topic: ")
-    search_engine.ingest_data(topic_input)
-
-    while True:
-        print("What do you want to ask about?")
-        question_input = input("\nYour question: ")
-        if question_input.lower() == "exit":
-            break
-
-        completion = search_engine.get_completions(question_input)
-        print(completion)
-
-    print("Thank you for using the NewsAskAI system. Goodbye!")
+    app = NewsAskUI()
+    app.run()

src/news_ask_ai/ask/news_ask_ui.py

@@ -0,0 +1,120 @@
+from textual.app import App, ComposeResult
+from textual.widgets import Footer, Header, Input, Button, Static
+from textual.containers import Horizontal, Container
+
+from textual.widget import Widget
+
+from news_ask_ai.ask.news_ask_engine import NewsAskEngine
+from news_ask_ai.utils.logger import setup_logger
+
+logger = setup_logger()
+
+
+class ConversationalContainer(Container, can_focus=True):
+    """Conversational container widget."""
+
+
+class MessageBox(Widget):
+    def __init__(self, text: str, role: str) -> None:
+        self.text = text
+        self.role = role
+        super().__init__()
+
+    def compose(self) -> ComposeResult:
+        yield Static(self.text, classes=f"message {self.role}")
+
+
+class NewsAskUI(App):  # type: ignore
+    TITLE = "NewsAskAI"
+    SUB_TITLE = "Ask questions about the news directly in your terminal"
+    CSS_PATH = "../static/css/styles.css"
+
+    def __init__(self) -> None:
+        super().__init__()
+        logger.info("Initializing RAG engine...")
+        self.search_engine = NewsAskEngine(collection_name="news-collection")
+
+    def compose(self) -> ComposeResult:
+        yield Header()
+
+        # -- Container for topic input and ingest button.
+        with Container(id="news_topic_container"):
+            yield MessageBox(
+                "Welcome to NewsAskAI!\n"
+                "Get the latest news insights with just a few clicks.\n"
+                "Enter a topic of interest below and press 'Ingest news' to start.\n"
+                "Need help? Use the commands at the bottom of your screen.",
+                role="info",
+            )
+
+            with Horizontal(id="input_box"):
+                yield Input(placeholder="Enter news topic...", id="news_topic_input")
+                yield Button(label="Ingest news", variant="success", id="news_ingest_button")
+
+        # -- Container for conversation UI, hidden by default.
+        with Container(id="conversation_container"):
+            with ConversationalContainer(id="conversation_box"):
+                yield MessageBox(
+                    "You're all set to explore the news!\n"
+                    "Type a question about the ingested news and press 'Ask' to get your answer.\n"
+                    "Wait for the response...\n"
+                    "Need assistance? Commands are listed at the bottom.",
+                    role="info",
+                )
+
+                with Horizontal(id="input_box"):
+                    yield Input(placeholder="Enter your question", id="conversation_input")
+                    yield Button(label="Ask", variant="success", id="conversation_button")
+
+        yield Footer()
+
+    def on_mount(self) -> None:
+        """Called when app is first mounted."""
+        self.query_one("#news_topic_input", Input).focus()
+
+        conversation_container = self.query_one("#conversation_container")
+        conversation_container.display = False
+
+    async def on_button_pressed(self, event: Button.Pressed) -> None:
+        """Handle button pressed events."""
+        button = event.button
+
+        if button.id == "news_ingest_button":
+            news_topic_input = self.query_one("#news_topic_input", Input)
+            if not news_topic_input.value.strip():
+                return
+
+            conversation_container = self.query_one("#news_topic_container")
+            conversation_container.display = False
+
+            conversation_container = self.query_one("#conversation_container")
+            conversation_container.display = True
+
+            self.search_engine.ingest_data(news_topic_input.value)
+
+        elif button.id == "conversation_button":
+            await self.conversation()
+
+    async def conversation(self) -> None:
+        message_input = self.query_one("#conversation_input", Input)
+        if message_input.value == "":
+            return
+
+        conversation_box = self.query_one("#conversation_box")
+
+        message_box = MessageBox(message_input.value, "question")
+        conversation_box.mount(message_box)
+        conversation_box.scroll_end(animate=True)
+
+        # Clean up the input without triggering events
+        with message_input.prevent(Input.Changed):
+            message_input.value = ""
+
+        completion = self.search_engine.get_completions(message_box.text)
+
+        conversation_box.mount(
+            MessageBox(
+                text=completion,
+                role="answer",
+            )
+        )

src/news_ask_ai/services/llm_completion_service.py

@@ -1,3 +1,4 @@
+import logging
 from typing import cast
 
 import torch
@@ -11,6 +12,9 @@
 
 logger = setup_logger()
 
+logging.getLogger("transformers").setLevel(logging.ERROR)
+logging.getLogger("torch").setLevel(logging.ERROR)
+
 # Input Format:
 # The phi-4 model is best suited for prompts formatted as a chat, using special tokens.
 # Each message should follow this structure:
@@ -36,6 +40,7 @@ def __init__(self, model_name: str = "microsoft/Phi-3.5-mini-instruct") -> None:
 
         torch.random.manual_seed(0)
         self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
         logger.info(f"Using device: {self.device}")
 
         model_kwargs = {

src/news_ask_ai/static/css/styles.css

@@ -0,0 +1,70 @@
+Screen {
+    background: #212529;
+}
+
+MessageBox {
+    layout: horizontal;
+    height: auto;
+    align-horizontal: center;
+}
+
+.message {
+    width: 50%;
+    min-width: 25%;
+    max-width: 50%;
+    border: tall black;
+    padding: 1 3;
+    margin: 1 0;
+    background: #343a40;
+}
+
+.info {
+    width: auto;
+    text-align: center;
+}
+
+.question {
+    margin: 1 0 1 25;
+}
+
+.answer {
+    margin: 1 25 1 0;
+}
+
+#conversation_box {
+    overflow-y: auto;
+    height: 100%;
+}
+
+#input_box {
+    dock: bottom;
+    height: auto;
+    width: 100%;
+    margin: 0 0 2 0;
+    align_horizontal: center;
+}
+
+/* News style */
+
+#news_topic_input {
+    width: 50%;
+    background: #343a40;
+}
+
+#news_ingest_button {
+    width: auto;
+}
+
+/* Conversation style */
+
+#conversation_input {
+    width: 50%;
+    background: #343a40;
+}
+
+
+#conversation_button {
+    width: auto;
+}
+
+

src/news_ask_ai/utils/logger.py

@@ -1,7 +1,10 @@
 import logging
+from logging.handlers import RotatingFileHandler
 
 
-def setup_logger(level: int = logging.INFO) -> logging.Logger:
+
+
+def setup_logger(level: int = logging.INFO, log_file: str = "app.log") -> logging.Logger:
     """
     Set up a logger with the specified name, log file, and level.
 
@@ -13,22 +16,18 @@ def setup_logger(level: int = logging.INFO) -> logging.Logger:
     Returns:
         logging.Logger: Configured logger.
     """
-    # Create a logger
     logger = logging.getLogger()
     if logger.hasHandlers():
         logger.handlers.clear()
 
     logger.setLevel(level)
 
-    # Create a console handler for outputting logs to the console
-    console_handler = logging.StreamHandler()
-    console_handler.setLevel(level)
+    file_handler = RotatingFileHandler(log_file, maxBytes=1024 * 1024, backupCount=5)
+    file_handler.setLevel(level)
 
-    # Define a log format
     formatter = logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")
-    console_handler.setFormatter(formatter)
+    file_handler.setFormatter(formatter)
 
-    # Add handlers to the logger
-    logger.addHandler(console_handler)
+    logger.addHandler(file_handler)
 
     return logger