Add AGA-WebUI (#89)

Co-authored-by: Yuan Tian <[email protected]>
Co-authored-by: Anirudh Dagar <[email protected]>

.gitignore

@@ -5,12 +5,15 @@
 .aim/
 AutogluonModels/
 data/
+user_data/
+sample_dataset/
 results/
 cookies.json
 cdk.out/
 .metaflow
 *.feather
 .langchain.db
+output_*.csv
 
 # Byte-compiled / optimized / DLL files
 __pycache__/

README.md

@@ -93,3 +93,29 @@ aga ./data ./config --output-filename my_output.csv --config-overrides "autogluo
 ```
 
 `autogluon-assistant-tools` provides more functionality and utilities for benchmarking, wrapped around autogluon-assistant. Please check out the [repo](https://github.com/autogluon/autogluon-assistant-tools/) for more details.
+
+## Autogluon Assistant Web UI
+The Autogluon Assistant Web UI is a user-friendly application that allows users to leverage the capabilities of the Autogluon-Assistant library through an intuitive web interface.
+
+The web UI enables users to upload datasets, configure Autogluon-Assistant runs with customized settings, preview data, monitor execution progress, view and download results, and supports secure, isolated sessions for concurrent users.
+
+#### To run the Autogluon Assistant Web UI:
+Navigate to the project directory and run the app:
+````
+cd src/autogluon_assistant/ui && streamlit run app.py
+````
+The Autogluon Assistant Web UI should now be accessible in your web browser at `http://localhost:8501`
+
+#### Add GPT4 Model to the LLM Option:
+If you’d like to add additional GPT4 model to the language model (LLM) dropdown:
+
+1. Navigate to src/autogluon_assistant/WebUI/constants.py
+
+2. Locate the `LLM_OPTIONS` variable, which looks like this:
+````
+LLM_OPTIONS = ["Claude 3.5 with Amazon Bedrock"]
+````
+3. Add "GPT 4o" to the list
+````
+LLM_OPTIONS = ["Claude 3.5 with Amazon Bedrock", "GPT 4o"]
+````

pyproject.toml

@@ -28,7 +28,12 @@ dependencies = [
     "sentence-transformers>=3.1.0",
     "tenacity>=8.2.2,<10.0",
     "gensim>=4.3",
-    "pandas>=2.2"
+    "pandas>=2.2",
+    "stqdm==0.0.5",
+    "streamlit==1.37.1",
+    "streamlit-aggrid==1.0.5",
+    "streamlit-extras==0.4.7",
+    "psutil>=5.9.8",
 ]
 
 [project.optional-dependencies]

src/autogluon_assistant/ui/.streamlit/config.toml

@@ -0,0 +1,10 @@
+[server]
+maxUploadSize = 4096
+enableStaticServing = true
+[theme]
+primaryColor="#4C7DE7"
+backgroundColor="#FFFFFF"
+secondaryBackgroundColor="#fbfcfc"
+textColor="#404040"
+[client]
+showSidebarNavigation = false

src/autogluon_assistant/ui/app.py

@@ -0,0 +1,82 @@
+from copy import deepcopy
+
+import streamlit as st
+import streamlit.components.v1 as components
+from constants import DEFAULT_SESSION_VALUES, LOGO_PATH
+from pages.demo import main as demo
+from pages.feature import main as feature
+from pages.nav_bar import nav_bar
+from pages.preview import main as preview
+from pages.task import main as run
+from pages.tutorial import main as tutorial
+
+st.set_page_config(
+    page_title="AutoGluon Assistant",
+    page_icon=LOGO_PATH,
+    layout="wide",
+    initial_sidebar_state="collapsed",
+)
+
+# fontawesome
+st.markdown(
+    """
+    <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.4.0/css/all.min.css">
+    """,
+    unsafe_allow_html=True,
+)
+
+# Bootstrap 4.1.3
+st.markdown(
+    """
+    <link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/css/bootstrap.min.css" integrity="sha384-MCw98/SFnGE8fJT3GXwEOngsV7Zt27NXFoaoApmYm81iuXoPkFOJwJ8ERdknLPMO" crossorigin="anonymous">
+    """,
+    unsafe_allow_html=True,
+)
+with open("style.css") as f:
+    st.markdown(f"<style>{f.read()}</style>", unsafe_allow_html=True)
+
+
+reload_warning = """
+<script>
+  window.onbeforeunload = function () {
+
+    return  "Are you sure want to LOGOUT the session ?";
+}; 
+</script>
+"""
+
+components.html(reload_warning, height=0)
+
+
+def initial_session_state():
+    """
+    Initial Session State
+    """
+    for key, default_value in DEFAULT_SESSION_VALUES.items():
+        if key not in st.session_state:
+            st.session_state[key] = (
+                deepcopy(default_value) if isinstance(default_value, (dict, list)) else default_value
+            )
+
+
+def main():
+    initial_session_state()
+    nav_bar()
+    tutorial()
+    demo()
+    feature()
+    run()
+    preview()
+
+    st.markdown(
+        """
+    <script src="https://code.jquery.com/jquery-3.3.1.slim.min.js" integrity="sha384-q8i/X+965DzO0rT7abK41JStQIAqVgRVzpbzo5smXKp4YfRvH+8abtTE1Pi6jizo" crossorigin="anonymous"></script>
+    <script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/umd/popper.min.js" integrity="sha384-ZMP7rVo3mIykV+2+9J3UJ46jBk0WLaUAdn689aCwoqbBJiSnjAK/l8WvCWPIPm49" crossorigin="anonymous"></script>
+    <script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/js/bootstrap.min.js" integrity="sha384-ChfqqxuZUCnJSK3+MXmPNIyE6ZbWh2IMqE241rYiqJxyMiZ6OW/JmZQ5stwEULTy" crossorigin="anonymous"></script>
+    """,
+        unsafe_allow_html=True,
+    )
+
+
+if __name__ == "__main__":
+    main()

src/autogluon_assistant/ui/constants.py

@@ -0,0 +1,134 @@
+from copy import deepcopy
+
+BASE_DATA_DIR = "./user_data"
+
+
+# Preset configurations
+PRESET_DEFAULT_CONFIG = {
+    "Best Quality": {"time_limit": "4 hrs", "feature_generation": True},
+    "High Quality": {"time_limit": "1 hr", "feature_generation": False},
+    "Medium Quality": {"time_limit": "10 mins", "feature_generation": False},
+}
+DEFAULT_PRESET = "Medium Quality"
+
+PRESET_MAPPING = {
+    "Best Quality": "best_quality",
+    "High Quality": "high_quality",
+    "Medium Quality": "medium_quality",
+}
+PRESET_OPTIONS = ["Best Quality", "High Quality", "Medium Quality"]
+
+# Time limit configurations (in seconds)
+TIME_LIMIT_MAPPING = {
+    "1 min": 60,
+    "10 mins": 600,
+    "30 mins": 1800,
+    "1 hr": 3600,
+    "2 hrs": 7200,
+    "4 hrs": 14400,
+}
+
+DEFAULT_TIME_LIMIT = "10 mins"
+
+TIME_LIMIT_OPTIONS = ["1 min", "10 mins", "30 mins", "1 hr", "2 hrs", "4 hrs"]
+
+# LLM configurations
+LLM_MAPPING = {
+    "Claude 3.5 with Amazon Bedrock": "anthropic.claude-3-5-sonnet-20241022-v2:0",
+    "GPT 4o": "gpt-4o-mini-2024-07-18",
+}
+
+LLM_OPTIONS = ["Claude 3.5 with Amazon Bedrock"]
+
+# Provider configuration
+PROVIDER_MAPPING = {"Claude 3.5 with Amazon Bedrock": "bedrock", "GPT 4o": "openai"}
+
+
+API_KEY_LOCATION = {"Claude 3.5 with Amazon Bedrock": "BEDROCK_API_KEY", "GPT 4o": "OPENAI_API_KEY"}
+
+INITIAL_STAGE = {
+    "Task Understanding": [],
+    "Feature Generation": [],
+    "Model Training": [],
+    "Prediction": [],
+}
+# Initial Session state
+DEFAULT_SESSION_VALUES = {
+    "config_overrides": [],
+    "preset": DEFAULT_PRESET,
+    "time_limit": DEFAULT_TIME_LIMIT,
+    "llm": None,
+    "pid": None,
+    "logs": "",
+    "process": None,
+    "clicked": False,
+    "task_running": False,
+    "output_file": None,
+    "output_filename": None,
+    "task_description": "",
+    "sample_description": "",
+    "return_code": None,
+    "task_canceled": False,
+    "uploaded_files": {},
+    "sample_files": {},
+    "selected_dataset": None,
+    "sample_dataset_dir": None,
+    "description_uploader_key": 0,
+    "sample_dataset_selector": None,
+    "current_stage": None,
+    "feature_generation": False,
+    "stage_status": {},
+    "show_remaining_time": False,
+    "model_path": None,
+    "increment_time": 0,
+    "progress_bar": None,
+    "increment": 2,
+    "zip_path": None,
+    "stage_container": deepcopy(INITIAL_STAGE),
+}
+
+# Message to display different logging stage
+STATUS_BAR_STAGE = {
+    "Task loaded!": 10,
+    "Model training starts": 25,
+    "Fitting model": 50,
+    "AutoGluon training complete": 80,
+    "Prediction starts": 90,
+}
+
+STAGE_COMPLETE_SIGNAL = [
+    "Task understanding complete",
+    "Automatic feature generation complete",
+    "Model training complete",
+    "Prediction complete",
+]
+
+# Stage Names
+STAGE_TASK_UNDERSTANDING = "Task Understanding"
+STAGE_FEATURE_GENERATION = "Feature Generation"
+STAGE_MODEL_TRAINING = "Model Training"
+STAGE_PREDICTION = "Prediction"
+
+# Log Messages
+MSG_TASK_UNDERSTANDING = "Task understanding starts"
+MSG_FEATURE_GENERATION = "Automatic feature generation starts"
+MSG_MODEL_TRAINING = "Model training starts"
+MSG_PREDICTION = "Prediction starts"
+
+# Mapping
+STAGE_MESSAGES = {
+    MSG_TASK_UNDERSTANDING: STAGE_TASK_UNDERSTANDING,
+    MSG_FEATURE_GENERATION: STAGE_FEATURE_GENERATION,
+    MSG_MODEL_TRAINING: STAGE_MODEL_TRAINING,
+    MSG_PREDICTION: STAGE_PREDICTION,
+}
+# DataSet Options
+DATASET_OPTIONS = ["Sample Dataset", "Upload Dataset"]
+
+# Captions under DataSet Options
+CAPTIONS = ["Run with sample dataset", "Upload Train, Test and Output (Optional) Dataset"]
+
+DEMO_URL = "https://automl-mm-bench.s3.amazonaws.com/autogluon-assistant/aga-kaggle-demo.mp4"
+
+SAMPLE_DATASET_DESCRIPTION = """You are solving this data science tasks:The dataset presented here (knot theory) comprises a lot of numerical features. Some of the features may be missing, with nan value. Your task is to predict the 'signature', which has 18 unique integers. The evaluation metric is the classification accuracy."""
+LOGO_PATH = "static/page_icon.png"

src/autogluon_assistant/ui/file_uploader.py

@@ -0,0 +1,57 @@
+import os
+
+import pandas as pd
+import streamlit as st
+from utils import get_user_data_dir
+
+
+def save_description_file(description):
+    """
+    Save the task description to a file in the user's data directory.
+
+    Args:
+        description (str): The task description to save.
+    """
+    try:
+        user_data_dir = get_user_data_dir()
+        description_file = os.path.join(user_data_dir, "description.txt")
+        with open(description_file, "w") as f:
+            f.write(description)
+    except Exception as e:
+        print(f"Error saving file: {str(e)}")
+
+
+def description_file_uploader():
+    """
+    Handle Description file uploads
+    """
+    uploaded_file = st.file_uploader(
+        "Upload task description file",
+        type="txt",
+        key=st.session_state.description_uploader_key,
+        help="Accepted file format: .txt",
+        label_visibility="collapsed",
+    )
+    if uploaded_file:
+        task_description = uploaded_file.read().decode("utf-8")
+        st.session_state.task_description = task_description
+        save_description_file(st.session_state.task_description)
+        st.session_state.description_uploader_key += 1
+        st.rerun()
+
+
+def file_uploader():
+    """
+    Handle file uploads
+    """
+    st.markdown("#### Upload Dataset")
+    uploaded_files = st.file_uploader(
+        "Select the dataset", accept_multiple_files=True, label_visibility="collapsed", type=["csv", "xlsx"]
+    )
+    st.session_state.uploaded_files = {}
+    for file in uploaded_files:
+        if file.name.endswith(".csv"):
+            df = pd.read_csv(file)
+        elif file.name.endswith(".xlsx"):
+            df = pd.read_excel(file)
+        st.session_state.uploaded_files[file.name] = {"file": file, "df": df}