TrainerML

import gradio as gr import pandas as pd import io import pickle import matplotlib.pyplot as plt import seaborn as sns from lazypredict.Supervised import LazyClassifier, LazyRegressor from sklearn.model_selection import train_test_split from ydata_profiling import ProfileReport import tempfile import requests def load_data(file_input): """Loads CSV data from either a local file upload or a public URL.""" if file_input is None: return None try: # For local file uploads, file_input is a temporary file object if hasattr(file_input, 'name'): file_path = file_input.name with open(file_path, 'rb') as f: file_bytes = f.read() df = pd.read_csv(io.BytesIO(file_bytes)) # For URL text input elif isinstance(file_input, str) and file_input.startswith('http'): response = requests.get(file_input) response.raise_for_status() df = pd.read_csv(io.StringIO(response.text)) else: return None return df except Exception as e: gr.Warning(f"Failed to load or parse data: {e}") return None def analyze_and_model(df, target_column): """Internal function to perform EDA, model training, and visualization.""" # ... (This function's content is unchanged) profile = ProfileReport(df, title="EDA Report", minimal=True) with tempfile.NamedTemporaryFile(delete=False, suffix=".html") as temp_html: profile.to_file(temp_html.name) profile_path = temp_html.name X = df.drop(columns=[target_column]) y = df[target_column] task = "classification" if y.nunique() <= 10 else "regression" X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42) model = LazyClassifier(ignore_warnings=True, verbose=0) if task == "classification" else LazyRegressor(ignore_warnings=True, verbose=0) models, _ = model.fit(X_train, X_test, y_train, y_test) sort_metric = "Accuracy" if task == "classification" else "R-Squared" best_model_name = models.sort_values(by=sort_metric, ascending=False).index[0] best_model = model.models[best_model_name] with tempfile.NamedTemporaryFile(delete=False, suffix=".pkl") as temp_pkl: pickle.dump(best_model, temp_pkl) pickle_path = temp_pkl.name plt.figure(figsize=(10, 6)) plot_column = "Accuracy" if task == "classification" else "R-Squared" sns.barplot(x=models[plot_column].head(10), y=models.head(10).index) plt.title(f"Top 10 Models by {plot_column}") plt.tight_layout() with tempfile.NamedTemporaryFile(delete=False, suffix=".png") as temp_png: plt.savefig(temp_png.name) plot_path = temp_png.name plt.close() models_reset = models.reset_index().rename(columns={'index': 'Model'}) return profile_path, task, models_reset, plot_path, pickle_path def run_pipeline(data_source, target_column): """ This single function drives the entire application. It's exposed as the primary tool for the MCP server. :param data_source: A local file path (from gr.File) or a URL (from gr.Textbox). :param target_column: The name of the target column for prediction. """ # --- 1. Input Validation --- if not data_source or not target_column: error_msg = "Error: Data source and target column must be provided." gr.Warning(error_msg) return None, error_msg, None, None, None, "Please provide all inputs." gr.Info("Starting analysis...") # --- 2. Data Loading --- df = load_data(data_source) if df is None: return None, "Error: Could not load data.", None, None, None, None if target_column not in df.columns: error_msg = f"Error: Target column '{target_column}' not found in the dataset. Available: {list(df.columns)}" gr.Warning(error_msg) return None, error_msg, None, None, None, None # --- 3. Analysis and Modeling --- profile_path, task, models_df, plot_path, pickle_path = analyze_and_model(df, target_column) # --- 4. Explanation --- best_model_name = models_df.iloc[0]['Model'] llm_explanation = f"AI explanation for the '{task}' task: The top performing model was **{best_model_name}**." gr.Info("Analysis complete!") return profile_path, task, models_df, plot_path, pickle_path, llm_explanation # --- Gradio UI --- with gr.Blocks(title="AutoML Trainer", theme=gr.themes.Soft()) as demo: gr.Markdown("## 🤖 AutoML Trainer") gr.Markdown("Enter a CSV data source (local file or public URL) and a target column to run the analysis. This interface is now friendly for both humans and AI agents.") with gr.Row(): with gr.Column(scale=1): # Using gr.File allows for both upload and is compatible with agents file_input = gr.File(label="Upload Local CSV File") url_input = gr.Textbox(label="Or Enter Public CSV URL", placeholder="e.g., https://.../data.csv") target_column_input = gr.Textbox(label="Enter Target Column Name", placeholder="e.g., approved") run_button = gr.Button("Run Analysis & AutoML", variant="primary") with gr.Column(scale=2): task_output = gr.Textbox(label="Detected Task", interactive=False) llm_output = gr.Textbox(label="AI Explanation (WIP)", lines=3, interactive=False) metrics_output = gr.Dataframe(label="Model Performance Metrics") with gr.Row(): vis_output = gr.Image(label="Top Models Comparison") with gr.Column(): eda_output = gr.File(label="Download Full EDA Report") model_output = gr.File(label="Download Best Model (.pkl)") # The single click event that powers the whole app # A helper function decides whether to use the file or URL input def process_inputs(file_data, url_data, target): data_source = file_data if file_data is not None else url_data return run_pipeline(data_source, target) run_button.click( fn=process_inputs, inputs=[file_input, url_input, target_column_input], outputs=[eda_output, task_output, metrics_output, vis_output, model_output, llm_output] ) demo.launch( server_name="0.0.0.0", server_port=7860, share=True, show_api=True, inbrowser=True, mcp_server=True )