import numpy as np import pandas as pd import matplotlib matplotlib.use("Agg") import matplotlib.pyplot as plt import gradio as gr import spaces @spaces.GPU def _gpu_placeholder(): return "ok" MAX_ROWS = 200_000 def profile(file): if file is None: raise gr.Error("Please upload a CSV file first.") path = getattr(file, "name", file) try: df = pd.read_csv(path, nrows=MAX_ROWS) except Exception as e: raise gr.Error(f"Could not read the file as a CSV: {e}") n_rows, n_cols = df.shape n_dupes = int(df.duplicated().sum()) total_missing = int(df.isna().sum().sum()) # ---- Summary ---- summary = ( f"### Overview\n" f"- **Rows:** {n_rows:,} \n" f"- **Columns:** {n_cols} \n" f"- **Duplicate rows:** {n_dupes:,} \n" f"- **Missing cells:** {total_missing:,} \n" ) if n_rows == MAX_ROWS: summary += f"\n*Only the first {MAX_ROWS:,} rows were analysed.*" # ---- Column table ---- cols = pd.DataFrame({ "column": df.columns, "dtype": df.dtypes.astype(str).values, "missing": df.isna().sum().values, "missing_%": (df.isna().mean() * 100).round(1).values, "unique": df.nunique().values, }) # ---- Outliers (IQR rule) on numeric columns ---- out_rows = [] for c in df.select_dtypes(include=[np.number]).columns: s = df[c].dropna() if s.empty: continue q1, q3 = s.quantile(0.25), s.quantile(0.75) iqr = q3 - q1 low, high = q1 - 1.5 * iqr, q3 + 1.5 * iqr n_out = int(((s < low) | (s > high)).sum()) out_rows.append({ "column": c, "outliers": n_out, "outlier_%": round(n_out / len(s) * 100, 1), "lower_bound": round(low, 3), "upper_bound": round(high, 3), }) outliers = pd.DataFrame( out_rows, columns=["column", "outliers", "outlier_%", "lower_bound", "upper_bound"] ) # ---- Likely issues ---- issues = [] for c in df.columns[df.isna().mean() > 0.5]: issues.append(f"- **{c}** is more than 50% empty.") for c in df.columns[df.nunique() <= 1]: issues.append(f"- **{c}** has a single value (or is empty) and adds no information.") for c in df.select_dtypes(include="object").columns: converted = pd.to_numeric(df[c], errors="coerce") share = converted.notna().sum() / max(df[c].notna().sum(), 1) if share > 0.8: issues.append(f"- **{c}** looks numeric but is stored as text. Check for stray symbols or typos.") if n_dupes: issues.append(f"- {n_dupes:,} fully duplicated rows found.") issues_md = "### Likely issues\n" + ("\n".join(issues) if issues else "No obvious issues found.") # ---- Chart: missing values ---- fig, ax = plt.subplots(figsize=(7, 4)) miss = cols.sort_values("missing_%", ascending=False).head(15) ax.barh(miss["column"][::-1], miss["missing_%"][::-1]) ax.set_xlabel("% missing") ax.set_title("Missing values by column (top 15)") fig.tight_layout() return summary, issues_md, cols, outliers, fig with gr.Blocks(title="CSV Data Quality Profiler") as demo: gr.Markdown( "# CSV Data Quality Profiler\n" "Upload a CSV to check missing values, duplicates, outliers and " "columns with the wrong data type. Nothing is stored, and no AI is used. " "All checks are deterministic rules." ) file_in = gr.File(label="Upload CSV", file_types=[".csv"]) run_btn = gr.Button("Run checks", variant="primary") summary_out = gr.Markdown() issues_out = gr.Markdown() cols_out = gr.Dataframe(label="Column profile") outliers_out = gr.Dataframe(label="Outliers (IQR rule)") plot_out = gr.Plot(label="Missing values") run_btn.click( profile, file_in, [summary_out, issues_out, cols_out, outliers_out, plot_out], ) demo.launch(ssr_mode=False)