Update app.py

app.py

@@ -1,4 +1,5 @@
 # app.py — ST_Sonic_Ts (Shear Slowness Ts)
+
 import io, json, os, base64, math
 from pathlib import Path
 import streamlit as st
@@ -22,9 +23,16 @@ from sklearn.metrics import mean_squared_error, mean_absolute_error
 APP_NAME = "ST_Log_Sonic (Ts)"
 TAGLINE  = "Real-Time Shear Slowness (Ts) Prediction"
 
-# Defaults (will be overridden by meta if present)
-FEATURES  = ["WOB(klbf)", "TORQUE(kft.lbf)", "SPP(psi)", "RPM(1/min)", "ROP(ft/h)", "Flow Rate, gpm"]
-TARGET    = "Ts"
+# Defaults (overridden by ts_meta.json if present)
+FEATURES  = [
+    "WOB (klbf)",
+    "Torque (kft.lbf)",
+    "SPP (psi)",
+    "RPM (1/min)",
+    "ROP (ft/h)",
+    "Flow Rate (gpm)",
+]
+TARGET    = "Ts (us/ft_Actual)"
 PRED_COL  = "Ts_Pred"
 
 MODELS_DIR = Path("models")
@@ -32,7 +40,7 @@ DEFAULT_MODEL = MODELS_DIR / "ts_model.joblib"
 MODEL_FALLBACKS = [MODELS_DIR / "model.joblib", MODELS_DIR / "model.pkl"]
 COLORS = {"pred": "#1f77b4", "actual": "#f2b702", "ref": "#5a5a5a"}
 
-# Toggle to show strict version banner from meta
+# Optional env banner from meta
 STRICT_VERSION_CHECK = False
 
 # ---- Plot sizing ----
@@ -50,14 +58,11 @@ st.set_page_config(page_title=APP_NAME, page_icon="logo.png", layout="wide")
 st.markdown("""
 <style>
  .brand-logo { width: 200px; height: auto; object-fit: contain; }
- .sidebar-header { display:flex; align-items:center; gap:12px; }
- .sidebar-header .text h1 { font-size: 1.05rem; margin:0; line-height:1.1; }
- .sidebar-header .text .tag { font-size: .85rem; color:#6b7280; margin:2px 0 0; }
  .centered-container { display: flex; flex-direction: column; align-items: center; text-align: center; }
  .st-message-box { background-color: #f0f2f6; color: #333; padding: 10px; border-radius: 10px; border: 1px solid #e6e9ef; }
  .st-message-box.st-success { background-color: #d4edda; color: #155724; border-color: #c3e6cb; }
  .st-message-box.st-warning { background-color: #fff3cd; color: #856404; border-color: #ffeeba; }
- .st-message-box.st-error { background-color: #f8d7da; color: #721c24; border-color: #f5c6cb; }
+ .st-message-box.st-error   { background-color: #f8d7da; color: #721c24; border-color: #f5c6cb; }
  .main .block-container { overflow: unset !important; }
  div[data-testid="stVerticalBlock"] { overflow: unset !important; }
  div[data-testid="stExpander"] > details > summary {
@@ -140,47 +145,49 @@ def parse_excel(data_bytes: bytes):
     xl = pd.ExcelFile(bio)
     return {sh: xl.parse(sh) for sh in xl.sheet_names}
 
-def read_book_bytes(b: bytes): 
+def read_book_bytes(b: bytes):
     return parse_excel(b) if b else {}
 
 # ---- Canonical feature aliasing ------------------------------------------
 def _build_alias_map(canonical_features: list[str], target_name: str) -> dict:
     """
-    Returns a dict of common header variants -> canonical names as per the model's FEATURES.
-    We choose the canonical for each family by checking which string exists in canonical_features.
+    Map common header variants -> the *canonical* names in canonical_features.
+    Whatever appears in canonical_features (from ts_meta.json) wins.
     """
-    def pick(expected_list, family_variants):
-        # pick the first variant that exists in expected_list, else fall back to first in family_variants
-        for v in family_variants:
+    def pick(expected_list, variants):
+        for v in variants:
             if v in expected_list:
                 return v
-        return family_variants[0]
+        return variants[0]
 
-    can_WOB    = pick(canonical_features, ["WOB, klbf","WOB(klbf)","WOB (klbf)"])
-    can_TORQUE = pick(canonical_features, ["Torque(kft.lbf)","TORQUE(kft.lbf)"])
-    can_SPP    = pick(canonical_features, ["SPP(psi)"])
-    can_RPM    = pick(canonical_features, ["RPM(1/min)","RPM (1/min)"])
-    can_ROP    = pick(canonical_features, ["ROP(ft/h)","ROP (ft/h)"])
-    can_FR     = pick(canonical_features, ["Flow Rate, gpm","Flow Rate , gpm","Flow Rate,gpm"])
+    can_WOB    = pick(canonical_features, ["WOB (klbf)", "WOB, klbf", "WOB(klbf)", "WOB( klbf)"])
+    can_TORQUE = pick(canonical_features, ["Torque (kft.lbf)", "Torque(kft.lbf)", "TORQUE(kft.lbf)"])
+    can_SPP    = pick(canonical_features, ["SPP (psi)", "SPP(psi)"])
+    can_RPM    = pick(canonical_features, ["RPM (1/min)", "RPM(1/min)"])
+    can_ROP    = pick(canonical_features, ["ROP (ft/h)", "ROP(ft/h)"])
+    can_FR     = pick(canonical_features, ["Flow Rate (gpm)", "Flow Rate, gpm", "Flow Rate,gpm", "Flow Rate , gpm", "Fow Rate, gpm", "Fow Rate, gpm "])
+    can_DEPTH  = "Depth (ft)"
 
     alias = {
-        # WOB
-        "WOB, klbf": can_WOB, "WOB(klbf)": can_WOB, "WOB (klbf)": can_WOB, "WOB( klbf)": can_WOB, "WOB , klbf": can_WOB,
-        # Torque
-        "Torque(kft.lbf)": can_TORQUE, "TORQUE(kft.lbf)": can_TORQUE,
-        # SPP
-        "SPP(psi)": can_SPP,
-        # RPM
-        "RPM(1/min)": can_RPM, "RPM (1/min)": can_RPM,
-        # ROP
-        "ROP(ft/h)": can_ROP, "ROP (ft/h)": can_ROP,
-        # Flow
-        "Flow Rate, gpm": can_FR, "Fow Rate, gpm": can_FR, "Fow Rate, gpm ": can_FR, "Flow Rate , gpm": can_FR, "Flow Rate,gpm": can_FR,
-        # Depth (for plotting only)
-        "Depth, ft": "Depth, ft", "Depth(ft)": "Depth, ft", "DEPTH, ft": "Depth, ft",
-        # Ts targets (map all to the chosen TARGET)
-        "Ts": target_name, "Ts,us/ft_Actual": target_name, "Ts, us/ft_Actual": target_name,
-        "TS_Actual": target_name, "Ts (us/ft)_Actual": target_name
+        # Features
+        "WOB (klbf)": can_WOB, "WOB, klbf": can_WOB, "WOB(klbf)": can_WOB, "WOB( klbf)": can_WOB,
+        "Torque (kft.lbf)": can_TORQUE, "Torque(kft.lbf)": can_TORQUE, "TORQUE(kft.lbf)": can_TORQUE,
+        "SPP (psi)": can_SPP, "SPP(psi)": can_SPP,
+        "RPM (1/min)": can_RPM, "RPM(1/min)": can_RPM,
+        "ROP (ft/h)": can_ROP, "ROP(ft/h)": can_ROP,
+        "Flow Rate (gpm)": can_FR, "Flow Rate, gpm": can_FR, "Flow Rate,gpm": can_FR, "Flow Rate , gpm": can_FR,
+        "Fow Rate, gpm": can_FR, "Fow Rate, gpm ": can_FR,
+
+        # Depth (plot only)
+        "Depth (ft)": can_DEPTH, "Depth, ft": can_DEPTH, "Depth(ft)": can_DEPTH, "DEPTH, ft": can_DEPTH,
+
+        # Target family
+        "Ts (us/ft_Actual)": target_name,
+        "Ts,us/ft_Actual":   target_name,
+        "Ts, us/ft_Actual":  target_name,
+        "Ts":                target_name,
+        "TS_Actual":         target_name,
+        "Ts (us/ft)_Actual": target_name,
     }
     return alias
 
@@ -188,7 +195,6 @@ def _normalize_columns(df: pd.DataFrame, canonical_features: list[str], target_n
     out = df.copy()
     out.columns = [str(c).strip().replace(" ,", ",").replace(",  ", ", ").replace("  ", " ") for c in out.columns]
     alias = _build_alias_map(canonical_features, target_name)
-    # only rename keys that actually exist
     actual = {k: v for k, v in alias.items() if k in out.columns and k != v}
     return out.rename(columns=actual)
 
@@ -219,6 +225,17 @@ def df_centered_rounded(df: pd.DataFrame, hide_index=True):
     )
     st.dataframe(styler, use_container_width=True, hide_index=hide_index)
 
+# ---------- Build X exactly as trained ----------
+def _make_X(df: pd.DataFrame, features: list[str]) -> pd.DataFrame:
+    """
+    Reindex columns to the exact training feature order and coerce to numeric.
+    Prevents scikit-learn 'feature names should match' errors.
+    """
+    X = df.reindex(columns=features, copy=False)
+    for c in X.columns:
+        X[c] = pd.to_numeric(X[c], errors="coerce")
+    return X
+
 # === Excel export helpers =================================================
 def _excel_engine() -> str:
     try:
@@ -497,27 +514,54 @@ def track_plot(df, include_actual=True):
 
 # ---------- Preview (matplotlib) ----------
 def preview_tracks(df: pd.DataFrame, cols: list[str]):
+    """
+    Quick-look multi-track preview:
+      - one subplot per selected column
+      - distinct stable colors per column
+      - shared & reversed Y-axis (Depth downwards)
+    """
     cols = [c for c in cols if c in df.columns]
     n = len(cols)
     if n == 0:
         fig, ax = plt.subplots(figsize=(4, 2))
-        ax.text(0.5,0.5,"No selected columns",ha="center",va="center"); ax.axis("off")
+        ax.text(0.5, 0.5, "No selected columns", ha="center", va="center")
+        ax.axis("off")
         return fig
-    fig, axes = plt.subplots(1, n, figsize=(2.2*n, 7.0), sharey=True, dpi=100)
-    if n == 1: axes = [axes]
+
+    # Depth or fallback to index
     depth_col = next((c for c in df.columns if 'depth' in str(c).lower()), None)
     if depth_col is not None:
-        idx = pd.to_numeric(df[depth_col], errors="coerce"); y_label = depth_col
+        idx = pd.to_numeric(df[depth_col], errors="coerce")
+        y_label = depth_col
     else:
-        idx = pd.Series(np.arange(1, len(df) + 1)); y_label = "Point Index"
+        idx = pd.Series(np.arange(1, len(df) + 1))
+        y_label = "Point Index"
+
+    y_min, y_max = float(idx.min()), float(idx.max())
+
+    # Stable qualitative palette
+    cmap = plt.get_cmap("tab20")
+    col_colors = {col: cmap(i % cmap.N) for i, col in enumerate(cols)}
+
+    fig, axes = plt.subplots(1, n, figsize=(2.3 * n, 7.0), sharey=True, dpi=100)
+    if n == 1:
+        axes = [axes]
+
     for i, (ax, col) in enumerate(zip(axes, cols)):
-        ax.plot(pd.to_numeric(df[col], errors="coerce"), idx, '-', lw=1.6, color="#333")
-        ax.set_xlabel(col); ax.xaxis.set_label_position('top'); ax.xaxis.tick_top()
-        ax.set_ylim(float(idx.max()), float(idx.min()))
+        x = pd.to_numeric(df[col], errors="coerce")
+        ax.plot(x, idx, '-', lw=1.8, color=col_colors[col])
+        ax.set_xlabel(col)
+        ax.xaxis.set_label_position('top')
+        ax.xaxis.tick_top()
+        ax.set_ylim(y_max, y_min)  # reversed Y (Depth down)
         ax.grid(True, linestyle=":", alpha=0.3)
-        if i == 0: ax.set_ylabel(y_label)
+
+        if i == 0:
+            ax.set_ylabel(y_label)
         else:
-            ax.tick_params(labelleft=False); ax.set_ylabel("")
+            ax.tick_params(labelleft=False)
+            ax.set_ylabel("")
+
     fig.tight_layout()
     return fig
 
@@ -551,7 +595,7 @@ except Exception as e:
     st.error(f"Failed to load model: {e}")
     st.stop()
 
-# Prefer Ts meta
+# Load meta (prefer Ts-specific)
 meta = {}
 meta_candidates = [MODELS_DIR / "ts_meta.json", MODELS_DIR / "meta.json", MODELS_DIR / "ym_meta.json"]
 meta_path = next((p for p in meta_candidates if p.exists()), None)
@@ -564,7 +608,7 @@ if meta_path:
     except Exception as e:
         st.warning(f"Could not parse meta file ({meta_path.name}): {e}")
 
-# Optional: version banner (silenced by default)
+# Optional: version banner
 if STRICT_VERSION_CHECK and meta.get("versions"):
     import numpy as _np, sklearn as _skl
     mv = meta["versions"]; msg=[]
@@ -671,7 +715,6 @@ if st.session_state.app_step == "dev":
             st.markdown('<div class="st-message-box st-error">Workbook must include Train/Training/training2 and Test/Testing/testing2 sheets.</div>', unsafe_allow_html=True)
             st.stop()
 
-        # Use meta FEATURES as canonical when normalizing
         tr = _normalize_columns(book[sh_train].copy(), FEATURES, TARGET)
         te = _normalize_columns(book[sh_test].copy(),  FEATURES, TARGET)
 
@@ -679,9 +722,9 @@ if st.session_state.app_step == "dev":
             st.markdown('<div class="st-message-box st-error">Missing required columns.</div>', unsafe_allow_html=True)
             st.stop()
 
-        # Predict with exactly the columns the model was trained on
-        tr[PRED_COL] = model.predict(tr[FEATURES])
-        te[PRED_COL] = model.predict(te[FEATURES])
+        # Predict with exactly the training feature order
+        tr[PRED_COL] = model.predict(_make_X(tr, FEATURES))
+        te[PRED_COL] = model.predict(_make_X(te, FEATURES))
 
         st.session_state.results["Train"]=tr; st.session_state.results["Test"]=te
         st.session_state.results["m_train"]={
@@ -746,9 +789,9 @@ if st.session_state.app_step == "validate":
         book = read_book_bytes(up.getvalue())
         name = find_sheet(book, ["Validation","Validate","validation2","Val","val"]) or list(book.keys())[0]
         df = _normalize_columns(book[name].copy(), FEATURES, TARGET)
-        if not ensure_cols(df, FEATURES+[TARGET]): 
+        if not ensure_cols(df, FEATURES+[TARGET]):
             st.markdown('<div class="st-message-box st-error">Missing required columns.</div>', unsafe_allow_html=True); st.stop()
-        df[PRED_COL] = model.predict(df[FEATURES])
+        df[PRED_COL] = model.predict(_make_X(df, FEATURES))
         st.session_state.results["Validate"]=df
 
         ranges = st.session_state.train_ranges; oor_pct = 0.0; tbl=None
@@ -759,7 +802,9 @@ if st.session_state.app_step == "validate":
                 tbl = df.loc[any_viol, FEATURES].copy()
                 for c in FEATURES:
                     if pd.api.types.is_numeric_dtype(tbl[c]): tbl[c] = tbl[c].round(2)
-                tbl["Violations"] = pd.DataFrame({f:(df[f]<ranges[f][0])|(df[f]>ranges[f][1]) for f in FEATURES}).loc[any_viol].apply(lambda r:", ".join([c for c,v in r.items() if v]), axis=1)
+                tbl["Violations"] = pd.DataFrame({f:(df[f]<ranges[f][0])|(df[f]>ranges[f][1]) for f in FEATURES}).loc[any_viol].apply(
+                    lambda r:", ".join([c for c,v in r.items() if v]), axis=1
+                )
         st.session_state.results["m_val"]={
             "R": pearson_r(df[TARGET], df[PRED_COL]),
             "RMSE": rmse(df[TARGET], df[PRED_COL]),
@@ -818,9 +863,9 @@ if st.session_state.app_step == "predict":
     if go_btn and up is not None:
         book = read_book_bytes(up.getvalue()); name = list(book.keys())[0]
         df = _normalize_columns(book[name].copy(), FEATURES, TARGET)
-        if not ensure_cols(df, FEATURES): 
+        if not ensure_cols(df, FEATURES):
             st.markdown('<div class="st-message-box st-error">Missing required columns.</div>', unsafe_allow_html=True); st.stop()
-        df[PRED_COL] = model.predict(df[FEATURES])
+        df[PRED_COL] = model.predict(_make_X(df, FEATURES))
         st.session_state.results["PredictOnly"]=df
 
         ranges = st.session_state.train_ranges; oor_pct = 0.0