Genesis AI deploy

DEPLOY.md

@@ -0,0 +1,105 @@
+# Deploy to Hugging Face Spaces – Step by Step
+
+## Prerequisites
+- Hugging Face account (free) at https://huggingface.co
+- Git installed on your machine
+
+---
+
+## Step 1 – Create a new Space
+
+1. Go to https://huggingface.co/spaces
+2. Click **Create new Space**
+3. Fill in:
+   - **Space name:** `genesisai-forecasting`
+   - **License:** MIT
+   - **SDK:** Docker   ← important, select Docker not Gradio/Streamlit
+   - **Visibility:** Public (or Private if you prefer)
+4. Click **Create Space**
+
+Your URL will be:
+**https://huggingface.co/spaces/YOUR_USERNAME/genesisai-forecasting**
+
+---
+
+## Step 2 – Clone and push your code
+
+Open terminal / Git Bash:
+
+```bash
+# Clone the empty Space (replace YOUR_USERNAME)
+git clone https://huggingface.co/spaces/YOUR_USERNAME/genesisai-forecasting
+cd genesisai-forecasting
+
+# Copy all your files into this folder:
+#   app.py
+#   dabur_baseline_forecast_v2.py
+#   dabur_demand_platform.html
+#   requirements.txt
+#   Dockerfile
+#   README.md
+#   .gitignore
+
+# Commit and push
+git add .
+git commit -m "Genesis AI initial deploy"
+git push
+```
+
+HF will automatically detect the Dockerfile and start building.
+Build takes ~3-5 minutes (installing statsforecast).
+
+---
+
+## Step 3 – Set Secrets (passwords)
+
+1. Go to your Space → **Settings** → **Repository secrets**
+2. Add these secrets:
+
+| Secret name | Value |
+|-------------|-------|
+| `ANALYST_PASSWORD` | your secure analyst password |
+| `VIEWER_PASSWORD` | your secure viewer password |
+| `SECRET_KEY` | any random string (e.g. `hf-dabur-2025-xk92`) |
+
+3. After adding secrets → **Factory reboot** the Space
+
+---
+
+## Step 4 – Open your Space
+
+Go to: **https://huggingface.co/spaces/YOUR_USERNAME/genesisai-forecasting**
+
+Log in with `analyst` and your analyst password.
+
+---
+
+## Sharing with viewers
+
+Send viewers this URL and the viewer password.
+They log in as `viewer` and see the latest forecast output immediately
+(no upload, no run button — read only).
+
+---
+
+## Pushing updates
+
+```bash
+cd genesisai-forecasting
+# make your changes to the files
+git add .
+git commit -m "update"
+git push
+```
+
+HF rebuilds automatically in ~3 minutes.
+
+---
+
+## Notes
+
+- **No sleep** – HF Spaces never sleep on public Spaces
+- **16GB RAM, 2 vCPU** – plenty for statsforecast on your dataset
+- **Forecast resets on restart** – results are in-memory.
+  If you redeploy, analyst needs to run once before viewer can see results.
+- **Private Space** – costs $0.05/hour on HF Pro. Public is always free.

Dockerfile

@@ -0,0 +1,33 @@
+# Genesis AI – Dabur Demand Intelligence Platform
+# HF Spaces Docker deployment
+
+FROM python:3.11-slim
+
+# HF Spaces runs as non-root user 1000
+RUN useradd -m -u 1000 appuser
+
+WORKDIR /app
+
+# Install dependencies first (layer caching)
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy application files
+COPY app.py .
+COPY dabur_baseline_forecast_v2.py .
+COPY dabur_demand_platform.html .
+
+# HF Spaces requires port 7860
+EXPOSE 7860
+
+# Switch to non-root user
+USER appuser
+
+# Run with gunicorn – threaded for background jobs, no timeout
+CMD ["gunicorn", "app:app", \
+     "--bind", "0.0.0.0:7860", \
+     "--workers", "1", \
+     "--threads", "8", \
+     "--timeout", "0", \
+     "--worker-class", "gthread", \
+     "--access-logfile", "-"]

README.md

@@ -1,11 +1,35 @@
 ---
-title: Genesisai Forecasting
-emoji: 📚
+title: Genesis AI Forecasting
+emoji: 📈
 colorFrom: red
-colorTo: yellow
+colorTo: orange
 sdk: docker
 pinned: false
-short_description: genesisai test app
+app_port: 7860
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Genesis AI – Dabur Demand Intelligence Platform
+
+Baseline volume forecasting across Channel × Design Brand × Account using ensemble time-series models (AutoETS, AutoARIMA, AutoTheta, HoltWinters, MSTL).
+
+## Users
+
+| Username | Password | Access |
+|----------|----------|--------|
+| `analyst` | set via Secret `ANALYST_PASSWORD` | Upload data, run forecast, view output |
+| `viewer` | set via Secret `VIEWER_PASSWORD` | View latest forecast output only |
+
+## How to Use
+
+1. Log in as **analyst**
+2. Go to **Forecast Engine** → upload your `.xlsx` file
+3. Click **Run Baseline Forecast** — pipeline runs in background
+4. Results appear automatically in **Forecast Output**
+5. Share the Space URL with viewers — they log in and see the output instantly
+
+## Setting Passwords (recommended)
+
+In your HF Space → **Settings** → **Repository secrets**:
+- `ANALYST_PASSWORD` → your secure password
+- `VIEWER_PASSWORD` → your secure password
+- `SECRET_KEY` → any random string for session encryption

app.py

@@ -0,0 +1,209 @@
+"""
+Genesis AI – Dabur Demand Intelligence Platform
+Hugging Face Spaces entry point.
+
+HF Spaces requires:
+  - Entry file named app.py
+  - App listening on port 7860
+  - SDK: docker (we use Flask, not Gradio/Streamlit)
+"""
+
+from flask import Flask, request, jsonify, send_from_directory, session
+import io, traceback, os, uuid, threading, time
+
+from dabur_baseline_forecast_v2 import run_pipeline, CONFIG
+
+BASE_DIR   = os.path.dirname(os.path.abspath(__file__))
+SECRET_KEY = os.environ.get("SECRET_KEY", "dabur-genesisai-hf-2025")
+PORT       = int(os.environ.get("PORT", 7860))   # HF Spaces requires 7860
+
+app = Flask(__name__, static_folder=BASE_DIR, static_url_path="")
+app.secret_key = SECRET_KEY
+
+# =============================================================================
+# USERS  – set via HF Space Secrets for security
+#   ANALYST_PASSWORD  /  VIEWER_PASSWORD
+# =============================================================================
+USERS = {
+    "analyst": {
+        "password": os.environ.get("ANALYST_PASSWORD", "dabur@analyst"),
+        "role": "analyst",
+    },
+    "viewer": {
+        "password": os.environ.get("VIEWER_PASSWORD", "dabur@viewer"),
+        "role": "viewer",
+    },
+}
+
+# =============================================================================
+# JOB STORE – background thread per forecast run
+# Frontend submits → gets job_id instantly → polls /job/<id> every 5s
+# =============================================================================
+_jobs: dict        = {}
+_latest_forecast   = None
+_jobs_lock         = threading.Lock()
+
+
+def _run_job(job_id: str, file_bytes: bytes, cfg: dict):
+    """Pipeline runs in daemon thread. Safe to run long without timeout."""
+    global _latest_forecast
+    try:
+        result = run_pipeline(io.BytesIO(file_bytes), cfg)
+        with _jobs_lock:
+            _jobs[job_id].update({
+                "status":      "done",
+                "result":      result,
+                "finished_at": time.time(),
+            })
+        _latest_forecast = result
+        print(f"[JOB {job_id[:8]}] Done – {result.get('n_series',0)} series")
+    except Exception as e:
+        tb = traceback.format_exc()
+        print(f"[JOB {job_id[:8]}] ERROR:\n{tb}")
+        with _jobs_lock:
+            _jobs[job_id].update({
+                "status":      "error",
+                "error":       str(e),
+                "traceback":   tb,
+                "finished_at": time.time(),
+            })
+    # Purge jobs older than 2 hours
+    now = time.time()
+    with _jobs_lock:
+        for jid in [k for k, v in _jobs.items()
+                    if v.get("finished_at") and now - v["finished_at"] > 7200]:
+            del _jobs[jid]
+
+
+# =============================================================================
+# ROUTES
+# =============================================================================
+
+@app.route("/")
+def index():
+    return send_from_directory(BASE_DIR, "dabur_demand_platform.html")
+
+
+# ── Auth ───────────────────────────────────────────────────────────────
+@app.route("/login", methods=["POST"])
+def login():
+    data     = request.get_json() or {}
+    username = data.get("username", "").strip().lower()
+    password = data.get("password", "")
+    user     = USERS.get(username)
+    if not user or user["password"] != password:
+        return jsonify({"status": "error", "message": "Invalid username or password"}), 401
+    session["username"] = username
+    session["role"]     = user["role"]
+    print(f"[AUTH] Login: {username} ({user['role']})")
+    return jsonify({"status": "ok", "username": username, "role": user["role"]})
+
+
+@app.route("/logout", methods=["POST"])
+def logout():
+    print(f"[AUTH] Logout: {session.get('username','?')}")
+    session.clear()
+    return jsonify({"status": "ok"})
+
+
+@app.route("/me", methods=["GET"])
+def me():
+    if "username" in session:
+        return jsonify({"status": "ok",
+                        "username": session["username"],
+                        "role":     session["role"]})
+    return jsonify({"status": "error", "message": "Not logged in"}), 401
+
+
+# ── Health ─────────────────────────────────────────────────────────────
+@app.route("/health", methods=["GET"])
+def health():
+    return jsonify({"status": "ok", "message": "Genesis AI – running on HF Spaces"})
+
+
+# ── Submit forecast job ────────────────────────────────────────────────
+@app.route("/run", methods=["POST"])
+def run():
+    if session.get("role") != "analyst":
+        return jsonify({"status": "error",
+                        "message": "Access denied – analyst role required"}), 403
+    if "file" not in request.files:
+        return jsonify({"status": "error", "message": "No file uploaded"}), 400
+
+    f = request.files["file"]
+    if not f.filename.lower().endswith((".xlsx", ".xls")):
+        return jsonify({"status": "error",
+                        "message": "Only .xlsx files are supported"}), 400
+
+    cfg = CONFIG.copy()
+    try:
+        if request.form.get("horizon"):    cfg["forecast_horizon"] = int(request.form["horizon"])
+        if request.form.get("season"):     cfg["season_length"]    = int(request.form["season"])
+        if request.form.get("cv_horizon"): cfg["cv_horizon"]       = int(request.form["cv_horizon"])
+        if request.form.get("top_n"):      cfg["top_n_ensemble"]   = int(request.form["top_n"])
+    except ValueError as e:
+        return jsonify({"status": "error", "message": f"Invalid config value: {e}"}), 400
+
+    file_bytes = f.read()
+    job_id     = str(uuid.uuid4())
+
+    with _jobs_lock:
+        _jobs[job_id] = {
+            "status":      "running",
+            "result":      None,
+            "error":       None,
+            "started_at":  time.time(),
+            "finished_at": None,
+        }
+
+    threading.Thread(target=_run_job, args=(job_id, file_bytes, cfg),
+                     daemon=True).start()
+
+    print(f"[JOB {job_id[:8]}] Started by {session.get('username','?')}")
+    return jsonify({"status": "started", "job_id": job_id})
+
+
+# ── Poll job status ────────────────────────────────────────────────────
+@app.route("/job/<job_id>", methods=["GET"])
+def job_status(job_id):
+    if "username" not in session:
+        return jsonify({"status": "error", "message": "Not logged in"}), 401
+
+    with _jobs_lock:
+        job = _jobs.get(job_id)
+
+    if not job:
+        return jsonify({"status": "error", "message": "Job not found"}), 404
+
+    if job["status"] == "running":
+        return jsonify({"status": "running",
+                        "elapsed_seconds": round(time.time() - job["started_at"])})
+
+    if job["status"] == "error":
+        return jsonify({"status": "error", "message": job["error"]}), 500
+
+    return jsonify({"status": "done", "result": job["result"]})
+
+
+# ── Latest forecast for viewer ─────────────────────────────────────────
+@app.route("/forecast-result", methods=["GET"])
+def forecast_result():
+    if "username" not in session:
+        return jsonify({"status": "error", "message": "Not logged in"}), 401
+    if _latest_forecast is None:
+        return jsonify({"status": "empty",
+                        "message": "No forecast has been run yet. Ask the analyst to run the pipeline."})
+    return jsonify(_latest_forecast)
+
+
+# =============================================================================
+# BOOT
+# =============================================================================
+if __name__ == "__main__":
+    print("=" * 60)
+    print("  Genesis AI – Dabur Demand Intelligence Platform")
+    print(f"  Running on port {PORT}")
+    print(f"  analyst / {USERS['analyst']['password']}  →  Full access")
+    print(f"  viewer  / {USERS['viewer']['password']}   →  Output only")
+    print("=" * 60)
+    app.run(host="0.0.0.0", port=PORT, debug=False, threaded=True)

dabur_baseline_forecast_v2.py

@@ -0,0 +1,556 @@
+# =============================================================================
+# DABUR BASELINE FORECAST PIPELINE  v2 – Excel Input Compatible
+# Format  : Long format with Driver Level 3 as row identifier
+# Scope   : Multi Design Brand × Multi Account (loops automatically)
+# Target  : Forecast Baseline Volume for next 12 months
+# Input   : Excel file (.xlsx) – reads "Base Sheet"
+# Output  : JSON to stdout (consumed by the web application)
+# =============================================================================
+
+import io, json, warnings, sys
+import numpy as np
+import pandas as pd
+
+from statsforecast import StatsForecast
+from statsforecast.models import (
+    AutoETS, AutoARIMA, AutoTheta,
+    MSTL, HoltWinters, SeasonalNaive, WindowAverage,
+)
+
+warnings.filterwarnings("ignore")
+pd.set_option("display.float_format", "{:.4f}".format)
+
+
+# =============================================================================
+# SECTION 0 – CONFIGURATION
+# =============================================================================
+
+CONFIG = {
+    "channel_col"     : "Channel",
+    "category_col"    : "Category",
+    "division_col"    : "Division",
+    "brand_group_col" : "Brand Group",
+    "account_col"     : "Account",
+    "key_col"         : "Key",
+    "design_brand_col": "Design Brand",
+    "date_col"        : "DATE",
+    "fy_col"          : "FY",
+    "volume_col"      : "Volume",
+    "price_col"       : "Price",
+    "value_col"       : "Value",
+    "driver_col"      : "Driver Level 3",
+
+    "baseline_label"  : "Baseline",
+    "actual_label"    : "Actual Volume",
+    "forecast_label"  : "Baseline Forecast",
+
+    "freq"            : "MS",
+    "season_length"   : 12,
+    "forecast_horizon": 12,
+    "top_n_ensemble"  : 3,
+    "cv_horizon"      : 6,
+    "cv_windows"      : 2,
+
+    # Minimum number of baseline data points required to run a forecast.
+    # Series with fewer than this many non-null, non-zero baseline months
+    # are skipped entirely – no CV, no forecast, reported in skipped_series.
+    "min_data_points" : 10,
+
+    # Excel sheet to read
+    "sheet_name"      : "Base Sheet",
+}
+
+SERIES_KEYS = ["Channel", "Design Brand", "Account"]
+
+
+# =============================================================================
+# SECTION 1 – DATA INGESTION
+#   Accepts: file path (str), bytes (from web upload), or pd.DataFrame
+# =============================================================================
+
+def load_data(source, cfg: dict) -> pd.DataFrame:
+    if isinstance(source, pd.DataFrame):
+        return source.copy()
+    if isinstance(source, (str, bytes, io.BytesIO)):
+        df = pd.read_excel(source, sheet_name=cfg["sheet_name"])
+    else:
+        raise ValueError(f"Unsupported source type: {type(source)}")
+
+    # Ensure DATE is datetime
+    df[cfg["date_col"]] = pd.to_datetime(df[cfg["date_col"]], errors="coerce")
+    df[cfg["date_col"]] = df[cfg["date_col"]].dt.to_period("M").dt.to_timestamp()
+    return df
+
+
+# =============================================================================
+# SECTION 2 – PREPROCESSING  (long → wide, one row per series × date)
+# =============================================================================
+
+def get_fy(date: pd.Timestamp) -> str:
+    """Dabur fiscal year: April–March."""
+    yr = date.year + 1 if date.month >= 4 else date.year
+    return f"FY{str(yr)[2:]}"
+
+
+def preprocess_long_to_wide(df: pd.DataFrame, cfg: dict) -> pd.DataFrame:
+    dc    = cfg["driver_col"]
+    vc    = cfg["volume_col"]
+    pc    = cfg["price_col"]
+    datec = cfg["date_col"]
+    bl    = cfg["baseline_label"]
+
+    # Keep only Baseline and Actual Volume rows
+    df = df[df[dc].isin([bl, cfg["actual_label"]])].copy()
+
+    dim_cols = [
+        cfg["channel_col"], cfg["category_col"], cfg["division_col"],
+        cfg["brand_group_col"], cfg["account_col"], cfg["key_col"],
+        cfg["design_brand_col"],
+    ]
+
+    pivot = df.pivot_table(
+        index=dim_cols + [datec],
+        columns=dc,
+        values=vc,
+        aggfunc="first",
+    ).reset_index()
+
+    pivot.columns.name = None
+    rename = {datec: "ds"}
+    if bl in pivot.columns:
+        rename[bl] = "baseline"
+    if cfg["actual_label"] in pivot.columns:
+        rename[cfg["actual_label"]] = "actual"
+    pivot.rename(columns=rename, inplace=True)
+
+    # Carry price from Baseline rows
+    price_map = (
+        df[df[dc] == bl]
+        .set_index(dim_cols + [datec])[pc]
+        .rename("price")
+    )
+    pivot = pivot.join(price_map, on=dim_cols + ["ds"])
+
+    pivot.sort_values(dim_cols + ["ds"], inplace=True)
+    pivot.reset_index(drop=True, inplace=True)
+
+    pivot["unique_id"] = pivot[SERIES_KEYS].apply(
+        lambda r: " | ".join(r.astype(str)), axis=1
+    )
+
+    for col in ["baseline", "actual"]:
+        if col in pivot.columns:
+            neg = (pivot[col] < 0).sum()
+            if neg:
+                pivot[col] = pivot[col].clip(lower=0)
+
+    return pivot
+
+
+def check_gaps(series_df: pd.DataFrame, uid: str) -> pd.DataFrame:
+    full_range = pd.date_range(series_df["ds"].min(), series_df["ds"].max(), freq="MS")
+    missing = full_range.difference(series_df["ds"])
+    if len(missing):
+        fill = pd.DataFrame({"ds": missing})
+        series_df = (
+            pd.concat([series_df, fill], ignore_index=True)
+            .sort_values("ds")
+            .ffill()
+        )
+    return series_df.reset_index(drop=True)
+
+
+# =============================================================================
+# SECTION 2b – PRE-FLIGHT VALIDATION
+#   Classifies every series as PASS or SKIP before any modelling begins.
+#   Returns two lists: valid_ids (will be forecast) and skipped (will not).
+# =============================================================================
+
+def validate_series(wide_df: pd.DataFrame, cfg: dict) -> tuple[list, list]:
+    """
+    Check each unique series for minimum data requirements.
+
+    Rules
+    -----
+    1. Baseline column must exist and have at least cfg["min_data_points"]
+       non-null, non-zero values.
+    2. Series date range must span at least cfg["min_data_points"] months
+       (guards against duplicate-date inflation).
+
+    Returns
+    -------
+    valid_ids   : list of unique_id strings that pass – will be forecast
+    skipped     : list of dicts with uid + reason – will be excluded
+    """
+    min_pts   = cfg["min_data_points"]
+    valid_ids = []
+    skipped   = []
+
+    all_ids = wide_df["unique_id"].unique().tolist()
+
+    print(f"\n{'='*70}")
+    print(f"  PRE-FLIGHT VALIDATION  ({len(all_ids)} series found in data)")
+    print(f"  Minimum data points required : {min_pts} months of Baseline")
+    print(f"{'='*70}")
+
+    for uid in all_ids:
+        s_df = wide_df[wide_df["unique_id"] == uid]
+
+        # Count non-null, non-zero baseline points
+        if "baseline" not in s_df.columns:
+            reason = "SKIP – 'Baseline' column missing after pivot"
+            skipped.append({"uid": uid, "reason": reason, "n_points": 0})
+            print(f"  [SKIP]  {uid}")
+            print(f"          Reason  : {reason}")
+            continue
+
+        valid_pts  = s_df["baseline"].replace(0, np.nan).dropna().shape[0]
+        date_span  = s_df["ds"].nunique()
+
+        if valid_pts < min_pts:
+            reason = (
+                f"SKIP – only {valid_pts} valid baseline point(s), "
+                f"need ≥ {min_pts}. Date range spans {date_span} month(s)."
+            )
+            skipped.append({"uid": uid, "reason": reason, "n_points": valid_pts})
+            print(f"  [SKIP]  {uid}")
+            print(f"          Points  : {valid_pts} valid baseline months  "
+                  f"(threshold = {min_pts})")
+            print(f"          Action  : Series excluded from forecast")
+        else:
+            valid_ids.append(uid)
+            print(f"  [PASS]  {uid}  ({valid_pts} pts)")
+
+    n_pass = len(valid_ids)
+    n_skip = len(skipped)
+    print(f"\n  Summary : {n_pass} series will be forecast  |  "
+          f"{n_skip} series skipped")
+    if skipped:
+        print(f"\n  Skipped series list:")
+        for s in skipped:
+            print(f"    • {s['uid']}  [{s['n_points']} pts]")
+    print(f"{'='*70}\n")
+
+    return valid_ids, skipped
+
+
+# =============================================================================
+# SECTION 3 – MODEL BANK
+# =============================================================================
+
+def build_models(season_length: int) -> list:
+    return [
+        AutoETS(season_length=season_length),
+        AutoARIMA(season_length=season_length),
+        AutoTheta(season_length=season_length),
+        HoltWinters(season_length=season_length, error_type="A", alias="HWAdd"),
+        HoltWinters(season_length=season_length, error_type="M", alias="HWMult"),
+        MSTL(season_length=season_length, trend_forecaster=AutoTheta(), alias="MSTL_Theta"),
+        SeasonalNaive(season_length=season_length),
+    ]
+
+
+# =============================================================================
+# SECTION 4 – CROSS-VALIDATION
+# =============================================================================
+
+def cross_validate_series(sf_df: pd.DataFrame, cfg: dict) -> pd.DataFrame:
+    """
+    Rank models by CV MAE, auto-scaling settings for the available series length.
+
+    Statsforecast minimum-length formula:  h + (n_windows - 1) * step + 1
+    The loop shrinks n_windows first, then h, until the series fits.
+    Falls back to a fixed default ranking if even h=2, nw=1 is too large.
+    """
+    DEFAULT_RANKING = [
+        "AutoETS", "AutoARIMA", "AutoTheta",
+        "HWAdd", "HWMult", "MSTL_Theta", "SeasonalNaive",
+    ]
+    n    = len(sf_df)
+    h    = cfg["cv_horizon"]
+    nw   = cfg["cv_windows"]
+    step = 3
+    orig_h, orig_nw = h, nw
+
+    while True:
+        needed = h + (nw - 1) * step + 1
+        if n >= needed:
+            break
+        if nw > 1:
+            nw -= 1
+        elif h > 2:
+            h -= 1
+        else:
+            print(f"      [CV]  Series too short even for minimal CV "
+                  f"(n={n}) – using default model ranking.")
+            return pd.DataFrame({
+                "Model": DEFAULT_RANKING,
+                "MAE":   [float("nan")] * len(DEFAULT_RANKING),
+            })
+
+    if h != orig_h or nw != orig_nw:
+        print(f"      [CV]  Short series (n={n}): CV params reduced from "
+              f"h={orig_h}/nw={orig_nw} → h={h}/nw={nw}")
+
+    sf = StatsForecast(
+        models=build_models(cfg["season_length"]),
+        freq=cfg["freq"],
+        fallback_model=WindowAverage(window_size=6),
+        n_jobs=1,
+    )
+    cv = sf.cross_validation(df=sf_df, h=h, n_windows=nw, step_size=step)
+    model_cols = [c for c in cv.columns if c not in ["unique_id", "ds", "cutoff", "y"]]
+    mae = {m: np.mean(np.abs(cv[m] - cv["y"])) for m in model_cols}
+    return (
+        pd.DataFrame.from_dict(mae, orient="index", columns=["MAE"])
+        .sort_values("MAE")
+        .reset_index()
+        .rename(columns={"index": "Model"})
+    )
+
+
+# =============================================================================
+# SECTION 5 – FORECAST GENERATION
+# =============================================================================
+
+def forecast_series(sf_df: pd.DataFrame, top_models: list, cfg: dict) -> pd.DataFrame:
+    sf = StatsForecast(
+        models=build_models(cfg["season_length"]),
+        freq=cfg["freq"],
+        fallback_model=WindowAverage(window_size=6),
+        n_jobs=1,
+    )
+    fcst = sf.forecast(df=sf_df, h=cfg["forecast_horizon"])
+    # Newer statsforecast versions may return ds as the index – normalise to column
+    if "ds" not in fcst.columns:
+        fcst = fcst.reset_index()
+    if "ds" not in fcst.columns:
+        for col in fcst.columns:
+            if col not in ["unique_id"] and pd.api.types.is_datetime64_any_dtype(fcst[col]):
+                fcst = fcst.rename(columns={col: "ds"})
+                break
+    if "ds" not in fcst.columns:
+        last_date = pd.to_datetime(sf_df["ds"].max())
+        fcst.insert(0, "ds",
+                    pd.date_range(last_date, periods=cfg["forecast_horizon"] + 1,
+                                  freq=cfg["freq"])[1:])
+    fcst["ds"] = pd.to_datetime(fcst["ds"])
+    available = [m for m in top_models if m in fcst.columns]
+    fcst["ensemble_baseline_forecast"] = fcst[available].mean(axis=1).clip(lower=0)
+    return fcst
+
+
+# =============================================================================
+# SECTION 6 – OUTPUT FORMATTER
+# =============================================================================
+
+def build_output_rows(wide_df, fcst_df, dim_meta, cfg) -> pd.DataFrame:
+    last_price = wide_df["price"].dropna().tail(3).mean()
+    rows = []
+    for _, row in fcst_df.iterrows():
+        vol   = round(row["ensemble_baseline_forecast"], 6)
+        price = round(last_price, 7)
+        value = round(vol * price, 3)
+        rows.append({
+            cfg["channel_col"]     : dim_meta[cfg["channel_col"]],
+            cfg["category_col"]    : dim_meta[cfg["category_col"]],
+            cfg["division_col"]    : dim_meta[cfg["division_col"]],
+            cfg["brand_group_col"] : dim_meta[cfg["brand_group_col"]],
+            cfg["account_col"]     : dim_meta[cfg["account_col"]],
+            cfg["key_col"]         : dim_meta[cfg["key_col"]],
+            cfg["design_brand_col"]: dim_meta[cfg["design_brand_col"]],
+            cfg["date_col"]        : row["ds"].strftime("%Y-%m-%d"),
+            cfg["fy_col"]          : get_fy(row["ds"]),
+            cfg["volume_col"]      : vol,
+            cfg["price_col"]       : price,
+            cfg["value_col"]       : value,
+            cfg["driver_col"]      : cfg["forecast_label"],
+        })
+    return pd.DataFrame(rows)
+
+
+# =============================================================================
+# SECTION 7 – MAIN PIPELINE  (returns JSON to stdout)
+# =============================================================================
+
+def run_pipeline(source, cfg: dict = CONFIG) -> dict:
+
+    # ── Step 1: Load ────────────────────────────────────────────────────
+    print("\n[STEP 1/5]  Loading data...")
+    raw_df = load_data(source, cfg)
+    print(f"            Rows loaded    : {len(raw_df):,}")
+    print(f"            Columns        : {raw_df.shape[1]}")
+    print(f"            Date range     : {raw_df[cfg['date_col']].min().date()} "
+          f"→ {raw_df[cfg['date_col']].max().date()}")
+
+    # ── Step 2: Preprocess ──────────────────────────────────────────────
+    print("\n[STEP 2/5]  Preprocessing (long → wide pivot)...")
+    wide_df = preprocess_long_to_wide(raw_df, cfg)
+    total_series = wide_df["unique_id"].nunique()
+    print(f"            Unique series found : {total_series}  "
+          f"(Channel × Design Brand × Account)")
+
+    # ── Step 3: Validate ────────────────────────────────────────────────
+    print("\n[STEP 3/5]  Running pre-flight validation...")
+    valid_ids, skipped_series = validate_series(wide_df, cfg)
+
+    if not valid_ids:
+        msg = ("All series were skipped due to insufficient data. "
+               f"Minimum required: {cfg['min_data_points']} data points.")
+        print(f"\n[ERROR]  {msg}")
+        output = {
+            "status"        : "error",
+            "message"       : msg,
+            "skipped_series": skipped_series,
+        }
+        print(json.dumps(output))
+        return output
+
+    # ── Step 4: Model + Forecast ────────────────────────────────────────
+    print(f"\n[STEP 4/5]  Forecasting {len(valid_ids)} series "
+          f"(horizon = {cfg['forecast_horizon']} months)...")
+
+    all_forecast_rows = []
+    mae_summary       = []
+    series_results    = []
+
+    for i, uid in enumerate(valid_ids, 1):
+        print(f"\n  [{i:>3}/{len(valid_ids)}]  {uid}")
+
+        s_df = wide_df[wide_df["unique_id"] == uid].copy()
+        s_df = check_gaps(s_df, uid)
+        n_pts = s_df["baseline"].replace(0, np.nan).dropna().shape[0]
+        print(f"      Data    : {n_pts} valid baseline months  "
+              f"({s_df['ds'].min().strftime('%b %Y')} – "
+              f"{s_df['ds'].max().strftime('%b %Y')})")
+
+        sf_input = s_df[["unique_id", "ds", "baseline"]].rename(columns={"baseline": "y"})
+
+        # Cross-validation
+        print(f"      CV      : ranking models by MAE...")
+        mae_df = cross_validate_series(sf_input, cfg)
+        top_models = mae_df["Model"].head(cfg["top_n_ensemble"]).tolist()
+        mae_df["series"] = uid
+        mae_summary.append(mae_df)
+        print(f"      Ensemble: {' + '.join(top_models)}")
+
+        # Forecast
+        print(f"      Forecast: generating {cfg['forecast_horizon']}-month baseline...")
+        fcst_df = forecast_series(sf_input, top_models, cfg)
+
+        dim_meta = s_df.iloc[0][[
+            cfg["channel_col"], cfg["category_col"], cfg["division_col"],
+            cfg["brand_group_col"], cfg["account_col"], cfg["key_col"],
+            cfg["design_brand_col"],
+        ]].to_dict()
+
+        out_rows = build_output_rows(s_df, fcst_df, dim_meta, cfg)
+        all_forecast_rows.append(out_rows)
+
+        total_vol = round(out_rows[cfg["volume_col"]].sum(), 2)
+        print(f"      Result  : Total forecast volume = {total_vol:,.2f}")
+
+        # Historical data for this series (Baseline + Actual)
+        hist_baseline = s_df[["ds", "baseline", "actual", "price"]].copy()
+        hist_baseline["ds"] = hist_baseline["ds"].dt.strftime("%Y-%m-%d")
+
+        series_results.append({
+            "uid"              : uid,
+            "channel"          : dim_meta[cfg["channel_col"]],
+            "design_brand"     : dim_meta[cfg["design_brand_col"]],
+            "account"          : dim_meta[cfg["account_col"]],
+            "top_models"       : top_models,
+            "n_data_points"    : n_pts,
+            "mae_table"        : mae_df[["Model", "MAE"]].round(2).to_dict(orient="records"),
+            "history"          : hist_baseline.rename(columns={
+                "ds": "date", "baseline": "baseline_vol",
+                "actual": "actual_vol", "price": "price"
+            }).to_dict(orient="records"),
+            "forecast"         : [
+                {
+                    "date"  : r[cfg["date_col"]],
+                    "fy"    : r[cfg["fy_col"]],
+                    "volume": round(r[cfg["volume_col"]], 2),
+                    "price" : round(r[cfg["price_col"]], 4),
+                    "value" : round(r[cfg["value_col"]], 0),
+                }
+                for _, r in out_rows.iterrows()
+            ],
+            "total_forecast_vol": total_vol,
+        })
+
+    # ── Step 5: Build output payload ────────────────────────────────────
+    print(f"\n[STEP 5/5]  Building output payload...")
+
+    # Decomp data for Page 1 (all drivers, not just Baseline)
+    decomp_drivers = raw_df[~raw_df[cfg["driver_col"]].isin([
+        cfg["actual_label"], "Predicted Volume"
+    ])].copy()
+    decomp_drivers["DATE_str"] = decomp_drivers[cfg["date_col"]].dt.strftime("%Y-%m-%d")
+
+    decomp_records = decomp_drivers[[
+        cfg["channel_col"], cfg["design_brand_col"], cfg["account_col"],
+        "DATE_str", cfg["fy_col"],
+        "Driver Level 1", "Driver Level 2", cfg["driver_col"],
+        cfg["volume_col"], cfg["price_col"], cfg["value_col"],
+    ]].rename(columns={
+        cfg["channel_col"]     : "channel",
+        cfg["design_brand_col"]: "design_brand",
+        cfg["account_col"]     : "account",
+        "DATE_str"             : "date",
+        cfg["fy_col"]          : "fy",
+        cfg["driver_col"]      : "driver_l3",
+        cfg["volume_col"]      : "volume",
+        cfg["price_col"]       : "price",
+        cfg["value_col"]       : "value",
+    }).to_dict(orient="records")
+
+    # Actual volume for overlay
+    actual_records = raw_df[raw_df[cfg["driver_col"]] == cfg["actual_label"]].copy()
+    actual_records["DATE_str"] = actual_records[cfg["date_col"]].dt.strftime("%Y-%m-%d")
+    actual_out = actual_records[[
+        cfg["design_brand_col"], cfg["account_col"], "DATE_str", cfg["volume_col"]
+    ]].rename(columns={
+        cfg["design_brand_col"]: "design_brand",
+        cfg["account_col"]     : "account",
+        "DATE_str"             : "date",
+        cfg["volume_col"]      : "actual_volume",
+    }).to_dict(orient="records")
+
+    # Final summary
+    total_fcst_vol = sum(r["total_forecast_vol"] for r in series_results)
+    print(f"\n{'='*70}")
+    print(f"  PIPELINE COMPLETE")
+    print(f"  Series forecast   : {len(series_results)}")
+    print(f"  Series skipped    : {len(skipped_series)}")
+    print(f"  Total forecast vol: {total_fcst_vol:,.2f}")
+    print(f"{'='*70}\n")
+
+    output = {
+        "status"          : "success",
+        "n_series"        : len(valid_ids),
+        "n_skipped"       : len(skipped_series),
+        "series_ids"      : valid_ids,
+        "skipped_series"  : skipped_series,
+        "series"          : series_results,
+        "decomp_data"     : decomp_records,
+        "actual_data"     : actual_out,
+    }
+
+    print(json.dumps(output))
+    return output
+
+
+# =============================================================================
+# ENTRY POINT  – run with:  python dabur_baseline_forecast_v2.py data.xlsx
+# =============================================================================
+
+if __name__ == "__main__":
+    src = sys.argv[1] if len(sys.argv) > 1 else None
+    if src is None:
+        print(json.dumps({
+            "status" : "error",
+            "message": "No input file provided. Usage: python dabur_baseline_forecast_v2.py <path_to_excel>"
+        }))
+        sys.exit(1)
+    run_pipeline(src)

dabur_demand_platform.html

@@ -0,0 +1,1273 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+<meta charset="UTF-8"/>
+<meta name="viewport" content="width=device-width,initial-scale=1.0"/>
+<title>Dabur · Demand Intelligence Platform</title>
+<link rel="preconnect" href="https://fonts.googleapis.com"/>
+<link href="https://fonts.googleapis.com/css2?family=DM+Sans:wght@300;400;500;600&family=DM+Mono:wght@400;500&display=swap" rel="stylesheet"/>
+<!-- Pyodide removed: pipeline now runs via local Flask server on port 5050 -->
+<script src="https://cdnjs.cloudflare.com/ajax/libs/Chart.js/4.4.1/chart.umd.min.js"></script>
+<style>
+*,*::before,*::after{box-sizing:border-box;margin:0;padding:0}
+:root{
+  --bg:#F4F3EF;
+  --white:#FFFFFF;
+  --surface:#FAFAF8;
+  --border:#E2E0D8;
+  --border2:#C8C5BA;
+  --text:#1A1916;
+  --text2:#6B6860;
+  --text3:#A39E92;
+  --accent:#C84B31;
+  --accent-soft:#F5E6E2;
+  --green:#2D6A4F;
+  --green-soft:#E8F4EE;
+  --blue:#1D4E89;
+  --blue-soft:#E5EDF7;
+  --amber:#C77A06;
+  --amber-soft:#FEF3D8;
+  --teal:#1A7A6E;
+  --teal-soft:#E3F4F1;
+  --font:'DM Sans',sans-serif;
+  --mono:'DM Mono',monospace;
+  --r:8px;
+  --r2:12px;
+  --shadow:0 1px 3px rgba(0,0,0,0.08),0 1px 2px rgba(0,0,0,0.04);
+  --shadow2:0 4px 16px rgba(0,0,0,0.1);
+}
+body{background:var(--bg);color:var(--text);font-family:var(--font);min-height:100vh;font-size:14px;line-height:1.5}
+
+/* ── TOPBAR ── */
+.topbar{height:56px;background:var(--white);border-bottom:1px solid var(--border);display:flex;align-items:center;justify-content:space-between;padding:0 24px;position:sticky;top:0;z-index:200;box-shadow:var(--shadow)}
+.topbar-brand{display:flex;align-items:center;gap:10px}
+.brand-logo{width:32px;height:32px;background:var(--accent);border-radius:6px;display:flex;align-items:center;justify-content:center;color:#fff;font-weight:600;font-size:13px;letter-spacing:0.02em}
+.brand-name{font-size:15px;font-weight:600;letter-spacing:-0.01em}
+.brand-sub{font-size:11px;color:var(--text3);font-family:var(--mono)}
+.topbar-nav{display:flex;gap:2px}
+.nav-item{padding:6px 16px;border-radius:6px;font-size:13px;font-weight:500;cursor:pointer;border:none;background:none;color:var(--text2);transition:all 0.15s;display:flex;align-items:center;gap:6px}
+.nav-item:hover{background:var(--bg);color:var(--text)}
+.nav-item.active{background:var(--accent-soft);color:var(--accent)}
+.nav-badge{background:var(--accent);color:#fff;border-radius:99px;padding:1px 6px;font-size:10px;font-family:var(--mono)}
+.topbar-right{display:flex;align-items:center;gap:8px}
+.status-chip{display:flex;align-items:center;gap:6px;padding:4px 10px;border-radius:99px;border:1px solid var(--border);font-size:12px;font-family:var(--mono);color:var(--text3)}
+.status-dot{width:6px;height:6px;border-radius:50%}
+.sd-loading{background:var(--amber);animation:blink 1s infinite}
+.sd-ready{background:var(--green)}
+.sd-error{background:var(--accent)}
+@keyframes blink{0%,100%{opacity:1}50%{opacity:0.3}}
+
+/* ── PAGES ── */
+.page{display:none;min-height:calc(100vh - 56px);padding:24px}
+.page.active{display:block}
+
+/* ── SECTION HEADER ── */
+.section-header{margin-bottom:20px}
+.section-header h1{font-size:20px;font-weight:600;letter-spacing:-0.02em;margin-bottom:4px}
+.section-header p{font-size:13px;color:var(--text2)}
+
+/* ── CARDS ── */
+.card{background:var(--white);border:1px solid var(--border);border-radius:var(--r2);box-shadow:var(--shadow)}
+.card-header{padding:14px 20px;border-bottom:1px solid var(--border);display:flex;align-items:center;justify-content:space-between}
+.card-title{font-size:13px;font-weight:600;color:var(--text)}
+.card-body{padding:20px}
+
+/* ── GRID ── */
+.grid-2{display:grid;grid-template-columns:1fr 1fr;gap:16px}
+.grid-3{display:grid;grid-template-columns:repeat(3,1fr);gap:16px}
+.grid-4{display:grid;grid-template-columns:repeat(4,1fr);gap:12px}
+
+/* ── METRIC CARDS ── */
+.metric{background:var(--white);border:1px solid var(--border);border-radius:var(--r2);padding:16px 20px}
+.metric-label{font-size:11px;color:var(--text3);font-weight:500;text-transform:uppercase;letter-spacing:0.06em;margin-bottom:6px}
+.metric-value{font-size:26px;font-weight:600;font-family:var(--mono);letter-spacing:-0.02em;line-height:1}
+.metric-sub{font-size:11px;color:var(--text3);margin-top:4px;font-family:var(--mono)}
+.metric.accent .metric-value{color:var(--accent)}
+.metric.green  .metric-value{color:var(--green)}
+.metric.blue   .metric-value{color:var(--blue)}
+.metric.amber  .metric-value{color:var(--amber)}
+
+/* ── FILE UPLOAD ZONE ── */
+.upload-zone{border:1.5px dashed var(--border2);border-radius:var(--r2);padding:32px;text-align:center;cursor:pointer;transition:all 0.2s;position:relative;background:var(--surface)}
+.upload-zone:hover,.upload-zone.drag{border-color:var(--accent);background:var(--accent-soft)}
+.upload-zone input{position:absolute;inset:0;opacity:0;cursor:pointer;width:100%;height:100%}
+.upload-icon{font-size:28px;margin-bottom:10px;display:block}
+.upload-title{font-size:14px;font-weight:500;margin-bottom:4px}
+.upload-hint{font-size:12px;color:var(--text3);font-family:var(--mono)}
+.file-loaded{display:flex;align-items:center;gap:10px;padding:10px 14px;background:var(--green-soft);border:1px solid #C3E6D4;border-radius:var(--r);margin-top:10px;font-size:12px;font-family:var(--mono);color:var(--green)}
+.file-dot{width:8px;height:8px;border-radius:50%;background:var(--green);flex-shrink:0}
+
+/* ── BUTTONS ── */
+.btn{display:inline-flex;align-items:center;gap:6px;padding:8px 16px;border-radius:var(--r);font-size:13px;font-weight:500;cursor:pointer;border:none;font-family:var(--font);transition:all 0.15s}
+.btn-primary{background:var(--accent);color:#fff}
+.btn-primary:hover{background:#b04028}
+.btn-primary:active{transform:scale(0.98)}
+.btn-primary:disabled{background:var(--border2);color:var(--text3);cursor:not-allowed;transform:none}
+.btn-outline{background:#fff;color:var(--text);border:1px solid var(--border2)}
+.btn-outline:hover{background:var(--bg);border-color:var(--text3)}
+.btn-sm{padding:5px 12px;font-size:12px}
+.btn-run{background:var(--green);color:#fff;font-weight:600;padding:10px 24px}
+.btn-run:hover{background:#22573e}
+.btn-run:disabled{background:var(--border2);color:var(--text3);cursor:not-allowed}
+
+/* ── FILTERS ── */
+.filter-bar{display:flex;align-items:center;gap:10px;flex-wrap:wrap;margin-bottom:16px}
+.filter-label{font-size:12px;font-weight:500;color:var(--text3)}
+select{border:1px solid var(--border);border-radius:var(--r);padding:6px 10px;font-size:12px;font-family:var(--font);color:var(--text);background:var(--white);cursor:pointer;outline:none}
+select:focus{border-color:var(--accent)}
+
+/* ── CHART CANVAS ── */
+.chart-wrap{position:relative;width:100%}
+.chart-wrap canvas{width:100%!important}
+
+/* ── STACKED BAR DECOMP ── */
+.decomp-legend{display:flex;flex-wrap:wrap;gap:8px;margin-bottom:12px}
+.legend-item{display:flex;align-items:center;gap:5px;font-size:11px;color:var(--text2)}
+.legend-dot{width:10px;height:10px;border-radius:2px;flex-shrink:0}
+
+/* ── CODE EDITOR ── */
+.editor-wrap{background:#1E1F23;border-radius:var(--r2);overflow:hidden}
+.editor-bar{background:#2A2B30;padding:8px 16px;display:flex;align-items:center;justify-content:space-between;border-bottom:1px solid #363840}
+.editor-filename{font-size:12px;font-family:var(--mono);color:#9DA3B0}
+.editor-actions{display:flex;gap:6px}
+.editor-btn{padding:4px 12px;border-radius:5px;font-size:11px;font-family:var(--font);font-weight:600;cursor:pointer;border:none;transition:all 0.15s}
+.btn-code-run{background:#2D6A4F;color:#fff}
+.btn-code-run:hover{background:#3a8562}
+.btn-code-run:disabled{background:#363840;color:#555;cursor:not-allowed}
+.btn-code-ghost{background:#363840;color:#9DA3B0}
+.btn-code-ghost:hover{background:#404248;color:#E0E4EE}
+textarea.editor{background:#1E1F23;color:#E0E4EE;font-family:var(--mono);font-size:12.5px;line-height:1.7;border:none;outline:none;resize:none;width:100%;padding:16px;tab-size:4;min-height:360px}
+textarea.editor::selection{background:rgba(200,75,49,0.25)}
+
+/* ── CONSOLE ── */
+.console{background:#0F1012;border-radius:0 0 var(--r2) var(--r2);max-height:180px;overflow-y:auto;padding:12px 16px;font-family:var(--mono);font-size:11.5px;line-height:1.8}
+.c-dim{color:#4A4F5C}
+.c-info{color:#4FA3C8}
+.c-ok{color:#7AC59A}
+.c-warn{color:#D4A853}
+.c-err{color:#E06060}
+
+/* ── TABLES ── */
+.data-table{width:100%;border-collapse:collapse;font-size:12px}
+.data-table th{font-size:10px;font-weight:600;color:var(--text3);text-transform:uppercase;letter-spacing:0.06em;padding:8px 12px;border-bottom:1px solid var(--border);text-align:left;background:var(--surface)}
+.data-table td{padding:9px 12px;border-bottom:1px solid var(--border);color:var(--text);font-family:var(--mono);font-size:12px}
+.data-table tr:last-child td{border-bottom:none}
+.data-table tr:hover td{background:var(--bg)}
+
+/* ── PILLS / BADGES ── */
+.pill{display:inline-flex;align-items:center;padding:2px 8px;border-radius:99px;font-size:10px;font-weight:600;font-family:var(--mono)}
+.pill-green{background:var(--green-soft);color:var(--green)}
+.pill-blue{background:var(--blue-soft);color:var(--blue)}
+.pill-amber{background:var(--amber-soft);color:var(--amber)}
+.pill-red{background:var(--accent-soft);color:var(--accent)}
+.pill-teal{background:var(--teal-soft);color:var(--teal)}
+
+/* ── PROGRESS ── */
+.progress-bar{height:6px;background:var(--bg);border-radius:3px;overflow:hidden}
+.progress-fill{height:100%;border-radius:3px;transition:width 0.5s ease}
+
+/* ── SERIES TABS ── */
+.series-tabs{display:flex;gap:4px;margin-bottom:16px}
+.stab{padding:6px 14px;border-radius:6px;font-size:12px;font-weight:500;cursor:pointer;border:1px solid var(--border);background:var(--white);color:var(--text2);transition:all 0.15s}
+.stab:hover{border-color:var(--text3)}
+.stab.active{background:var(--accent);color:#fff;border-color:var(--accent)}
+
+/* ── EMPTY STATE ── */
+.empty{text-align:center;padding:48px 24px;color:var(--text3)}
+.empty-icon{font-size:36px;margin-bottom:12px;display:block;opacity:0.5}
+.empty h3{font-size:15px;font-weight:600;color:var(--text2);margin-bottom:6px}
+.empty p{font-size:12px;line-height:1.6}
+
+/* ── SECTION DIVIDER ── */
+.divider{height:1px;background:var(--border);margin:20px 0}
+
+/* ── FORECAST OUTPUT ── */
+.forecast-summary{display:grid;grid-template-columns:repeat(5,1fr);gap:12px;margin-bottom:20px}
+.model-badge{display:inline-flex;align-items:center;gap:4px;padding:2px 8px;border-radius:4px;font-size:11px;font-family:var(--mono);background:var(--blue-soft);color:var(--blue);margin-right:4px}
+
+/* ── PAGE 4 PLACEHOLDER ── */
+.coming-soon{min-height:60vh;display:flex;align-items:center;justify-content:center;flex-direction:column;gap:16px;text-align:center}
+.coming-soon .big-icon{font-size:60px;opacity:0.15}
+.coming-soon h2{font-size:22px;font-weight:600;color:var(--text2)}
+.coming-soon p{font-size:14px;color:var(--text3);max-width:400px;line-height:1.7}
+.coming-soon .placeholder-boxes{display:grid;grid-template-columns:repeat(3,1fr);gap:16px;width:100%;max-width:700px;margin-top:20px}
+.ph-box{background:var(--white);border:1.5px dashed var(--border2);border-radius:var(--r2);padding:24px;text-align:center}
+.ph-box .ph-icon{font-size:24px;margin-bottom:8px;opacity:0.3}
+.ph-box p{font-size:12px;color:var(--text3)}
+
+/* ── SCROLLBARS ── */
+::-webkit-scrollbar{width:5px;height:5px}
+::-webkit-scrollbar-track{background:transparent}
+::-webkit-scrollbar-thumb{background:var(--border2);border-radius:3px}
+
+/* ── RESPONSIVE ── */
+@media(max-width:900px){
+  .grid-4{grid-template-columns:repeat(2,1fr)}
+  .forecast-summary{grid-template-columns:repeat(3,1fr)}
+}
+
+/* ── LOGIN SCREEN ── */
+#login-screen{position:fixed;inset:0;z-index:1000;background:var(--bg);display:flex;align-items:center;justify-content:center}
+#login-screen.hidden{display:none}
+.login-box{background:var(--white);border:1px solid var(--border);border-radius:16px;box-shadow:var(--shadow2);padding:40px;width:380px}
+.login-logo{width:48px;height:48px;background:var(--accent);border-radius:10px;display:flex;align-items:center;justify-content:center;color:#fff;font-weight:700;font-size:18px;margin:0 auto 16px}
+.login-title{text-align:center;font-size:20px;font-weight:600;letter-spacing:-0.02em;margin-bottom:4px}
+.login-sub{text-align:center;font-size:12px;color:var(--text3);font-family:var(--mono);margin-bottom:28px}
+.login-field{margin-bottom:14px}
+.login-label{font-size:11px;font-weight:600;color:var(--text3);text-transform:uppercase;letter-spacing:0.06em;margin-bottom:5px;display:block}
+.login-input{width:100%;border:1px solid var(--border);border-radius:var(--r);padding:9px 12px;font-size:13px;font-family:var(--font);color:var(--text);background:var(--white);outline:none;transition:border 0.15s}
+.login-input:focus{border-color:var(--accent)}
+.login-btn{width:100%;padding:10px;border-radius:var(--r);background:var(--accent);color:#fff;font-size:14px;font-weight:600;border:none;cursor:pointer;font-family:var(--font);transition:background 0.15s;margin-top:4px}
+.login-btn:hover{background:#b04028}
+.login-btn:disabled{background:var(--border2);cursor:not-allowed}
+.login-error{color:var(--accent);font-size:12px;text-align:center;margin-top:10px;min-height:18px;font-family:var(--mono)}
+
+/* ── USER CHIP in topbar ── */
+.user-chip{display:flex;align-items:center;gap:8px;padding:4px 12px 4px 8px;border-radius:99px;border:1px solid var(--border);background:var(--white);font-size:12px;cursor:pointer;transition:background 0.15s}
+.user-chip:hover{background:var(--bg)}
+.user-avatar{width:22px;height:22px;border-radius:50%;background:var(--accent);color:#fff;font-size:10px;font-weight:700;display:flex;align-items:center;justify-content:center}
+.user-name{font-weight:500;color:var(--text)}
+.user-role{font-family:var(--mono);color:var(--text3);font-size:10px}
+
+</style>
+</head>
+<body>
+
+<!-- ══════════════════════════════════════════════════════════════════
+  LOGIN SCREEN
+══════════════════════════════════════════════════════════════════ -->
+<div id="login-screen">
+  <div class="login-box">
+    <div class="login-logo">DB</div>
+    <div class="login-title">Demand Intelligence Platform</div>
+    <div class="login-sub">Dabur · MMM Decomp & Forecasting</div>
+    <div class="login-field">
+      <label class="login-label">Username</label>
+      <input class="login-input" id="login-user" type="text" placeholder="analyst or viewer" autocomplete="username"/>
+    </div>
+    <div class="login-field">
+      <label class="login-label">Password</label>
+      <input class="login-input" id="login-pass" type="password" placeholder="••••••••••" autocomplete="current-password"/>
+    </div>
+    <button class="login-btn" id="login-btn" onclick="doLogin()">Sign In</button>
+    <div class="login-error" id="login-error"></div>
+  </div>
+</div>
+
+
+<!-- ══════════════════════════════════════════════════════════════════
+  TOP BAR
+══════════════════════════════════════════════════════════════════ -->
+<div class="topbar">
+  <div class="topbar-brand">
+    <div class="brand-logo">DB</div>
+    <div>
+      <div class="brand-name">Demand Intelligence Platform</div>
+      <div class="brand-sub">Dabur · MMM Decomp & Forecasting</div>
+    </div>
+  </div>
+  <div class="topbar-nav" id="topbar-nav">
+    <button class="nav-item active" id="nav-p1" onclick="goPage('p1',this)">📊 Decomposition</button>
+    <button class="nav-item" id="nav-p2" onclick="goPage('p2',this)">⚙️ Forecast Engine</button>
+    <button class="nav-item" id="nav-p3" onclick="goPage('p3',this)">📈 Forecast Output</button>
+    <button class="nav-item" id="nav-p4" onclick="goPage('p4',this)">🗺️ Full Demand <span class="nav-badge">Soon</span></button>
+  </div>
+  <div class="topbar-right" style="gap:10px">
+    <div class="status-chip" id="status-chip" style="display:none">
+      <div class="status-dot sd-loading" id="sdot"></div>
+      <span id="stext">loading…</span>
+    </div>
+    <div class="user-chip" id="user-chip" style="display:none" onclick="doLogout()">
+      <div class="user-avatar" id="user-avatar">?</div>
+      <div>
+        <div class="user-name" id="user-name-label">—</div>
+        <div class="user-role" id="user-role-label">—</div>
+      </div>
+      <span style="font-size:10px;color:var(--text3);margin-left:2px">⏏</span>
+    </div>
+  </div>
+</div>
+
+
+<!-- ══════════════════════════════════════════════════════════════════
+  PAGE 1 – DECOMPOSED VOLUME VISUALISATION
+══════════════════════════════════════════════════════════════════ -->
+<div class="page active" id="p1">
+  <div class="section-header">
+    <h1>Volume Decomposition by Drivers</h1>
+    <p>Historical baseline and driver contribution breakdown from the input decomp file</p>
+  </div>
+
+  <!-- File upload -->
+  <div class="card" style="margin-bottom:16px">
+    <div class="card-header"><span class="card-title">Upload Decomposition File</span><span class="pill pill-amber">Required to start</span></div>
+    <div class="card-body">
+      <div class="grid-2" style="gap:20px;align-items:start">
+        <div>
+          <div class="upload-zone" id="decomp-upload-zone">
+            <input type="file" id="decomp-file-input" accept=".xlsx,.csv"/>
+            <span class="upload-icon">📂</span>
+            <div class="upload-title">Drop your decomposition file here</div>
+            <div class="upload-hint">.xlsx (Base Sheet) · .csv supported</div>
+          </div>
+          <div id="decomp-file-loaded" style="display:none" class="file-loaded">
+            <div class="file-dot"></div><span id="decomp-file-name">—</span>
+          </div>
+        </div>
+        <div>
+          <p style="font-size:12px;color:var(--text2);line-height:1.8;margin-bottom:12px">
+            Expected sheet: <span style="font-family:var(--mono);color:var(--text)">Base Sheet</span><br>
+            Key columns: <span style="font-family:var(--mono);color:var(--text)">Channel, Design Brand, Account, DATE, Volume, Driver Level 3</span><br>
+            Drivers read: <span style="font-family:var(--mono);color:var(--text)">Baseline, Actual Volume, and all incremental drivers</span>
+          </p>
+          <button class="btn btn-primary" id="load-decomp-btn" onclick="loadDecompFile()" disabled>Load &amp; Visualise</button>
+        </div>
+      </div>
+    </div>
+  </div>
+
+  <!-- Filters -->
+  <div class="filter-bar" id="decomp-filters" style="display:none">
+    <span class="filter-label">Filter:</span>
+    <select id="f-channel" onchange="renderDecomp()"><option value="ALL">All Channels</option></select>
+    <select id="f-brand" onchange="renderDecomp()"><option value="ALL">All Brands</option></select>
+    <select id="f-account" onchange="renderDecomp()"><option value="ALL">All Accounts</option></select>
+    <select id="f-fy" onchange="renderDecomp()"><option value="ALL">All FY</option></select>
+  </div>
+
+  <!-- KPI row -->
+  <div class="grid-4" id="decomp-kpis" style="display:none;margin-bottom:16px">
+    <div class="metric accent"><div class="metric-label">Total Actual Volume</div><div class="metric-value" id="kpi-actual">—</div><div class="metric-sub">all periods</div></div>
+    <div class="metric green"><div class="metric-label">Total Baseline Vol</div><div class="metric-value" id="kpi-baseline">—</div><div class="metric-sub">structural demand</div></div>
+    <div class="metric blue"><div class="metric-label">Incremental Vol</div><div class="metric-value" id="kpi-incr">—</div><div class="metric-sub">promo + visibility + price</div></div>
+    <div class="metric amber"><div class="metric-label">Baseline Share</div><div class="metric-value" id="kpi-bshare">—</div><div class="metric-sub">of actual volume</div></div>
+  </div>
+
+  <!-- Charts -->
+  <div id="decomp-charts" style="display:none">
+    <div class="grid-2" style="margin-bottom:16px">
+      <div class="card">
+        <div class="card-header"><span class="card-title">Stacked Driver Decomposition · Monthly</span></div>
+        <div class="card-body"><div class="chart-wrap" style="height:280px"><canvas id="stackedChart"></canvas></div></div>
+      </div>
+      <div class="card">
+        <div class="card-header"><span class="card-title">Actual vs Baseline Volume</span></div>
+        <div class="card-body"><div class="chart-wrap" style="height:280px"><canvas id="actualBaseChart"></canvas></div></div>
+      </div>
+    </div>
+    <div class="card" style="margin-bottom:16px">
+      <div class="card-header">
+        <span class="card-title">Driver Contribution Share</span>
+        <div id="decomp-legend" class="decomp-legend" style="margin-bottom:0"></div>
+      </div>
+      <div class="card-body"><div class="chart-wrap" style="height:220px"><canvas id="driverShareChart"></canvas></div></div>
+    </div>
+    <div class="card">
+      <div class="card-header"><span class="card-title">Driver Contribution Summary</span></div>
+      <div class="card-body" style="padding:0 0 8px">
+        <table class="data-table" id="driver-table"><thead><tr><th>Driver</th><th>Level 1</th><th>Level 2</th><th>Total Volume</th><th>Avg Monthly</th><th>Share</th></tr></thead><tbody id="driver-tbody"></tbody></table>
+      </div>
+    </div>
+  </div>
+
+  <div id="decomp-empty" class="empty">
+    <span class="empty-icon">📊</span>
+    <h3>No data loaded</h3>
+    <p>Upload a decomposition Excel file above to see the driver visualisation</p>
+  </div>
+</div>
+
+
+<!-- ══════════════════════════════════════════════════════════════════
+  PAGE 2 – FORECAST ENGINE
+══════════════════════════════════════════════════════════════════ -->
+<div class="page" id="p2">
+  <div class="section-header">
+    <h1>Baseline Forecast Engine</h1>
+    <p>Upload the forecasting script and data file · run via in-browser Python interpreter (Pyodide)</p>
+  </div>
+
+  <div class="grid-2" style="gap:16px;margin-bottom:16px;align-items:start">
+    <!-- Upload panel -->
+    <div>
+      <div class="card" style="margin-bottom:12px">
+        <div class="card-header"><span class="card-title">Forecast Script</span><span class="pill pill-blue">.py</span></div>
+        <div class="card-body">
+          <div class="upload-zone" id="script-upload-zone">
+            <input type="file" id="script-file-input" accept=".py,.ipynb"/>
+            <span class="upload-icon">📜</span>
+            <div class="upload-title">Upload forecast script</div>
+            <div class="upload-hint">dabur_baseline_forecast_v2.py</div>
+          </div>
+          <div id="script-file-loaded" style="display:none" class="file-loaded">
+            <div class="file-dot"></div><span id="script-file-name">—</span>
+          </div>
+        </div>
+      </div>
+      <div class="card">
+        <div class="card-header"><span class="card-title">Decomp Input Data</span><span class="pill pill-amber">.xlsx</span></div>
+        <div class="card-body">
+          <div class="upload-zone" id="data-upload-zone">
+            <input type="file" id="data-file-input" accept=".xlsx,.csv"/>
+            <span class="upload-icon">📑</span>
+            <div class="upload-title">Upload decomp Excel file</div>
+            <div class="upload-hint">Base Sheet · same file as Page 1</div>
+          </div>
+          <div id="data-file-loaded" style="display:none" class="file-loaded">
+            <div class="file-dot"></div><span id="data-file-name">—</span>
+          </div>
+        </div>
+      </div>
+    </div>
+
+    <!-- Config panel -->
+    <div class="card" style="height:100%">
+      <div class="card-header"><span class="card-title">Run Configuration</span></div>
+      <div class="card-body">
+        <div style="display:grid;grid-template-columns:1fr 1fr;gap:12px;margin-bottom:16px">
+          <div>
+            <div style="font-size:11px;font-weight:600;color:var(--text3);margin-bottom:4px;text-transform:uppercase;letter-spacing:0.06em">Forecast horizon (months)</div>
+            <input type="number" id="cfg-horizon" value="12" min="1" max="36" style="width:100%;border:1px solid var(--border);border-radius:var(--r);padding:7px 10px;font-size:13px;font-family:var(--mono)"/>
+          </div>
+          <div>
+            <div style="font-size:11px;font-weight:600;color:var(--text3);margin-bottom:4px;text-transform:uppercase;letter-spacing:0.06em">Season length</div>
+            <input type="number" id="cfg-season" value="12" min="1" max="52" style="width:100%;border:1px solid var(--border);border-radius:var(--r);padding:7px 10px;font-size:13px;font-family:var(--mono)"/>
+          </div>
+          <div>
+            <div style="font-size:11px;font-weight:600;color:var(--text3);margin-bottom:4px;text-transform:uppercase;letter-spacing:0.06em">CV horizon</div>
+            <input type="number" id="cfg-cvh" value="6" min="1" max="12" style="width:100%;border:1px solid var(--border);border-radius:var(--r);padding:7px 10px;font-size:13px;font-family:var(--mono)"/>
+          </div>
+          <div>
+            <div style="font-size:11px;font-weight:600;color:var(--text3);margin-bottom:4px;text-transform:uppercase;letter-spacing:0.06em">Top N ensemble</div>
+            <input type="number" id="cfg-topn" value="3" min="1" max="7" style="width:100%;border:1px solid var(--border);border-radius:var(--r);padding:7px 10px;font-size:13px;font-family:var(--mono)"/>
+          </div>
+        </div>
+        <div style="padding:10px 14px;background:var(--bg);border-radius:var(--r);font-size:12px;color:var(--text2);margin-bottom:16px;line-height:1.7">
+          Pipeline: AutoETS · AutoARIMA · AutoTheta · HoltWinters (Additive + Multiplicative) · MSTL+Theta · SeasonalNaive<br>
+          Ensemble: top-N models by cross-validation MAE · fallback: WindowAverage(6)
+        </div>
+        <button class="btn btn-run" id="run-pipeline-btn" onclick="runForecastPipeline()" disabled style="width:100%">▶ Run Baseline Forecast</button>
+      </div>
+    </div>
+  </div>
+
+  <!-- Editor -->
+  <div style="margin-bottom:16px">
+    <div class="editor-wrap">
+      <div class="editor-bar">
+        <span class="editor-filename" id="p2-editor-filename">no script loaded</span>
+        <div class="editor-actions">
+          <button class="editor-btn btn-code-ghost" onclick="loadDemoScript()">Load demo script</button>
+          <button class="editor-btn btn-code-ghost" onclick="document.getElementById('editor-code').value=''">Clear</button>
+          <button class="editor-btn btn-code-run" id="run-editor-btn" onclick="runForecastPipeline()" disabled>▶ Run</button>
+        </div>
+      </div>
+      <textarea class="editor" id="editor-code" spellcheck="false" placeholder="Upload a script or click 'Load demo script'…"></textarea>
+      <div class="console" id="p2-console"><div class="c-dim">» Pyodide interpreter initialising…</div></div>
+    </div>
+  </div>
+</div>
+
+
+<!-- ══════════════════════════════════════════════════════════════════
+  PAGE 3 – FORECAST OUTPUT
+══════════════════════════════════════════════════════════════════ -->
+<div class="page" id="p3">
+  <div class="section-header">
+    <h1>Baseline Forecast Output</h1>
+    <p>Results from the last pipeline run · 12-month baseline projection per series</p>
+  </div>
+
+  <div id="p3-empty" class="empty">
+    <span class="empty-icon">📈</span>
+    <h3>No forecast run yet</h3>
+    <p>Go to the Forecast Engine tab, upload your files and run the pipeline. Results will appear here automatically.</p>
+  </div>
+
+  <div id="p3-content" style="display:none">
+    <!-- Top KPIs -->
+    <div class="forecast-summary" id="p3-kpis" style="margin-bottom:16px"></div>
+
+    <!-- Series selector -->
+    <div class="series-tabs" id="p3-series-tabs"></div>
+
+    <!-- Per-series view -->
+    <div id="p3-series-view">
+      <div class="grid-2" style="margin-bottom:16px">
+        <div class="card">
+          <div class="card-header">
+            <span class="card-title" id="p3-chart-title">Baseline Forecast · 12-Month Horizon</span>
+            <div id="p3-top-models" style="display:flex;gap:4px;flex-wrap:wrap"></div>
+          </div>
+          <div class="card-body"><div class="chart-wrap" style="height:280px"><canvas id="forecastChart"></canvas></div></div>
+        </div>
+        <div class="card">
+          <div class="card-header"><span class="card-title">Model Accuracy (CV MAE)</span></div>
+          <div class="card-body" style="padding:0">
+            <table class="data-table" id="mae-table">
+              <thead><tr><th>Model</th><th>MAE</th><th>Rank</th></tr></thead>
+              <tbody id="mae-tbody"></tbody>
+            </table>
+          </div>
+        </div>
+      </div>
+      <div class="card" style="margin-bottom:16px">
+        <div class="card-header">
+          <span class="card-title">12-Month Forecast Table</span>
+          <button class="btn btn-outline btn-sm" onclick="downloadForecast()">⬇ Download CSV</button>
+        </div>
+        <div class="card-body" style="padding:0">
+          <table class="data-table" id="forecast-table">
+            <thead><tr><th>Date</th><th>FY</th><th>Forecast Volume</th><th>Price</th><th>Forecast Value</th><th>MoM Δ%</th></tr></thead>
+            <tbody id="forecast-tbody"></tbody>
+          </table>
+        </div>
+      </div>
+    </div>
+  </div>
+</div>
+
+
+<!-- ══════════════════════════════════════════════════════════════════
+  PAGE 4 – FULL DEMAND FORECAST (PLACEHOLDER)
+══════════════════════════════════════════════════════════════════ -->
+<div class="page" id="p4">
+  <div class="coming-soon">
+    <div class="big-icon">🗺️</div>
+    <h2>Full Demand Forecast</h2>
+    <p>This module will combine baseline forecast with plan data (CP spend, visibility, price) to produce total demand forecast across all channels and accounts.</p>
+    <div class="placeholder-boxes">
+      <div class="ph-box"><div class="ph-icon">📋</div><p>Plan Data Upload<br><span style="color:var(--text3);font-size:10px">CP · Visibility · Price Plans</span></p></div>
+      <div class="ph-box"><div class="ph-icon">⚡</div><p>Incremental Modelling<br><span style="color:var(--text3);font-size:10px">Promo uplift · ROI attribution</span></p></div>
+      <div class="ph-box"><div class="ph-icon">📊</div><p>Full Demand Output<br><span style="color:var(--text3);font-size:10px">Total Volume · Value · Share</span></p></div>
+    </div>
+    <p style="margin-top:24px;font-size:12px;background:var(--white);padding:10px 20px;border-radius:var(--r2);border:1px solid var(--border);color:var(--text3);font-family:var(--mono)">Status: Pending plan data integration · ETA: Next sprint</p>
+  </div>
+</div>
+
+
+<!-- ═══════════════════════════ JAVASCRIPT ═══════════════════════════ -->
+<script>
+// ─── GLOBAL STATE ─────────────────────────────────────────────────────
+const SERVER_URL = "";
+let decompData = null, actualData = null;
+let forecastResult = null;
+let activeSeriesIdx = 0;
+let dataFile = null;
+let chartInstances = {};
+let currentUser = null;   // { username, role }
+
+const DRIVER_COLORS = {
+  "Baseline"             : "#2D6A4F",
+  "Consumer Promotions"  : "#C84B31",
+  "Visibility"           : "#1D4E89",
+  "Rainfall"             : "#4A90D9",
+  "Own Price"            : "#C77A06",
+  "Baseline Forecast"    : "#7B4EA0",
+};
+const DEFAULT_COLORS = ["#E76F51","#264653","#2A9D8F","#E9C46A","#457B9D","#A8DADC","#F4A261"];
+
+// ─── AUTH ─────────────────────────────────────────────────────────────
+async function doLogin(){
+  const btn  = document.getElementById("login-btn");
+  const err  = document.getElementById("login-error");
+  const user = document.getElementById("login-user").value.trim();
+  const pass = document.getElementById("login-pass").value;
+  if(!user||!pass){ err.textContent="Enter username and password."; return; }
+  btn.disabled=true; btn.textContent="Signing in…"; err.textContent="";
+  try {
+    const r = await fetch("/login",{method:"POST",headers:{"Content-Type":"application/json"},body:JSON.stringify({username:user,password:pass})});
+    const j = await r.json();
+    if(j.status!=="ok") throw new Error(j.message||"Login failed");
+    onLoginSuccess(j);
+  } catch(e){
+    err.textContent = e.message;
+    btn.disabled=false; btn.textContent="Sign In";
+  }
+}
+
+function onLoginSuccess(user){
+  currentUser = user;
+  document.getElementById("login-screen").classList.add("hidden");
+  document.getElementById("user-chip").style.display="flex";
+  document.getElementById("status-chip").style.display="flex";
+  document.getElementById("user-avatar").textContent = user.username[0].toUpperCase();
+  document.getElementById("user-name-label").textContent = user.username;
+  document.getElementById("user-role-label").textContent = user.role;
+  applyRoleUI(user.role);
+  // ─── BOOT: restore session or show login ─────────────────────────────
+(async function boot(){
+  // Enter key on password field triggers login
+  document.getElementById("login-pass").addEventListener("keydown", e=>{
+    if(e.key==="Enter") doLogin();
+  });
+  try {
+    const r = await fetch("/me");
+    const j = await r.json();
+    if(j.status==="ok"){
+      onLoginSuccess(j);   // session still valid – skip login screen
+    }
+    // else: stay on login screen
+  } catch(e){ /* stay on login screen */ }
+})();
+  // If viewer, immediately try to load the latest forecast
+  if(user.role==="viewer") loadLatestForecast();
+}
+
+async function doLogout(){
+  if(!confirm("Sign out?")) return;
+  await fetch("/logout",{method:"POST"});
+  currentUser=null; forecastResult=null;
+  document.getElementById("login-screen").classList.remove("hidden");
+  document.getElementById("user-chip").style.display="none";
+  document.getElementById("status-chip").style.display="none";
+  document.getElementById("login-pass").value="";
+  document.getElementById("login-error").textContent="";
+}
+
+function applyRoleUI(role){
+  // analyst sees all 4 tabs; viewer sees only Forecast Output
+  const analystOnly = ["nav-p1","nav-p2","nav-p4"];
+  analystOnly.forEach(id=>{
+    const el=document.getElementById(id);
+    if(el) el.style.display = role==="analyst" ? "" : "none";
+  });
+  if(role==="viewer"){
+    // force viewer onto p3
+    document.querySelectorAll(".page").forEach(p=>p.classList.remove("active"));
+    document.getElementById("p3").classList.add("active");
+    document.querySelectorAll(".nav-item").forEach(n=>n.classList.remove("active"));
+    document.getElementById("nav-p3").classList.add("active");
+  }
+}
+
+async function loadLatestForecast(){
+  // Viewer polls the server for the latest forecast result
+  try {
+    const r = await fetch("/forecast-result");
+    const j = await r.json();
+    if(j.status==="success"){
+      forecastResult=j;
+      renderPage3();
+      setStatus("ready","forecast loaded ✓");
+    } else if(j.status==="empty"){
+      setStatus("ready","awaiting forecast");
+    }
+  } catch(e){ /* silent */ }
+}
+
+// ─── SERVER HEALTH CHECK (runs after login) ────────────────────────────
+async function initPyodide(){
+  if(currentUser?.role==="viewer"){
+    setStatus("ready","viewer mode");
+    return;
+  }
+  clog("» Checking server…","dim");
+  try {
+    const r=await fetch("/health");
+    const j=await r.json();
+    if(j.status==="ok"){
+      setStatus("ready","server ready ✓");
+      clog("» Server ready ✓  Upload your data file and run.","ok");
+      document.getElementById("run-pipeline-btn").disabled=false;
+      document.getElementById("run-editor-btn").disabled=false;
+    } else throw new Error(j.message);
+  } catch(e){
+    setStatus("error","server offline");
+    clog("✗ "+e.message,"err");
+  }
+}
+
+function setStatus(state,text){
+  const d=document.getElementById("sdot"),s=document.getElementById("stext");
+  d.className="status-dot sd-"+state; s.textContent=text;
+}
+function clog(msg,cls="ok"){
+  const el=document.getElementById("p2-console");
+  if(!el) return;
+  const div=document.createElement("div");
+  div.className="c-"+cls; div.textContent=msg;
+  el.appendChild(div); el.scrollTop=el.scrollHeight;
+}
+
+// ─── NAVIGATION ───────────────────────────────────────────────────────
+function goPage(id,el){
+  // viewers cannot navigate away from p3
+  if(currentUser?.role==="viewer" && id!=="p3") return;
+  document.querySelectorAll(".page").forEach(p=>p.classList.remove("active"));
+  document.querySelectorAll(".nav-item").forEach(n=>n.classList.remove("active"));
+  document.getElementById(id).classList.add("active");
+  el.classList.add("active");
+}
+
+// ─── FILE UPLOAD HELPERS ──────────────────────────────────────────────
+function setupUpload(zoneId,inputId,onLoaded){
+  const zone=document.getElementById(zoneId), inp=document.getElementById(inputId);
+  inp.addEventListener("change",e=>{[...e.target.files].forEach(onLoaded); e.target.value="";});
+  zone.addEventListener("dragover",e=>{e.preventDefault();zone.classList.add("drag");});
+  zone.addEventListener("dragleave",()=>zone.classList.remove("drag"));
+  zone.addEventListener("drop",e=>{e.preventDefault();zone.classList.remove("drag");[...e.dataTransfer.files].forEach(onLoaded);});
+}
+
+function showFileLoaded(nameEl,loadedEl,filename){
+  document.getElementById(nameEl).textContent=filename;
+  document.getElementById(loadedEl).style.display="flex";
+}
+
+// ─────────────────────────────────────────────────────────────────────
+// PAGE 1 – DECOMP FILE UPLOAD & VISUALISATION
+// ─────────────────────────────────────────────────────────────────────
+setupUpload("decomp-upload-zone","decomp-file-input",f=>{
+  showFileLoaded("decomp-file-name","decomp-file-loaded",f.name);
+  document.getElementById("load-decomp-btn").disabled=false;
+  window._decompFile=f;
+});
+
+async function loadDecompFile(){
+  const f=window._decompFile; if(!f) return;
+  const btn=document.getElementById("load-decomp-btn");
+  btn.disabled=true; btn.textContent="Loading…";
+  try {
+    const fd=new FormData();
+    fd.append("file",f);
+    // Reuse /run endpoint to get decomp_data and actual_data from the server
+    clog("» Sending file to server for parsing…","info");
+    // Submit and poll (same pattern as forecast engine)
+    const submitResp=await fetch(SERVER_URL+"/run",{method:"POST",body:fd});
+    if(!submitResp.ok) throw new Error("Server error "+submitResp.status);
+    const submitJson=await submitResp.json();
+    if(submitJson.status!=="started") throw new Error(submitJson.message||"Pipeline error");
+
+    // Poll until done
+    let json=null;
+    while(true){
+      await new Promise(r=>setTimeout(r,4000));
+      const poll=await (await fetch(SERVER_URL+"/job/"+submitJson.job_id)).json();
+      if(poll.status==="running"){ btn.textContent="Loading… "+poll.elapsed_seconds+"s"; continue; }
+      if(poll.status==="error") throw new Error(poll.message||"Pipeline error");
+      if(poll.status==="done"){ json=poll.result; break; }
+    }
+    if(json.status!=="success") throw new Error(json.message||"Pipeline error");
+    // Build records list from decomp_data returned by pipeline
+    const records=(json.decomp_data||[]).map(r=>({
+      channel:r.channel, design_brand:r.design_brand, account:r.account,
+      date:r.date, fy:r.fy, dl1:r["Driver Level 1"]||r.dl1||"",
+      dl2:r["Driver Level 2"]||r.dl2||"", dl3:r.driver_l3||r.dl3||"",
+      volume:r.volume, price:r.price, value:r.value
+    }));
+    // Also stash actual_data for overlay
+    actualData=json.actual_data||[];
+    processDecompData(records);
+    btn.textContent="✓ Loaded";
+    btn.style.background="var(--green)";
+    clog("» Decomp data loaded ✓  "+records.length+" records.","ok");
+  } catch(e){
+    btn.disabled=false; btn.textContent="Load & Visualise";
+    clog("✗ "+e.message,"err");
+    alert("Error loading file: "+e.message);
+  }
+}
+
+function processDecompData(records){
+  decompData=records;
+  // Populate filters
+  const channels=[...new Set(records.map(r=>r.Channel||r.channel))].filter(Boolean);
+  const brands=[...new Set(records.map(r=>r.design_brand))].filter(Boolean);
+  const accounts=[...new Set(records.map(r=>r.Account||r.account))].filter(Boolean);
+  const fys=[...new Set(records.map(r=>r.FY||r.fy))].filter(Boolean).sort();
+  populateSelect("f-channel",channels);
+  populateSelect("f-brand",brands);
+  populateSelect("f-account",accounts);
+  populateSelect("f-fy",fys);
+  document.getElementById("decomp-filters").style.display="flex";
+  document.getElementById("decomp-kpis").style.display="grid";
+  document.getElementById("decomp-charts").style.display="block";
+  document.getElementById("decomp-empty").style.display="none";
+  renderDecomp();
+}
+
+function populateSelect(id,opts){
+  const el=document.getElementById(id);
+  const first=el.options[0]; el.innerHTML=""; el.appendChild(first);
+  opts.forEach(o=>{const op=document.createElement("option");op.value=op.textContent=o;el.appendChild(op);});
+}
+
+function getFilteredDecomp(){
+  const ch=document.getElementById("f-channel").value;
+  const br=document.getElementById("f-brand").value;
+  const ac=document.getElementById("f-account").value;
+  const fy=document.getElementById("f-fy").value;
+  return (decompData||[]).filter(r=>{
+    const rc=r.Channel||r.channel, rb=r.design_brand, ra=r.Account||r.account, rf=r.FY||r.fy;
+    return (ch==="ALL"||rc===ch)&&(br==="ALL"||rb===br)&&(ac==="ALL"||ra===ac)&&(fy==="ALL"||rf===fy);
+  });
+}
+
+function renderDecomp(){
+  const rows=getFilteredDecomp();
+  if(!rows.length) return;
+
+  const baseline=rows.filter(r=>r.dl3==="Baseline");
+  const actual=rows.filter(r=>r.dl3==="Actual Volume");
+  const drivers=rows.filter(r=>!["Baseline","Actual Volume","Predicted Volume"].includes(r.dl3));
+
+  // KPIs
+  const totalActual=actual.reduce((s,r)=>s+(+r.volume||0),0);
+  const totalBaseline=baseline.reduce((s,r)=>s+(+r.volume||0),0);
+  const totalIncr=drivers.reduce((s,r)=>s+(+r.volume||0),0);
+  const bShare=totalActual>0?totalBaseline/totalActual*100:0;
+  document.getElementById("kpi-actual").textContent=fmtNum(totalActual);
+  document.getElementById("kpi-baseline").textContent=fmtNum(totalBaseline);
+  document.getElementById("kpi-incr").textContent=fmtNum(totalIncr);
+  document.getElementById("kpi-bshare").textContent=bShare.toFixed(1)+"%";
+
+  // Get sorted dates from baseline
+  const dates=[...new Set(baseline.map(r=>r.date))].sort();
+  const allDriverTypes=[...new Set(rows.map(r=>r.dl3))].filter(d=>!["Actual Volume","Predicted Volume"].includes(d));
+
+  // Stacked chart datasets
+  const stackedDatasets=allDriverTypes.map((dt,i)=>{
+    const color=DRIVER_COLORS[dt]||DEFAULT_COLORS[i%DEFAULT_COLORS.length];
+    const data=dates.map(d=>{
+      const found=rows.filter(r=>r.dl3===dt&&r.date===d);
+      return found.reduce((s,r)=>s+(+r.volume||0),0);
+    });
+    return {label:dt,data,backgroundColor:color+"CC",borderColor:color,borderWidth:1,stack:"s"};
+  });
+  renderChart("stackedChart","bar",dates,stackedDatasets,{stacked:true,height:280});
+
+  // Actual vs Baseline line chart
+  const actualVols=dates.map(d=>actual.filter(r=>r.date===d).reduce((s,r)=>s+(+r.volume||0),0)||null);
+  const baselineVols=dates.map(d=>baseline.filter(r=>r.date===d).reduce((s,r)=>s+(+r.volume||0),0)||null);
+  renderChart("actualBaseChart","line",dates,[
+    {label:"Actual Volume",data:actualVols,borderColor:"#1D4E89",backgroundColor:"#1D4E8920",borderWidth:2.5,pointRadius:3,tension:0.3,fill:false},
+    {label:"Baseline",data:baselineVols,borderColor:"#2D6A4F",backgroundColor:"#2D6A4F20",borderWidth:2.5,pointRadius:3,tension:0.3,fill:false},
+  ],{height:280});
+
+  // Driver share doughnut
+  const driverTotals={};
+  allDriverTypes.forEach(dt=>{
+    driverTotals[dt]=rows.filter(r=>r.dl3===dt).reduce((s,r)=>s+(+r.volume||0),0);
+  });
+  const dColors=allDriverTypes.map((dt,i)=>DRIVER_COLORS[dt]||DEFAULT_COLORS[i%DEFAULT_COLORS.length]);
+  renderChart("driverShareChart","bar",allDriverTypes,[
+    {label:"Total Volume",data:allDriverTypes.map(d=>Math.max(0,driverTotals[d]||0)),backgroundColor:dColors,borderRadius:4}
+  ],{indexAxis:"y",height:220,legend:false});
+
+  // Driver summary table
+  const totalAbsVol=Object.values(driverTotals).reduce((s,v)=>s+Math.abs(v),0);
+  const tbody=document.getElementById("driver-tbody");
+  tbody.innerHTML=allDriverTypes.sort((a,b)=>Math.abs(driverTotals[b])-Math.abs(driverTotals[a])).map(dt=>{
+    const count=[...new Set(rows.filter(r=>r.dl3===dt).map(r=>r.date))].length;
+    const tot=driverTotals[dt]||0;
+    const dl1=[...new Set(rows.filter(r=>r.dl3===dt).map(r=>r.dl1))].join(", ");
+    const dl2=[...new Set(rows.filter(r=>r.dl3===dt).map(r=>r.dl2))].join(", ");
+    const share=totalAbsVol>0?(Math.abs(tot)/totalAbsVol*100).toFixed(1):0;
+    const color=DRIVER_COLORS[dt]||"#888";
+    return `<tr>
+      <td><span style="display:inline-block;width:8px;height:8px;border-radius:2px;background:${color};margin-right:6px;vertical-align:middle"></span>${dt}</td>
+      <td style="color:var(--text2)">${dl1}</td>
+      <td style="color:var(--text2)">${dl2}</td>
+      <td>${fmtNum(tot)}</td>
+      <td>${count>0?fmtNum(tot/count):"—"}</td>
+      <td><div style="display:flex;align-items:center;gap:8px"><div class="progress-bar" style="width:80px"><div class="progress-fill" style="width:${share}%;background:${color}"></div></div><span>${share}%</span></div></td>
+    </tr>`;
+  }).join("");
+}
+
+// ─────────────────────────────────────────────────────────────────────
+// PAGE 2 – FORECAST ENGINE
+// ─────────────────────────────────────────────────────────────────────
+setupUpload("script-upload-zone","script-file-input",f=>{
+  showFileLoaded("script-file-name","script-file-loaded",f.name);
+  const reader=new FileReader();
+  reader.onload=ev=>{
+    document.getElementById("editor-code").value=ev.target.result;
+    document.getElementById("p2-editor-filename").textContent=f.name;
+    clog("» Script loaded: "+f.name,"info");
+  };
+  reader.readAsText(f);
+});
+
+setupUpload("data-upload-zone","data-file-input",f=>{
+  showFileLoaded("data-file-name","data-file-loaded",f.name);
+  dataFile=f;
+  clog("» Data file ready: "+f.name,"info");
+});
+
+function loadDemoScript(){
+  const script=getDemoScript();
+  document.getElementById("editor-code").value=script;
+  document.getElementById("p2-editor-filename").textContent="dabur_baseline_forecast_v2.py";
+  clog("» Demo script loaded into editor.","info");
+}
+
+async function runForecastPipeline(){
+  if(!dataFile){clog("✗ No data file uploaded.","err");return;}
+
+  const btn=document.getElementById("run-pipeline-btn");
+  const ebtn=document.getElementById("run-editor-btn");
+  btn.disabled=ebtn.disabled=true;
+  btn.textContent="⏳ Submitting…"; ebtn.textContent="⏳";
+  setStatus("loading","submitting job…");
+  document.getElementById("p2-console").innerHTML="";
+  clog("» Uploading file to Genesis AI server…","info");
+  clog(`» ${new Date().toLocaleTimeString()}`,"dim");
+
+  const horizon=parseInt(document.getElementById("cfg-horizon").value)||12;
+  const season=parseInt(document.getElementById("cfg-season").value)||12;
+  const cvh=parseInt(document.getElementById("cfg-cvh").value)||6;
+  const topn=parseInt(document.getElementById("cfg-topn").value)||3;
+
+  try {
+    const fd=new FormData();
+    fd.append("file",dataFile);
+    fd.append("horizon",horizon);
+    fd.append("season",season);
+    fd.append("cv_horizon",cvh);
+    fd.append("top_n",topn);
+
+    // Step 1: submit job – returns immediately with job_id
+    const submitResp=await fetch(SERVER_URL+"/run",{method:"POST",body:fd});
+    const submitJson=await submitResp.json();
+    if(submitJson.status!=="started") throw new Error(submitJson.message||"Failed to start job");
+
+    const jobId=submitJson.job_id;
+    clog("» Job started (id: "+jobId.slice(0,8)+"…)","info");
+    clog("» Pipeline running in background – polling for result…","dim");
+    btn.textContent="⏳ Running…";
+
+    // Step 2: poll /job/<id> every 5 seconds
+    await pollJob(jobId);
+
+  } catch(e){
+    clog("✗ "+e.message,"err");
+    setStatus("error","pipeline error");
+    btn.disabled=ebtn.disabled=false;
+    btn.textContent="▶ Run Baseline Forecast"; ebtn.textContent="▶ Run";
+  }
+}
+
+async function pollJob(jobId){
+  const btn=document.getElementById("run-pipeline-btn");
+  const ebtn=document.getElementById("run-editor-btn");
+  const start=Date.now();
+
+  while(true){
+    await new Promise(r=>setTimeout(r,5000));   // wait 5 seconds between polls
+
+    let poll;
+    try { poll=await (await fetch(SERVER_URL+"/job/"+jobId)).json(); }
+    catch(e){ clog("⚠ Poll error: "+e.message,"warn"); continue; }
+
+    if(poll.status==="running"){
+      const elapsed=poll.elapsed_seconds||Math.round((Date.now()-start)/1000);
+      const mins=Math.floor(elapsed/60), secs=elapsed%60;
+      const t=mins>0?`${mins}m ${secs}s`:`${secs}s`;
+      clog(`» Still running… (${t} elapsed)`,"dim");
+      btn.textContent=`⏳ Running ${t}…`;
+      setStatus("loading",`running ${t}…`);
+      continue;
+    }
+
+    if(poll.status==="error"){
+      clog("✗ Pipeline error: "+poll.message,"err");
+      setStatus("error","pipeline error");
+      break;
+    }
+
+    if(poll.status==="done"){
+      forecastResult=poll.result;
+      clog("» Pipeline complete ✓  "+forecastResult.n_series+" series processed.","ok");
+      clog("» Skipped: "+forecastResult.n_skipped+" series (insufficient data).","dim");
+      clog("» Result saved – viewer users can now see the output.","info");
+      renderPage3();
+      setStatus("ready","forecast complete ✓");
+      break;
+    }
+  }
+
+  btn.disabled=ebtn.disabled=false;
+  btn.textContent="▶ Run Baseline Forecast"; ebtn.textContent="▶ Run";
+}
+
+// ─────────────────────────────────────────────────────────────────────
+// PAGE 3 – FORECAST OUTPUT
+// ─────────────────────────────────────────────────────────────────────
+function renderPage3(){
+  if(!forecastResult) return;
+  document.getElementById("p3-empty").style.display="none";
+  document.getElementById("p3-content").style.display="block";
+
+  const series=forecastResult.series||[];
+  const totalFcstVol=series.reduce((s,sr)=>s+(sr.total_forecast_vol||0),0);
+
+  // KPIs
+  const kpiEl=document.getElementById("p3-kpis");
+  kpiEl.innerHTML=`
+    <div class="metric green"><div class="metric-label">Series Modelled</div><div class="metric-value">${series.length}</div><div class="metric-sub">Channel × Brand × Account</div></div>
+    <div class="metric blue"><div class="metric-label">Total Forecast Vol</div><div class="metric-value">${fmtNum(totalFcstVol)}</div><div class="metric-sub">12-month horizon</div></div>
+    <div class="metric accent"><div class="metric-label">Forecast Period</div><div class="metric-value">${series[0]?.forecast?.[0]?.fy||"—"}</div><div class="metric-sub">to ${series[0]?.forecast?.slice(-1)[0]?.fy||"—"}</div></div>
+    <div class="metric amber"><div class="metric-label">Ensemble Models</div><div class="metric-value">${(series[0]?.top_models||[]).length}</div><div class="metric-sub">top-N by CV MAE</div></div>
+    <div class="metric"><div class="metric-label">Date Run</div><div class="metric-value" style="font-size:16px">${new Date().toLocaleDateString()}</div><div class="metric-sub">${new Date().toLocaleTimeString()}</div></div>
+  `;
+
+  // Series tabs
+  const tabsEl=document.getElementById("p3-series-tabs");
+  tabsEl.innerHTML=series.map((s,i)=>`<div class="stab ${i===0?"active":""}" onclick="selectSeries(${i},this)">${s.design_brand} · ${s.account}</div>`).join("");
+
+  activeSeriesIdx=0;
+  renderSeriesView(0);
+}
+
+function selectSeries(idx,el){
+  document.querySelectorAll(".stab").forEach(t=>t.classList.remove("active"));
+  el.classList.add("active");
+  activeSeriesIdx=idx;
+  renderSeriesView(idx);
+}
+
+function renderSeriesView(idx){
+  const series=(forecastResult?.series||[])[idx];
+  if(!series) return;
+
+  document.getElementById("p3-chart-title").textContent=`Baseline Forecast · ${series.design_brand} × ${series.account}`;
+
+  // Top models chips
+  document.getElementById("p3-top-models").innerHTML=(series.top_models||[]).map(m=>`<span class="model-badge">${m}</span>`).join("");
+
+  // Forecast chart: history + forecast
+  const hist=series.history||[];
+  const fcst=series.forecast||[];
+  const histDates=hist.map(r=>r.date);
+  const fcstDates=fcst.map(r=>r.date);
+  const allDates=[...histDates,...fcstDates];
+  const actualLine=[...hist.map(r=>r.actual_vol),   ...Array(fcstDates.length).fill(null)];
+  const baseLine=  [...hist.map(r=>r.baseline_vol), ...Array(fcstDates.length).fill(null)];
+  const fcstLine=  [...Array(histDates.length).fill(null),...fcst.map(r=>r.volume)];
+  // Connect last historical baseline to first forecast
+  if(hist.length>0 && fcst.length>0){
+    fcstLine[histDates.length-1]=hist[hist.length-1].baseline_vol;
+  }
+
+  renderChart("forecastChart","line",allDates,[
+    {label:"Actual Volume",data:actualLine,borderColor:"#1D4E89",borderWidth:2.5,pointRadius:2,tension:0.3,fill:false,spanGaps:false},
+    {label:"Historical Baseline",data:baseLine,borderColor:"#2D6A4F",borderWidth:2.5,pointRadius:2,tension:0.3,fill:false,spanGaps:false},
+    {label:"Baseline Forecast (12M)",data:fcstLine,borderColor:"#C84B31",borderWidth:2.5,pointRadius:3,tension:0.3,fill:false,spanGaps:false,borderDash:[6,3]},
+  ],{height:280,annotation:hist.length>0?histDates[histDates.length-1]:null});
+
+  // MAE table
+  document.getElementById("mae-tbody").innerHTML=(series.mae_table||[]).map((r,i)=>`
+    <tr>
+      <td>${r.Model}<span style="margin-left:6px">${(series.top_models||[]).includes(r.Model)?'<span class="pill pill-green">ensemble</span>':""}</span></td>
+      <td style="font-family:var(--mono)">${(+r.MAE).toLocaleString(undefined,{maximumFractionDigits:0})}</td>
+      <td><span class="pill ${i===0?"pill-green":i<3?"pill-blue":"pill-amber"}">#${i+1}</span></td>
+    </tr>`).join("");
+
+  // Forecast table
+  document.getElementById("forecast-tbody").innerHTML=fcst.map((r,i)=>{
+    const prev=i>0?fcst[i-1].volume:null;
+    const mom=prev&&prev>0?((r.volume-prev)/prev*100):null;
+    const momStr=mom!==null?`<span class="${mom>=0?"pill pill-green":"pill pill-red"}">${mom>=0?"+":""}${mom.toFixed(1)}%</span>`:"—";
+    return `<tr>
+      <td style="font-family:var(--mono)">${r.date}</td>
+      <td><span class="pill pill-blue">${r.fy}</span></td>
+      <td style="font-family:var(--mono);font-weight:600">${fmtNum(r.volume)}</td>
+      <td style="font-family:var(--mono)">${(+r.price).toLocaleString(undefined,{minimumFractionDigits:2,maximumFractionDigits:2})}</td>
+      <td style="font-family:var(--mono)">₹${fmtNum(r.value)}</td>
+      <td>${momStr}</td>
+    </tr>`;
+  }).join("");
+}
+
+function downloadForecast(){
+  const series=(forecastResult?.series||[])[activeSeriesIdx];
+  if(!series) return;
+  const rows=[["Date","FY","Forecast Volume","Price","Forecast Value","Design Brand","Account","Channel"]];
+  (series.forecast||[]).forEach(r=>rows.push([r.date,r.fy,r.volume,r.price,r.value,series.design_brand,series.account,series.channel]));
+  const csv=rows.map(r=>r.join(",")).join("\n");
+  const a=document.createElement("a"); a.href="data:text/csv;charset=utf-8,"+encodeURIComponent(csv);
+  a.download=`dabur_baseline_forecast_${series.design_brand}_${series.account}.csv`.replace(/\s+/g,"_");
+  a.click();
+}
+
+// ─── CHART HELPER ─────────────────────────────────────────────────────
+function renderChart(id,type,labels,datasets,opts={}){
+  if(chartInstances[id]){chartInstances[id].destroy();}
+  const ctx=document.getElementById(id);
+  if(!ctx) return;
+  ctx.parentElement.style.height=(opts.height||260)+"px";
+  chartInstances[id]=new Chart(ctx,{
+    type,
+    data:{labels,datasets},
+    options:{
+      responsive:true,maintainAspectRatio:false,
+      animation:{duration:500},
+      plugins:{
+        legend:{display:opts.legend!==false,position:"top",labels:{boxWidth:10,font:{size:11,family:"DM Mono"},padding:12}},
+        tooltip:{backgroundColor:"#1A1916",titleFont:{size:11,family:"DM Mono"},bodyFont:{size:11,family:"DM Mono"},padding:10}
+      },
+      scales:{
+        x:{stacked:opts.stacked||false,ticks:{font:{size:10,family:"DM Mono"},maxRotation:45},grid:{display:false}},
+        y:{stacked:opts.stacked||false,ticks:{font:{size:10,family:"DM Mono"},callback:v=>fmtNum(v)},grid:{color:"#F4F3EF"}},
+        ...(opts.indexAxis==="y"?{y:{stacked:false,ticks:{font:{size:10,family:"DM Mono"}}},x:{ticks:{font:{size:10,family:"DM Mono"},callback:v=>fmtNum(v)}}}:{}),
+      },
+      indexAxis:opts.indexAxis||"x",
+    }
+  });
+}
+
+// ─── UTILS ────────────────────────────────────────────────────────────
+function fmtNum(n){
+  const v=+n||0;
+  if(Math.abs(v)>=1e6) return (v/1e6).toFixed(1)+"M";
+  if(Math.abs(v)>=1e3) return (v/1e3).toFixed(1)+"K";
+  return Math.round(v).toLocaleString();
+}
+
+// ─── DEMO SCRIPT (inline copy of v2) ──────────────────────────────────
+function getDemoScript(){
+return `# =============================================================================
+# DABUR BASELINE FORECAST PIPELINE  v2
+# Reads from Excel · outputs JSON to stdout for the web app
+# =============================================================================
+import io, json, warnings, sys
+import numpy as np
+import pandas as pd
+from statsforecast import StatsForecast
+from statsforecast.models import (
+    AutoETS, AutoARIMA, AutoTheta,
+    MSTL, HoltWinters, SeasonalNaive, WindowAverage,
+)
+warnings.filterwarnings("ignore")
+
+CONFIG = {
+    "channel_col":"Channel","category_col":"Category","division_col":"Division",
+    "brand_group_col":"Brand Group","account_col":"Account","key_col":"Key",
+    "design_brand_col":"Design Brand","date_col":"DATE","fy_col":"FY",
+    "volume_col":"Volume","price_col":"Price","value_col":"Value",
+    "driver_col":"Driver Level 3",
+    "baseline_label":"Baseline","actual_label":"Actual Volume",
+    "forecast_label":"Baseline Forecast",
+    "freq":"MS","season_length":12,"forecast_horizon":12,
+    "top_n_ensemble":3,"cv_horizon":6,"cv_windows":2,
+    "sheet_name":"Base Sheet",
+}
+SERIES_KEYS = ["Channel","Design Brand","Account"]
+
+def load_data(source, cfg):
+    if isinstance(source, pd.DataFrame): return source.copy()
+    df = pd.read_excel(source, sheet_name=cfg["sheet_name"])
+    df[cfg["date_col"]] = pd.to_datetime(df[cfg["date_col"]], errors="coerce")
+    df[cfg["date_col"]] = df[cfg["date_col"]].dt.to_period("M").dt.to_timestamp()
+    return df
+
+def get_fy(date):
+    yr = date.year+1 if date.month>=4 else date.year
+    return f"FY{str(yr)[2:]}"
+
+def preprocess_long_to_wide(df, cfg):
+    dc,vc,pc,datec,bl = cfg["driver_col"],cfg["volume_col"],cfg["price_col"],cfg["date_col"],cfg["baseline_label"]
+    df = df[df[dc].isin([bl,cfg["actual_label"]])].copy()
+    dim_cols = [cfg["channel_col"],cfg["category_col"],cfg["division_col"],cfg["brand_group_col"],cfg["account_col"],cfg["key_col"],cfg["design_brand_col"]]
+    pivot = df.pivot_table(index=dim_cols+[datec],columns=dc,values=vc,aggfunc="first").reset_index()
+    pivot.columns.name=None
+    rename={datec:"ds"}
+    if bl in pivot.columns: rename[bl]="baseline"
+    if cfg["actual_label"] in pivot.columns: rename[cfg["actual_label"]]="actual"
+    pivot.rename(columns=rename,inplace=True)
+    price_map = df[df[dc]==bl].set_index(dim_cols+[datec])[pc].rename("price")
+    pivot = pivot.join(price_map,on=dim_cols+["ds"])
+    pivot.sort_values(dim_cols+["ds"],inplace=True); pivot.reset_index(drop=True,inplace=True)
+    pivot["unique_id"] = pivot[SERIES_KEYS].apply(lambda r:" | ".join(r.astype(str)),axis=1)
+    for col in ["baseline","actual"]:
+        if col in pivot.columns: pivot[col]=pivot[col].clip(lower=0)
+    return pivot
+
+def check_gaps(s_df, uid):
+    full=pd.date_range(s_df["ds"].min(),s_df["ds"].max(),freq="MS")
+    missing=full.difference(s_df["ds"])
+    if len(missing):
+        fill=pd.DataFrame({"ds":missing})
+        s_df=pd.concat([s_df,fill],ignore_index=True).sort_values("ds").ffill()
+    return s_df.reset_index(drop=True)
+
+def build_models(sl):
+    return [AutoETS(season_length=sl),AutoARIMA(season_length=sl),AutoTheta(season_length=sl),
+            HoltWinters(season_length=sl,error_type="A",alias="HWAdd"),HoltWinters(season_length=sl,error_type="M",alias="HWMult"),
+            MSTL(season_length=sl,trend_forecaster=AutoTheta(),alias="MSTL_Theta"),SeasonalNaive(season_length=sl)]
+
+def cross_validate_series(sf_df, cfg):
+    sf=StatsForecast(models=build_models(cfg["season_length"]),freq=cfg["freq"],fallback_model=WindowAverage(window_size=6),n_jobs=1)
+    cv=sf.cross_validation(df=sf_df,h=cfg["cv_horizon"],n_windows=cfg["cv_windows"],step_size=3)
+    model_cols=[c for c in cv.columns if c not in ["unique_id","ds","cutoff","y"]]
+    mae={m:np.mean(np.abs(cv[m]-cv["y"])) for m in model_cols}
+    return pd.DataFrame.from_dict(mae,orient="index",columns=["MAE"]).sort_values("MAE").reset_index().rename(columns={"index":"Model"})
+
+def forecast_series(sf_df, top_models, cfg):
+    sf=StatsForecast(models=build_models(cfg["season_length"]),freq=cfg["freq"],fallback_model=WindowAverage(window_size=6),n_jobs=1)
+    fcst=sf.forecast(df=sf_df,h=cfg["forecast_horizon"])
+    if "ds" not in fcst.columns: fcst.reset_index(inplace=True)
+    available=[m for m in top_models if m in fcst.columns]
+    fcst["ensemble_baseline_forecast"]=fcst[available].mean(axis=1).clip(lower=0)
+    return fcst
+
+def build_output_rows(wide_df, fcst_df, dim_meta, cfg):
+    last_price=wide_df["price"].dropna().tail(3).mean()
+    rows=[]
+    for _,row in fcst_df.iterrows():
+        vol=round(row["ensemble_baseline_forecast"],6); price=round(last_price,7); value=round(vol*price,3)
+        rows.append({cfg["channel_col"]:dim_meta[cfg["channel_col"]],cfg["category_col"]:dim_meta[cfg["category_col"]],
+                     cfg["division_col"]:dim_meta[cfg["division_col"]],cfg["brand_group_col"]:dim_meta[cfg["brand_group_col"]],
+                     cfg["account_col"]:dim_meta[cfg["account_col"]],cfg["key_col"]:dim_meta[cfg["key_col"]],
+                     cfg["design_brand_col"]:dim_meta[cfg["design_brand_col"]],cfg["date_col"]:row["ds"].strftime("%Y-%m-%d"),
+                     cfg["fy_col"]:get_fy(row["ds"]),cfg["volume_col"]:vol,cfg["price_col"]:price,cfg["value_col"]:value,
+                     cfg["driver_col"]:cfg["forecast_label"]})
+    return pd.DataFrame(rows)
+
+def run_pipeline(source, cfg=CONFIG):
+    raw_df=load_data(source,cfg)
+    wide_df=preprocess_long_to_wide(raw_df,cfg)
+    series_list=wide_df["unique_id"].unique().tolist()
+    all_forecast_rows=[]; mae_summary=[]; series_results=[]
+    for uid in series_list:
+        s_df=wide_df[wide_df["unique_id"]==uid].copy(); s_df=check_gaps(s_df,uid)
+        sf_input=s_df[["unique_id","ds","baseline"]].rename(columns={"baseline":"y"})
+        mae_df=cross_validate_series(sf_input,cfg)
+        top_models=mae_df["Model"].head(cfg["top_n_ensemble"]).tolist()
+        mae_df["series"]=uid; mae_summary.append(mae_df)
+        fcst_df=forecast_series(sf_input,top_models,cfg)
+        dim_meta=s_df.iloc[0][[cfg["channel_col"],cfg["category_col"],cfg["division_col"],cfg["brand_group_col"],cfg["account_col"],cfg["key_col"],cfg["design_brand_col"]]].to_dict()
+        out_rows=build_output_rows(s_df,fcst_df,dim_meta,cfg)
+        all_forecast_rows.append(out_rows)
+        hist=s_df[["ds","baseline","actual","price"]].copy()
+        hist["ds"]=hist["ds"].dt.strftime("%Y-%m-%d")
+        series_results.append({
+            "uid":uid,"channel":dim_meta[cfg["channel_col"]],"design_brand":dim_meta[cfg["design_brand_col"]],
+            "account":dim_meta[cfg["account_col"]],"top_models":top_models,
+            "mae_table":mae_df[["Model","MAE"]].round(2).to_dict(orient="records"),
+            "history":hist.rename(columns={"ds":"date","baseline":"baseline_vol","actual":"actual_vol","price":"price"}).to_dict(orient="records"),
+            "forecast":[{"date":r[cfg["date_col"]],"fy":r[cfg["fy_col"]],"volume":round(r[cfg["volume_col"]],2),"price":round(r[cfg["price_col"]],4),"value":round(r[cfg["value_col"]],0)} for _,r in out_rows.iterrows()],
+            "total_forecast_vol":round(out_rows[cfg["volume_col"]].sum(),2),
+        })
+    decomp_drivers=raw_df[~raw_df[cfg["driver_col"]].isin([cfg["actual_label"],"Predicted Volume"])].copy()
+    decomp_drivers["DATE_str"]=decomp_drivers[cfg["date_col"]].dt.strftime("%Y-%m-%d")
+    decomp_records=decomp_drivers[[cfg["channel_col"],cfg["design_brand_col"],cfg["account_col"],"DATE_str",cfg["fy_col"],"Driver Level 1","Driver Level 2",cfg["driver_col"],cfg["volume_col"],cfg["price_col"],cfg["value_col"]]].rename(columns={cfg["channel_col"]:"channel",cfg["design_brand_col"]:"design_brand",cfg["account_col"]:"account","DATE_str":"date",cfg["fy_col"]:"fy",cfg["driver_col"]:"driver_l3",cfg["volume_col"]:"volume",cfg["price_col"]:"price",cfg["value_col"]:"value"}).to_dict(orient="records")
+    actual_records=raw_df[raw_df[cfg["driver_col"]]==cfg["actual_label"]].copy()
+    actual_records["DATE_str"]=actual_records[cfg["date_col"]].dt.strftime("%Y-%m-%d")
+    actual_out=actual_records[[cfg["design_brand_col"],cfg["account_col"],"DATE_str",cfg["volume_col"]]].rename(columns={cfg["design_brand_col"]:"design_brand",cfg["account_col"]:"account","DATE_str":"date",cfg["volume_col"]:"actual_volume"}).to_dict(orient="records")
+    output={"status":"success","n_series":len(series_list),"series_ids":series_list,"series":series_results,"decomp_data":decomp_records,"actual_data":actual_out}
+    print(json.dumps(output))
+    return output
+
+if __name__=="__main__":
+    src=sys.argv[1] if len(sys.argv)>1 else None
+    if src: run_pipeline(src)
+`;
+}
+
+// ─── BOOT ─────────────────────────────────────────────────────────────
+initPyodide();
+</script>
+</body>
+</html>

requirements.txt

@@ -0,0 +1,6 @@
+flask>=3.0.0
+gunicorn>=21.2.0
+numpy>=1.26.0
+pandas>=2.2.0
+openpyxl>=3.1.2
+statsforecast>=1.7.0