Initial commit: MyMidas personal finance tracker

Full-stack self-hosted finance app with FastAPI backend and React frontend.

Features:
- Accounts, transactions, budgets, investments with GBP base currency
- CSV import with auto-detection for 10 UK bank formats
- ML predictions: spending forecast, net worth projection, Monte Carlo
- 7 selectable themes (Obsidian, Arctic, Midnight, Vault, Terminal, Synthwave, Ledger)
- Receipt/document attachments on transactions (JPEG, PNG, WebP, PDF)
- AES-256-GCM field encryption, RS256 JWT, TOTP 2FA, RLS, audit log
- Encrypted nightly backups + key rotation script
- Mobile-responsive layout with bottom nav

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/app/ml/spending_forecast.py

@@ -0,0 +1,91 @@
+from __future__ import annotations
+
+import warnings
+from datetime import date
+from dateutil.relativedelta import relativedelta
+
+import numpy as np
+import pandas as pd
+
+warnings.filterwarnings("ignore")
+
+MIN_POINTS = 3
+FORECAST_MONTHS = 3
+
+
+def _next_month_starts(from_date: date, n: int) -> list[str]:
+    months = []
+    d = (from_date.replace(day=1) + relativedelta(months=1))
+    for _ in range(n):
+        months.append(d.strftime("%Y-%m-%d"))
+        d += relativedelta(months=1)
+    return months
+
+
+def _fit_holt(values: list[float], n: int) -> tuple[list[float], list[float], list[float]]:
+    from statsmodels.tsa.holtwinters import ExponentialSmoothing
+
+    try:
+        if len(values) >= 12:
+            model = ExponentialSmoothing(values, trend="add", seasonal="add", seasonal_periods=12)
+        elif len(values) >= 4:
+            model = ExponentialSmoothing(values, trend="add", seasonal=None)
+        else:
+            model = ExponentialSmoothing(values, trend=None, seasonal=None)
+
+        fit = model.fit(optimized=True, disp=False)
+        forecast = fit.forecast(n)
+        sigma = float(np.std(fit.resid)) if len(fit.resid) > 1 else float(np.mean(values) * 0.15)
+        lower = np.maximum(0, forecast - 1.28 * sigma)
+        upper = forecast + 1.28 * sigma
+        return forecast.tolist(), lower.tolist(), upper.tolist()
+    except Exception:
+        avg = float(np.mean(values))
+        sigma = float(np.std(values)) if len(values) > 1 else avg * 0.15
+        return [avg] * n, [max(0, avg - 1.28 * sigma)] * n, [(avg + 1.28 * sigma)] * n
+
+
+def forecast_spending(df: pd.DataFrame) -> list[dict]:
+    """
+    df columns: category_id, category_name, ds (monthly), y (amount)
+    Returns list of category forecast dicts.
+    """
+    if df.empty:
+        return []
+
+    today = date.today()
+    future_dates = _next_month_starts(today, FORECAST_MONTHS)
+    results = []
+
+    for (cat_id, cat_name), group in df.groupby(["category_id", "category_name"]):
+        group = group.sort_values("ds")
+        values = group["y"].tolist()
+        actuals = [
+            {"date": row["ds"].strftime("%Y-%m-%d"), "amount": row["y"]}
+            for _, row in group.iterrows()
+        ]
+
+        if len(values) < MIN_POINTS:
+            avg = float(np.mean(values))
+            forecast_pts = [
+                {"date": d, "amount": round(avg, 2), "lower": round(avg * 0.7, 2), "upper": round(avg * 1.3, 2)}
+                for d in future_dates
+            ]
+        else:
+            fcast, lower, upper = _fit_holt(values, FORECAST_MONTHS)
+            forecast_pts = [
+                {"date": d, "amount": round(max(0, f), 2), "lower": round(l, 2), "upper": round(u, 2)}
+                for d, f, l, u in zip(future_dates, fcast, lower, upper)
+            ]
+
+        results.append({
+            "category_id": cat_id,
+            "category_name": cat_name,
+            "monthly_avg": round(float(np.mean(values)), 2),
+            "actuals": actuals[-6:],  # last 6 months for display
+            "forecast": forecast_pts,
+        })
+
+    # Sort by monthly_avg descending (highest spend first)
+    results.sort(key=lambda x: x["monthly_avg"], reverse=True)
+    return results