app.py

# streamlit run app.py
#pip install streamlit pandas openai
#pip install scikit-learn


import os
import streamlit as st
import pandas as pd
from collections import defaultdict

# Optional AI (OpenAI)
try:
    import openai
    OPENAI_AVAILABLE = True
except:
    OPENAI_AVAILABLE = False

# -------------------------
# Page settings
# -------------------------
st.set_page_config(page_title="財務シミュレーター", layout="wide")


TITLE = "財務シミュレーター"

# -------------------------
# CSS / styling
# -------------------------
st.markdown("""
<style>
body { background: linear-gradient(135deg, #E8F0FF 0%, #FFFFFF 40%, #F9ECFF 100%); }

.app-title { color: #123A8A; font-size:28px; font-weight:700; margin-bottom:6px; }

.card {
    background: linear-gradient(180deg, rgba(255,255,255,0.95), rgba(250,250,250,0.95));
    padding: 18px; border-radius:12px;
    box-shadow: 0 6px 18px rgba(15,23,42,0.08);
    margin-bottom:18px;
}

.section-bar { width:100%; height:6px; background:#3B82F6; border-radius:4px; margin:18px 0 32px 0; }

.bs-table { width:100%; border-collapse: collapse; margin-top:10px; font-size:14px; }
.bs-table th { background:#e8f2ff; color:#0b3a66; padding:8px; border:1px solid #9aa6b2; text-align:left; }
.bs-table td { padding:8px; border:1px solid #c8d2db; }
.bs-total { background:#eef5ff; font-weight:700; }
.bs-section { background:#f3f7fb; font-weight:700; }

.section-title { margin-top:0px; font-size:22px; font-weight:600; color:#123A8A; }
.card {
    margin-top:0px !important;
}

input[type=number] { background-color:#ffffff !important; color:#000000 !important; }
.muted { color:#6b7280; font-size:13px; }
</style>
""", unsafe_allow_html=True)

st.markdown("""
<style>
.card:empty {
    display: none !important;
    visibility: hidden !important;
    height: 0 !important;
    padding: 0 !important;
    margin: 0 !important;
}
</style>
""", unsafe_allow_html=True)


# -------------------------
# Account master
# -------------------------
ACCOUNT_TYPES = {
    # 資産
    "現金": "資産", "預金": "資産", "売掛金": "資産", "棚卸資産": "資産",
    "有形固定資産": "資産", "無形資産": "資産", "前払費用": "資産", "投資有価証券": "資産",
    # 負債
    "買掛金": "負債", "短期借入金": "負債", "長期借入金": "負債", "未払金": "負債",
    "未払費用": "負債", "前受金": "負債",
    # 純資産
    "資本金": "純資産", "利益剰余金": "純資産", "資本剰余金": "純資産",
    # 収益
    "売上": "収益", "受取利息": "収益",
    # 費用
    "仕入": "費用", "給与": "費用", "旅費交通費": "費用", "通信費": "費用", "減価償却費": "費用",
}

# -------------------------
# Helper functions
# -------------------------
def create_initial_balances():
    return defaultdict(float)

def apply_journal_entries(balances, journal_entries):
    b = balances.copy()
    for j in journal_entries:
        d = j["借方"]; c = j["貸方"]; amt = j["金額"]
        b[d] += amt
        b[c] -= amt
    return b

def create_balance_sheet(balances):
    bs = {"資産": {}, "負債": {}, "純資産": {}}
    for acct, val in balances.items():
        typ = ACCOUNT_TYPES.get(acct)
        if typ in bs:
            bs[typ][acct] = val
    return bs

def create_profit_loss(balances):
    revenue = {}; expenses = {}
    for acct, val in balances.items():
        typ = ACCOUNT_TYPES.get(acct)
        if typ == "収益":
            revenue[acct] = abs(val)
        elif typ == "費用":
            expenses[acct] = abs(val)
        
    profit = sum(revenue.values()) - sum(expenses.values())
    return {"収益": revenue, "費用": expenses, "利益": profit}

def bs_totals(bs):
    totals = {}
    for k, v in bs.items():
        totals[k] = sum(v.values()) if isinstance(v, dict) else 0
    return totals

def format_yen(x):
    return f"{x:,.0f}"

# -------------------------
# UI: Header
# -------------------------
st.markdown(f"<div class='app-title'>{TITLE}</div>", unsafe_allow_html=True)
st.markdown("<div class='muted'>予想される仕訳を入力すると、貸借対照表（B/S）・損益計算書（P/L）が更新されます。AIは仕訳説明・簡易分析に使えます（APIキー任意）。</div>", unsafe_allow_html=True)

# -------------------------
# UI: B/S input
# -------------------------
st.markdown("<div class='card'>", unsafe_allow_html=True)
st.markdown("<div class='section-title'>貸借対照表（B/S）入力</div>", unsafe_allow_html=True)

col_l, col_r = st.columns([1,1])
with col_l:
    st.subheader("資産の部（現在・百万円単位）")
    cash = st.number_input("現金", value=0, step=1, key="in_cash")
    deposit = st.number_input("預金", value=0, step=1, key="in_deposit")
    ar = st.number_input("売掛金", value=0, step=1, key="in_ar")
    inventory = st.number_input("棚卸資産", value=0, step=1, key="in_inventory")
    ppe = st.number_input("有形固定資産",value=0, step=1, key="in_ppe")
    intangible = st.number_input("無形資産", value=0, step=1, key="in_intangible")
    prepay = st.number_input("前払費用", value=0, step=1, key="in_prepay")
    invest = st.number_input("投資有価証券", value=0, step=1, key="in_invest")
with col_r:
    st.subheader("負債・純資産の部")
    ap = st.number_input("買掛金", value=0, step=1, key="in_ap")
    short_loan = st.number_input("短期借入金", value=0, step=1, key="in_shortloan")
    long_loan = st.number_input("長期借入金", value=0, step=1, key="in_longloan")
    unpaid = st.number_input("未払金", value=0, step=1, key="in_unpaid")
    capital = st.number_input("資本金", value=0, step=1, key="in_capital")
    retained = st.number_input("利益剰余金", value=0, step=1, key="in_retained")
    capital_surplus = st.number_input("資本剰余金", value=0, step=1, key="in_capital_surplus")

st.markdown("</div>", unsafe_allow_html=True)

st.markdown("<div class='card'>", unsafe_allow_html=True)
st.markdown("<div class='section-title'>損益計算書（P/L）入力</div>", unsafe_allow_html=True)

col_l, col_r = st.columns([1, 1])

with st.container():
    st.subheader("損益計算書（P/L）（現在・百万円単位）")

    col1, col2 = st.columns(2)

    with col1:
        sales = st.number_input("売上高", value=0, step=1, key="pl_sales")
        cogs = st.number_input("売上原価", value=0, step=1, key="pl_cogs")
        sga = st.number_input("販管費", value=0, step=1, key="pl_sga")

    with col2:
        non_op_income = st.number_input("営業外収益", value=0, step=1, key="pl_non_op_income")
        non_op_expense = st.number_input("営業外費用", value=0, step=1, key="pl_non_op_expense")
        extraordinary_gain = st.number_input("特別利益", value=0, step=1, key="pl_extra_gain")
        extraordinary_loss = st.number_input("特別損失", value=0, step=1, key="pl_extra_loss")

gross_profit = sales - cogs
operating_profit = gross_profit - sga
ordinary_profit = operating_profit + non_op_income - non_op_expense
profit_before_tax = ordinary_profit + extraordinary_gain - extraordinary_loss
net_income = profit_before_tax   # 今は税金を簡略化






# -------------------------
# UI: Journal Entries
# -------------------------
st.markdown("<div class='card'>", unsafe_allow_html=True)
st.markdown("<div class='section-title'>仕訳（Journal Entries）</div>", unsafe_allow_html=True)
st.markdown("<div class='muted'>借方・貸方・金額を入力して追加。複数行対応。</div>", unsafe_allow_html=True)

if "journal_entries" not in st.session_state:
    st.session_state.journal_entries = []

cols = st.columns([2,2,1,1])
with cols[0]:
    sel_debit = st.selectbox("借方科目", options=list(ACCOUNT_TYPES.keys()), key="sel_debit")
with cols[1]:
    sel_credit = st.selectbox("貸方科目", options=list(ACCOUNT_TYPES.keys()), key="sel_credit")
with cols[2]:
    sel_amount = st.number_input("金額", min_value=0, value=0, step=1, key="je_amount")
with cols[3]:
    if st.button("追加"):
        if sel_amount <= 0:
            st.warning("金額は0より大きくしてください。")
        else:
            st.session_state.journal_entries.append({"借方": sel_debit, "貸方": sel_credit, "金額": float(sel_amount)})
            st.success("仕訳を追加しました。")

st.markdown("<div class='muted'>現在の仕訳一覧</div>", unsafe_allow_html=True)
if st.session_state.journal_entries:
    st.table(pd.DataFrame(st.session_state.journal_entries))
    if st.button("仕訳を全て消す"):
        st.session_state.journal_entries = []
else:
    st.info("仕訳はまだ追加されていません。")

st.markdown("</div>", unsafe_allow_html=True)

# -------------------------
# Build balances dict
# -------------------------
initial_balances = create_initial_balances()
initial_balances["現金"] = cash
initial_balances["預金"] = deposit
initial_balances["売掛金"] = ar
initial_balances["棚卸資産"] = inventory
initial_balances["有形固定資産"] = ppe
initial_balances["無形資産"] = intangible
initial_balances["前払費用"] = prepay
initial_balances["投資有価証券"] = invest
initial_balances["買掛金"] = ap
initial_balances["短期借入金"] = short_loan
initial_balances["長期借入金"] = long_loan
initial_balances["未払金"] = unpaid
initial_balances["資本金"] = capital
initial_balances["利益剰余金"] = retained
initial_balances["資本剰余金"] = capital_surplus

# -------------------------
# UI: Statements
# -------------------------
st.markdown("<div class='card'>", unsafe_allow_html=True)
st.markdown("<div class='section-title'>B/S、P/Lを予想</div>", unsafe_allow_html=True)

total_debit = sum(j["金額"] for j in st.session_state.journal_entries) if st.session_state.journal_entries else 0.0
st.markdown(f"<div class='muted'>仕訳合計（借方）: {format_yen(total_debit)}</div>", unsafe_allow_html=True)

if st.button("予想"):
    after_balances = apply_journal_entries(initial_balances, st.session_state.journal_entries)
    bs_after = create_balance_sheet(after_balances)

    totals = bs_totals(bs_after)
    assets_total = totals.get("資産",0)
    liabilities_total = totals.get("負債",0)
    equity_total = totals.get("純資産",0)
    right_total = liabilities_total + equity_total

    # 借方・貸方チェック
    if assets_total != right_total:
        st.warning(f"⚠️ 資産合計 ({format_yen(assets_total)}) と 負債・純資産合計 ({format_yen(right_total)}) が一致していません。")


    # --- B/S
    st.markdown("<h3>貸借対照表</h3>", unsafe_allow_html=True)
    st.markdown("<div style='overflow-x:auto'>", unsafe_allow_html=True)
    bs_html = "<table class='bs-table'>"
    for section in ["資産","負債","純資産"]:
        bs_html += f"<tr><th colspan='2'>{section}の部</th></tr>"
        for k,v in bs_after[section].items():
            bs_html += f"<tr><td>{k}</td><td style='text-align:right'>{format_yen(v)}</td></tr>"
        bs_html += f"<tr class='bs-total'><td>【{section}合計】</td><td style='text-align:right'>{format_yen(totals[section])}</td></tr>"
    bs_html += f"<tr class='bs-total' style='background:#dfe8ff'><td>【負債・純資産合計】</td><td style='text-align:right'>{format_yen(right_total)}</td></tr>"
    bs_html += "</table>"
    st.markdown(bs_html, unsafe_allow_html=True)
    st.markdown("</div>", unsafe_allow_html=True)

     # --- P/L
    st.markdown("<h3>損益計算書（P/L）</h3>", unsafe_allow_html=True)

    pl_after = pl_df.copy()

    
    # 仕訳で反映された収益／費用を優先して P/L を組み立てる
    sales_val = pl_after["収益"].get("売上", 0)
    cogs_val = pl_after["費用"].get("仕入", 0)
    sga_val = pl_after["費用"].get("販管費", 0)
    non_op_income_val = pl_after["収益"].get("営業外収益", 0)
    non_op_expense_val = pl_after["費用"].get("営業外費用", 0)
    extra_gain_val = pl_after["収益"].get("特別利益", 0)
    extra_loss_val = pl_after["費用"].get("特別損失", 0)

    gross_profit = sales_val - cogs_val
    operating_profit = gross_profit - sga_val
    ordinary_profit = operating_profit + non_op_income_val - non_op_expense_val
    profit_before_tax = ordinary_profit + extra_gain_val - extra_loss_val
    net_income_val = profit_before_tax  # 税考慮は省略

    pl_df = pd.DataFrame({
        "項目": [
            "売上高",
            "売上原価",
            "売上総利益",
            "販管費",
            "営業利益",
            "営業外収益",
            "営業外費用",
            "経常利益",
            "特別利益",
            "特別損失",
            "税引前当期純利益",
            "当期純利益"
        ],
        "金額": [
            sales_val,
            cogs_val,
            gross_profit,
            sga_val,
            operating_profit,
            non_op_income_val,
            non_op_expense_val,
            ordinary_profit,
            extra_gain_val,
            extra_loss_val,
            profit_before_tax,
            net_income_val
        ]
    })

    # 表示を整数にそろえる：ここで必ず pl_df が存在するので安全
    pl_df["金額"] = pl_df["金額"].astype(int)

    st.table(pl_df)

# -------------------------
# 回帰予測と異常値検知（修正版）
# -------------------------
import numpy as np
import warnings
warnings.filterwarnings("ignore", category=UserWarning)

# sklearn が無ければ親切に教える
try:
    from sklearn.ensemble import IsolationForest
    SKL_AVAILABLE = True
except Exception:
    SKL_AVAILABLE = False

# 1. 履歴 DataFrame 作成
if st.session_state.get("journal_entries"):
    df_journal = pd.DataFrame(st.session_state.journal_entries)
    # 安全化：列が存在するか、型変換
    if "金額" in df_journal.columns:
        df_journal["金額"] = pd.to_numeric(df_journal["金額"], errors="coerce").fillna(0.0)
    else:
        df_journal["金額"] = 0.0
else:
    df_journal = pd.DataFrame(columns=["借方", "貸方", "金額"])

# 2. 科目別の集計（借方はプラス、貸方はマイナス）
df_summary = pd.DataFrame(index=list(ACCOUNT_TYPES.keys()))
df_summary["金額"] = 0.0
for _, row in df_journal.iterrows():
    d = row.get("借方")
    c = row.get("貸方")
    amt = float(row.get("金額", 0.0))
    if d in df_summary.index:
        df_summary.loc[d, "金額"] += amt
    if c in df_summary.index:
        df_summary.loc[c, "金額"] -= amt

# 3. 回帰予測（簡易：ここでは直近値を使う。時系列が整えば本格回帰へ差し替え）
predictions = {}
for acct in df_summary.index:
    # ここでは履歴があまりないため、現状値をそのまま予測値とする
    val = df_summary.loc[acct, "金額"]
    predictions[acct] = float(val)

# 4. 異常値検知（IsolationForest） — sklearn が無ければスキップして通知
df_summary["異常"] = 0  # デフォルト 0（正常）
if SKL_AVAILABLE and len(df_summary) > 0:
    X = df_summary["金額"].values.reshape(-1, 1)
    # variance check: 常に同じ値だと IsolationForest は不安定なのでガード
    if np.nanstd(X) > 0:
        iso = IsolationForest(contamination=0.05, random_state=42)
        try:
            iso.fit(X)
            out = iso.predict(X)  # 1: normal, -1: outlier
            df_summary["異常"] = out
        except Exception:
            # 万が一失敗したらすべて正常扱いにする
            df_summary["異常"] = 1
    else:
        # 分散がゼロ（すべて同値）の場合は異常検知しない
        df_summary["異常"] = 1
else:
    if not SKL_AVAILABLE:
        st.info("※ 異常値検知には scikit-learn が必要です（pip install scikit-learn）。現在はスキップしています。")

# 5. 予測 P/L 作成（表示用 DataFrame を必ず作る）
pl_pred_items = ["売上", "受取利息", "仕入", "給与", "旅費交通費", "通信費", "減価償却費"]
pl_pred_values = [ predictions.get(x, 0.0) for x in pl_pred_items ]

pred_pl_df = pd.DataFrame({
    "項目": pl_pred_items,
    "予測金額": pl_pred_values
})

# 表示形式を整数に揃える（画面上は 0 表示になる）
# 浮動小数の負荷が無ければ int cast、負の値も扱うため int() で丸める
pred_pl_df["予測金額"] = pred_pl_df["予測金額"].apply(lambda x: int(round(x)))

# 6. 結果表示（カード内）
st.markdown("<div class='card'>", unsafe_allow_html=True)
st.markdown("<div class='section-title'>予測 P/L & 異常値</div>", unsafe_allow_html=True)

st.markdown("<div style='margin-bottom:8px;'><strong>予測 P/L（簡易）</strong></div>", unsafe_allow_html=True)
st.table(pred_pl_df)

st.markdown("<div style='margin-top:10px; margin-bottom:6px;'><strong>異常値チェック（1=正常, -1=外れ値）</strong></div>", unsafe_allow_html=True)
# 金額列を整数にして表示（df_summary は index が科目）
display_df = df_summary.reset_index().rename(columns={"index": "科目"})
display_df["金額"] = display_df["金額"].apply(lambda x: int(round(x)))
st.table(display_df[["科目", "金額", "異常"]])

st.markdown("</div>", unsafe_allow_html=True)