File size: 9,989 Bytes
83efdfc 2cc39e6 83efdfc 2cc39e6 83efdfc 2cc39e6 83efdfc 2cc39e6 83efdfc 5dd7efe 83efdfc 2cc39e6 83efdfc 5dd7efe 83efdfc | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 | # context/context_builder.py
"""
LLM ์
๋ ฅ์ฉ ์ปจํ
์คํธ ๋น๋ ํจ์ ๋ชจ์
"""
def _fmt_krw(val):
return "N/A" if val is None else f"{val:.2f}"
def _fmt_num(val):
if val is None or val != val:
return "N/A"
if isinstance(val, int):
return f"{val:,}"
if isinstance(val, float):
return f"{val:,.2f}"
return str(val)
def _pct(val):
if val is None or val != val:
return "N/A"
return f"{float(val) * 100:.1f}%"
def _upside(current, target):
if not current or not target:
return "N/A"
return f"{(float(target) / float(current) - 1) * 100:.1f}%"
def build_company_section(fd, ticker):
location = " / ".join(x for x in [fd.get("city"), fd.get("country")] if x)
return f"""[๊ธฐ์
์ ๋ณด]\nํ์ฌ๋ช
: {fd.get('company_name', ticker)} ({fd.get('symbol', ticker)})\n์นํฐ: {fd.get('sector', 'N/A')} / {fd.get('industry', 'N/A')}\n๊ฑฐ๋์: {fd.get('exchange', 'N/A')} / ํตํ: {fd.get('currency', 'N/A')}\n๋ณธ์ฌ: {location or 'N/A'} / ์ง์ ์: {_fmt_num(fd.get('full_time_employees'))}\n์๊ฐ์ด์ก: ${fd.get('market_cap_b', 'N/A')}B / EV: ${fd.get('enterprise_value_b', 'N/A')}B\n์ฌ์
์ค๋ช
: {fd.get('description', 'N/A')}"""
def build_price_section(pd):
return f"""[์ฃผ๊ฐ ํผํฌ๋จผ์ค]\nํ์ฌ๊ฐ: ${pd.get('current_price', 'N/A')}\n1์ฃผ ์์ต๋ฅ : {pd.get('change_1w_pct', 'N/A')}% / 1๊ฐ์: {pd.get('change_1m_pct', 'N/A')}% / YTD: {pd.get('change_ytd_pct', 'N/A')}%\n52์ฃผ ๊ณ ๊ฐ: ${pd.get('52w_high', 'N/A')} / ์ ๊ฐ: ${pd.get('52w_low', 'N/A')}\n30์ผ ํ๊ท ๊ฑฐ๋๋: {pd.get('avg_volume_30d', 'N/A'):,}"""
def build_valuation_section(fd):
return f"""[๋ฐธ๋ฅ์์ด์
& ์ฌ๋ฌด]\nPER(TTM): {fd.get('pe_ratio', 'N/A')} / ์ ํ PER: {fd.get('forward_pe', 'N/A')}\nPBR: {fd.get('pb_ratio', 'N/A')} / PSR: {fd.get('ps_ratio', 'N/A')} / EV/EBITDA: {fd.get('ev_to_ebitda', 'N/A')}\nEPS(TTM): {fd.get('trailing_eps', 'N/A')} / ์ ํ EPS: {fd.get('forward_eps', 'N/A')}\nROE: {_pct(fd.get('roe'))} / ROA: {_pct(fd.get('roa'))}\n๋งค์ถ์ด์ด์ต๋ฅ : {_pct(fd.get('gross_margin'))} / ์์
์ด์ต๋ฅ : {_pct(fd.get('operating_margin'))} / EBITDA ๋ง์ง: {_pct(fd.get('ebitda_margin'))}\n์์ด์ต๋ฅ : {_pct(fd.get('profit_margin'))}\n๋งค์ถ ์ฑ์ฅ๋ฅ : {_pct(fd.get('revenue_growth'))} / ์ด์ต ์ฑ์ฅ๋ฅ : {_pct(fd.get('earnings_growth'))}\n๋ถ์ฑ๋น์จ: {fd.get('debt_to_equity', 'N/A')} / ์ ๋๋น์จ: {fd.get('current_ratio', 'N/A')} / ๋น์ข๋น์จ: {fd.get('quick_ratio', 'N/A')}\n์ดํ๊ธ: ${fd.get('total_cash_b', 'N/A')}B / ์ด๋ถ์ฑ: ${fd.get('total_debt_b', 'N/A')}B\n์์
ํ๊ธํ๋ฆ: ${fd.get('operating_cashflow_b', 'N/A')}B / ์์ฌํ๊ธํ๋ฆ: ${fd.get('free_cashflow_b', 'N/A')}B\n๋ฐฐ๋น์์ต๋ฅ : {_pct(fd.get('dividend_yield'))} / ๋ฐฐ๋น์ฑํฅ: {_pct(fd.get('payout_ratio'))} / ๋ฒ ํ: {fd.get('beta', 'N/A')}"""
def build_analyst_section(fd, pd):
rec = fd.get("recommendation", "")
return f"""[์ ๋๋ฆฌ์คํธ ์ปจ์ผ์์ค]\nํ๊ท ๋ชฉํ์ฃผ๊ฐ: ${fd.get('analyst_target', 'N/A')} / ์๋จ: ${fd.get('target_high_price', 'N/A')} / ํ๋จ: ${fd.get('target_low_price', 'N/A')}\nํฌ์์๊ฒฌ: {rec.upper() if rec else 'N/A'} / ์ปค๋ฒ ์ ๋๋ฆฌ์คํธ ์: {_fmt_num(fd.get('analyst_opinion_count'))}\nํ์ฌ๊ฐ ๋๋น ์์น์ฌ๋ ฅ: {_upside(pd.get('current_price'), fd.get('analyst_target'))}"""
def build_capital_structure_section(fd):
return f"""[์๊ธ & ์๋ณธ๊ตฌ์กฐ]\n๋ฐํ์ฃผ์์: {_fmt_num(fd.get('shares_outstanding_b'))}B์ฃผ / ์ ํต์ฃผ์์: {_fmt_num(fd.get('float_shares_b'))}B์ฃผ\n๋ด๋ถ์ ๋ณด์ ์จ: {_pct(fd.get('held_percent_insiders'))} / ๊ธฐ๊ด ๋ณด์ ์จ: {_pct(fd.get('held_percent_institutions'))}\n๊ณต๋งค๋ ๋น์จ(Short Ratio): {fd.get('short_ratio', 'N/A')} / ์ ํต์ฃผ์ ๋๋น ๊ณต๋งค๋: {_pct(fd.get('short_percent_float'))}\n์ต๊ทผ ๋ฐฐ๋น๊ธ: {fd.get('dividend_rate', 'N/A')} / ๋ฐฐ๋น๋ฝ์ผ: {fd.get('ex_dividend_date', 'N/A')}"""
def build_technicals_section(td):
return f"""[๊ธฐ์ ์ ์งํ]\nMA20: ${td.get('ma20', 'N/A')} (ํ์ฌ๊ฐ MA20 {td.get('price_vs_ma20', 'N/A')})\nMA50: ${td.get('ma50', 'N/A')} / MA200: ${td.get('ma200', 'N/A')}\nRSI(14): {td.get('rsi_14', 'N/A')} โ {td.get('rsi_signal', 'N/A')}\nMACD ํ์คํ ๊ทธ๋จ: {td.get('macd_histogram', 'N/A')} โ {td.get('macd_signal', 'N/A')}\n๋ณผ๋ฆฐ์ ๋ฐด๋ ์์น: {td.get('bb_position', 'N/A')}\n๊ฑฐ๋๋ ๋น์จ(vs 20์ผ ํ๊ท ): {td.get('volume_ratio', 'N/A')}x"""
def build_earnings_section(ed, intent):
lines = ["[์ค์ ๋ฐ์ดํฐ]"]
if ed.get("next_earnings_date"):
lines.append(f"๋ค์ ์ค์ ๋ฐํ ์์ ์ผ: {ed['next_earnings_date']}")
filtered = ed.get("filtered_quarter")
if filtered and filtered.get("found") and filtered.get("data"):
lines.extend(_format_filtered_earnings(filtered))
elif filtered and not filtered.get("found"):
lines.append(f"\nโป {filtered['period']} ์ค์ ๋ฐ์ดํฐ๋ฅผ ์ฐพ์ ์ ์์ต๋๋ค.")
lines.extend(_format_quarterly_table(ed.get("quarterly_results", [])))
lines.extend(_format_annual_table(ed.get("annual_results", [])))
lines.extend(_format_eps_surprise(ed.get("earnings_surprise", [])))
return "\n".join(lines)
def _format_filtered_earnings(filtered):
lines = [f"\nโ {filtered['period']} ์ค์ (์กฐํ ๋์)"]
d = filtered["data"]
if filtered["type"] == "quarter":
lines += [
f" ๋งค์ถ: {_fmt_krw(d.get('revenue_b'))} ์กฐ์",
f" ์์
์ด์ต: {_fmt_krw(d.get('operating_income_b'))} ์กฐ์ (OPM: {d.get('operating_margin', 'N/A')}%)",
f" ์์ด์ต: {_fmt_krw(d.get('net_income_b'))} ์กฐ์",
]
if d.get("revenue_yoy_pct") is not None:
lines.append(f" ๋งค์ถ YoY: {d['revenue_yoy_pct']:+.1f}%")
if d.get("op_income_yoy_pct") is not None:
lines.append(f" ์์
์ด์ต YoY: {d['op_income_yoy_pct']:+.1f}%")
else:
lines += [
f" ๋งค์ถ: {_fmt_krw(d.get('revenue_t'))} ์กฐ์",
f" ์์
์ด์ต: {_fmt_krw(d.get('operating_income_t'))} ์กฐ์ (OPM: {d.get('operating_margin', 'N/A')}%)",
f" ์์ด์ต: {_fmt_krw(d.get('net_income_t'))} ์กฐ์",
]
return lines
def _format_quarterly_table(quarters):
if not quarters:
return []
lines = [
"\nโ ์ต๊ทผ ๋ถ๊ธฐ๋ณ ์ค์ ์ถ์ด",
f" {'๋ถ๊ธฐ':<8} {'๋งค์ถ(์กฐ์)':>10} {'์์
์ด์ต(์กฐ์)':>14} {'OPM':>6} {'๋งค์ถYoY':>8}",
" " + "-" * 54,
]
for q in quarters[:6]:
yoy = f"{q['revenue_yoy_pct']:+.1f}%" if q.get("revenue_yoy_pct") is not None else "N/A"
lines.append(
f" {q['period']:<8} "
f"{_fmt_krw(q.get('revenue_b')):>10} "
f"{_fmt_krw(q.get('operating_income_b')):>14} "
f"{str(q.get('operating_margin', 'N/A')) + '%':>6} "
f"{yoy:>8}"
)
return lines
def _format_annual_table(annual):
if not annual:
return []
lines = [
"\nโ ์ฐ๊ฐ ์ค์ ์ถ์ด",
f" {'์ฐ๋':<6} {'๋งค์ถ(์กฐ์)':>10} {'์์
์ด์ต(์กฐ์)':>14} {'OPM':>6}",
" " + "-" * 40,
]
for a in annual[:4]:
lines.append(
f" {a['year']:<6} "
f"{_fmt_krw(a.get('revenue_t')):>10} "
f"{_fmt_krw(a.get('operating_income_t')):>14} "
f"{str(a.get('operating_margin', 'N/A')) + '%':>6}"
)
return lines
def _format_eps_surprise(surprises):
if not surprises:
return []
lines = ["\nโ EPS ์ํ๋ผ์ด์ฆ (์ต๊ทผ)"]
for s in surprises[:4]:
surp = f"{s['surprise_pct']:+.1f}%" if s.get("surprise_pct") is not None else "N/A"
lines.append(
f" {s['period']:<10} "
f"์ค์ : {s.get('eps_actual', 'N/A')} "
f"์์: {s.get('eps_estimate', 'N/A')} "
f"์ํ๋ผ์ด์ฆ: {surp}"
)
return lines
def build_web_search_section(ws_results):
lines = ["[์น ๊ฒ์ ์ต์ ์ ๋ณด (์ค์๊ฐ)]"]
for i, block in enumerate(ws_results, 1):
text = block.get("text", "").strip()
citations = block.get("citations", [])
if text:
snippet = text[:800] + ("..." if len(text) > 800 else "")
lines.append(f"\nโ ๊ฒ์ ๊ฒฐ๊ณผ {i}")
lines.append(snippet)
if citations:
lines.append(" [์ถ์ฒ]")
for c in citations[:5]:
title = c.get("title", "")
url = c.get("url", "")
if title or url:
lines.append(f" โข {title} {url}")
return "\n".join(lines)
def build_context(market_data, intent, news_str=None):
sections = []
fd = market_data.fundamentals
pd = market_data.price_data
td = market_data.technicals
ed = market_data.earnings_data
ws = market_data.web_search_results
if fd:
sections.append(build_company_section(fd, market_data.ticker))
if pd:
sections.append(build_price_section(pd))
if fd and any(fd.get(k) for k in ["pe_ratio", "pb_ratio", "roe"]):
sections.append(build_valuation_section(fd))
if fd.get("analyst_target"):
sections.append(build_analyst_section(fd, pd))
if fd and any(fd.get(k) is not None for k in [
"shares_outstanding_b",
"held_percent_institutions",
"short_ratio",
"dividend_rate",
]):
sections.append(build_capital_structure_section(fd))
if td:
sections.append(build_technicals_section(td))
if ed:
sections.append(build_earnings_section(ed, intent))
if market_data.news_snippets:
news_text = "\n".join(f" โข {n}" for n in market_data.news_snippets[:6])
sections.append(f"[์ต๊ทผ ๋ด์ค ํค๋๋ผ์ธ (yfinance)]\n{news_text}")
if news_str:
sections.append(f"[๊ตฌ๊ธ ๋ด์ค ์์ฝ]\n{news_str}")
if ws:
sections.append(build_web_search_section(ws))
return "\n\n".join(sections)
|