Compare commits
4 Commits
feature/is
...
feature/is
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
d64e072f06 | ||
|
|
b2312fbe01 | ||
|
|
98c4a2413c | ||
| 6fba7c7ae8 |
291
src/main.py
291
src/main.py
@@ -10,14 +10,14 @@ import argparse
|
||||
import asyncio
|
||||
import logging
|
||||
import signal
|
||||
import sys
|
||||
from datetime import UTC, datetime
|
||||
from typing import Any
|
||||
|
||||
from src.analysis.scanner import MarketScanner
|
||||
from src.analysis.smart_scanner import ScanCandidate, SmartVolatilityScanner
|
||||
from src.analysis.volatility import VolatilityAnalyzer
|
||||
from src.brain.gemini_client import GeminiClient
|
||||
from src.brain.context_selector import ContextSelector
|
||||
from src.brain.gemini_client import GeminiClient, TradeDecision
|
||||
from src.broker.kis_api import KISBroker
|
||||
from src.broker.overseas import OverseasBroker
|
||||
from src.config import Settings
|
||||
@@ -31,6 +31,10 @@ from src.logging.decision_logger import DecisionLogger
|
||||
from src.logging_config import setup_logging
|
||||
from src.markets.schedule import MarketInfo, get_next_market_open, get_open_markets
|
||||
from src.notifications.telegram_client import TelegramClient, TelegramCommandHandler
|
||||
from src.strategy.models import DayPlaybook
|
||||
from src.strategy.playbook_store import PlaybookStore
|
||||
from src.strategy.pre_market_planner import PreMarketPlanner
|
||||
from src.strategy.scenario_engine import ScenarioEngine
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
@@ -75,7 +79,8 @@ TRADE_SESSION_INTERVAL_HOURS = 6 # Hours between sessions
|
||||
async def trading_cycle(
|
||||
broker: KISBroker,
|
||||
overseas_broker: OverseasBroker,
|
||||
brain: GeminiClient,
|
||||
scenario_engine: ScenarioEngine,
|
||||
playbook: DayPlaybook,
|
||||
risk: RiskManager,
|
||||
db_conn: Any,
|
||||
decision_logger: DecisionLogger,
|
||||
@@ -84,7 +89,7 @@ async def trading_cycle(
|
||||
telegram: TelegramClient,
|
||||
market: MarketInfo,
|
||||
stock_code: str,
|
||||
scan_candidates: dict[str, ScanCandidate],
|
||||
scan_candidates: dict[str, dict[str, ScanCandidate]],
|
||||
) -> None:
|
||||
"""Execute one trading cycle for a single stock."""
|
||||
cycle_start_time = asyncio.get_event_loop().time()
|
||||
@@ -135,13 +140,27 @@ async def trading_cycle(
|
||||
else 0.0
|
||||
)
|
||||
|
||||
market_data = {
|
||||
market_data: dict[str, Any] = {
|
||||
"stock_code": stock_code,
|
||||
"market_name": market.name,
|
||||
"current_price": current_price,
|
||||
"foreigner_net": foreigner_net,
|
||||
}
|
||||
|
||||
# Enrich market_data with scanner metrics for scenario engine
|
||||
market_candidates = scan_candidates.get(market.code, {})
|
||||
candidate = market_candidates.get(stock_code)
|
||||
if candidate:
|
||||
market_data["rsi"] = candidate.rsi
|
||||
market_data["volume_ratio"] = candidate.volume_ratio
|
||||
|
||||
# Build portfolio data for global rule evaluation
|
||||
portfolio_data = {
|
||||
"portfolio_pnl_pct": pnl_pct,
|
||||
"total_cash": total_cash,
|
||||
"total_eval": total_eval,
|
||||
}
|
||||
|
||||
# 1.5. Get volatility metrics from context store (L7_REALTIME)
|
||||
latest_timeframe = context_store.get_latest_timeframe(ContextLayer.L7_REALTIME)
|
||||
volatility_score = 50.0 # Default normal volatility
|
||||
@@ -178,8 +197,13 @@ async def trading_cycle(
|
||||
volume_surge,
|
||||
)
|
||||
|
||||
# 2. Ask the brain for a decision
|
||||
decision = await brain.decide(market_data)
|
||||
# 2. Evaluate scenario (local, no API call)
|
||||
match = scenario_engine.evaluate(playbook, stock_code, market_data, portfolio_data)
|
||||
decision = TradeDecision(
|
||||
action=match.action.value,
|
||||
confidence=match.confidence,
|
||||
rationale=match.rationale,
|
||||
)
|
||||
logger.info(
|
||||
"Decision for %s (%s): %s (confidence=%d)",
|
||||
stock_code,
|
||||
@@ -188,6 +212,19 @@ async def trading_cycle(
|
||||
decision.confidence,
|
||||
)
|
||||
|
||||
# 2.1. Notify scenario match
|
||||
if match.matched_scenario is not None:
|
||||
try:
|
||||
condition_parts = [f"{k}={v}" for k, v in match.match_details.items()]
|
||||
await telegram.notify_scenario_matched(
|
||||
stock_code=stock_code,
|
||||
action=decision.action,
|
||||
condition_summary=", ".join(condition_parts) if condition_parts else "matched",
|
||||
confidence=float(decision.confidence),
|
||||
)
|
||||
except Exception as exc:
|
||||
logger.warning("Scenario matched notification failed: %s", exc)
|
||||
|
||||
# 2.5. Log decision with context snapshot
|
||||
context_snapshot = {
|
||||
"L1": {
|
||||
@@ -200,7 +237,7 @@ async def trading_cycle(
|
||||
"purchase_total": purchase_total,
|
||||
"pnl_pct": pnl_pct,
|
||||
},
|
||||
# L3-L7 will be populated when context tree is implemented
|
||||
"scenario_match": match.match_details,
|
||||
}
|
||||
input_data = {
|
||||
"current_price": current_price,
|
||||
@@ -279,8 +316,8 @@ async def trading_cycle(
|
||||
|
||||
# 6. Log trade with selection context
|
||||
selection_context = None
|
||||
if stock_code in scan_candidates:
|
||||
candidate = scan_candidates[stock_code]
|
||||
if stock_code in market_candidates:
|
||||
candidate = market_candidates[stock_code]
|
||||
selection_context = {
|
||||
"rsi": candidate.rsi,
|
||||
"volume_ratio": candidate.volume_ratio,
|
||||
@@ -324,7 +361,9 @@ async def trading_cycle(
|
||||
async def run_daily_session(
|
||||
broker: KISBroker,
|
||||
overseas_broker: OverseasBroker,
|
||||
brain: GeminiClient,
|
||||
scenario_engine: ScenarioEngine,
|
||||
playbook_store: PlaybookStore,
|
||||
pre_market_planner: PreMarketPlanner,
|
||||
risk: RiskManager,
|
||||
db_conn: Any,
|
||||
decision_logger: DecisionLogger,
|
||||
@@ -336,10 +375,8 @@ async def run_daily_session(
|
||||
) -> None:
|
||||
"""Execute one daily trading session.
|
||||
|
||||
Designed for API efficiency with Gemini Free tier:
|
||||
- Batch decision making (1 API call per market)
|
||||
- Runs N times per day at fixed intervals
|
||||
- Minimizes API usage while maintaining trading capability
|
||||
V2 proactive strategy: 1 Gemini call for playbook generation,
|
||||
then local scenario evaluation per stock (0 API calls).
|
||||
"""
|
||||
# Get currently open markets
|
||||
open_markets = get_open_markets(settings.enabled_market_list)
|
||||
@@ -352,27 +389,66 @@ async def run_daily_session(
|
||||
|
||||
# Process each open market
|
||||
for market in open_markets:
|
||||
# Use market-local date for playbook keying
|
||||
market_today = datetime.now(market.timezone).date()
|
||||
|
||||
# Dynamic stock discovery via scanner (no static watchlists)
|
||||
candidates_list: list[ScanCandidate] = []
|
||||
try:
|
||||
candidates = await smart_scanner.scan()
|
||||
watchlist = [c.stock_code for c in candidates] if candidates else []
|
||||
candidates_list = await smart_scanner.scan() if smart_scanner else []
|
||||
except Exception as exc:
|
||||
logger.error("Smart Scanner failed for %s: %s", market.name, exc)
|
||||
watchlist = []
|
||||
|
||||
if not watchlist:
|
||||
if not candidates_list:
|
||||
logger.info("No scanner candidates for market %s — skipping", market.code)
|
||||
continue
|
||||
|
||||
watchlist = [c.stock_code for c in candidates_list]
|
||||
candidate_map = {c.stock_code: c for c in candidates_list}
|
||||
logger.info("Processing market: %s (%d stocks)", market.name, len(watchlist))
|
||||
|
||||
# Generate or load playbook (1 Gemini API call per market per day)
|
||||
playbook = playbook_store.load(market_today, market.code)
|
||||
if playbook is None:
|
||||
try:
|
||||
playbook = await pre_market_planner.generate_playbook(
|
||||
market=market.code,
|
||||
candidates=candidates_list,
|
||||
today=market_today,
|
||||
)
|
||||
playbook_store.save(playbook)
|
||||
try:
|
||||
await telegram.notify_playbook_generated(
|
||||
market=market.code,
|
||||
stock_count=playbook.stock_count,
|
||||
scenario_count=playbook.scenario_count,
|
||||
token_count=playbook.token_count,
|
||||
)
|
||||
except Exception as exc:
|
||||
logger.warning("Playbook notification failed: %s", exc)
|
||||
logger.info(
|
||||
"Generated playbook for %s: %d stocks, %d scenarios",
|
||||
market.code, playbook.stock_count, playbook.scenario_count,
|
||||
)
|
||||
except Exception as exc:
|
||||
logger.error("Playbook generation failed for %s: %s", market.code, exc)
|
||||
try:
|
||||
await telegram.notify_playbook_failed(
|
||||
market=market.code, reason=str(exc)[:200],
|
||||
)
|
||||
except Exception as notify_exc:
|
||||
logger.warning("Playbook failed notification error: %s", notify_exc)
|
||||
playbook = PreMarketPlanner._empty_playbook(market_today, market.code)
|
||||
|
||||
# Collect market data for all stocks from scanner
|
||||
stocks_data = []
|
||||
for stock_code in watchlist:
|
||||
try:
|
||||
if market.is_domestic:
|
||||
orderbook = await broker.get_orderbook(stock_code)
|
||||
current_price = safe_float(orderbook.get("output1", {}).get("stck_prpr", "0"))
|
||||
current_price = safe_float(
|
||||
orderbook.get("output1", {}).get("stck_prpr", "0")
|
||||
)
|
||||
foreigner_net = safe_float(
|
||||
orderbook.get("output1", {}).get("frgn_ntby_qty", "0")
|
||||
)
|
||||
@@ -380,17 +456,23 @@ async def run_daily_session(
|
||||
price_data = await overseas_broker.get_overseas_price(
|
||||
market.exchange_code, stock_code
|
||||
)
|
||||
current_price = safe_float(price_data.get("output", {}).get("last", "0"))
|
||||
current_price = safe_float(
|
||||
price_data.get("output", {}).get("last", "0")
|
||||
)
|
||||
foreigner_net = 0.0
|
||||
|
||||
stocks_data.append(
|
||||
{
|
||||
stock_data: dict[str, Any] = {
|
||||
"stock_code": stock_code,
|
||||
"market_name": market.name,
|
||||
"current_price": current_price,
|
||||
"foreigner_net": foreigner_net,
|
||||
}
|
||||
)
|
||||
# Enrich with scanner metrics
|
||||
cand = candidate_map.get(stock_code)
|
||||
if cand:
|
||||
stock_data["rsi"] = cand.rsi
|
||||
stock_data["volume_ratio"] = cand.volume_ratio
|
||||
stocks_data.append(stock_data)
|
||||
except Exception as exc:
|
||||
logger.error("Failed to fetch data for %s: %s", stock_code, exc)
|
||||
continue
|
||||
@@ -399,17 +481,19 @@ async def run_daily_session(
|
||||
logger.warning("No valid stock data for market %s", market.code)
|
||||
continue
|
||||
|
||||
# Get batch decisions (1 API call for all stocks in this market)
|
||||
logger.info("Requesting batch decision for %d stocks in %s", len(stocks_data), market.name)
|
||||
decisions = await brain.decide_batch(stocks_data)
|
||||
|
||||
# Get balance data once for the market
|
||||
if market.is_domestic:
|
||||
balance_data = await broker.get_balance()
|
||||
output2 = balance_data.get("output2", [{}])
|
||||
total_eval = safe_float(output2[0].get("tot_evlu_amt", "0")) if output2 else 0
|
||||
total_cash = safe_float(output2[0].get("dnca_tot_amt", "0")) if output2 else 0
|
||||
purchase_total = safe_float(output2[0].get("pchs_amt_smtl_amt", "0")) if output2 else 0
|
||||
total_eval = safe_float(
|
||||
output2[0].get("tot_evlu_amt", "0")
|
||||
) if output2 else 0
|
||||
total_cash = safe_float(
|
||||
output2[0].get("dnca_tot_amt", "0")
|
||||
) if output2 else 0
|
||||
purchase_total = safe_float(
|
||||
output2[0].get("pchs_amt_smtl_amt", "0")
|
||||
) if output2 else 0
|
||||
else:
|
||||
balance_data = await overseas_broker.get_overseas_balance(market.exchange_code)
|
||||
output2 = balance_data.get("output2", [{}])
|
||||
@@ -422,21 +506,37 @@ async def run_daily_session(
|
||||
|
||||
total_eval = safe_float(balance_info.get("frcr_evlu_tota", "0") or "0")
|
||||
total_cash = safe_float(balance_info.get("frcr_dncl_amt_2", "0") or "0")
|
||||
purchase_total = safe_float(balance_info.get("frcr_buy_amt_smtl", "0") or "0")
|
||||
purchase_total = safe_float(
|
||||
balance_info.get("frcr_buy_amt_smtl", "0") or "0"
|
||||
)
|
||||
|
||||
# Calculate daily P&L %
|
||||
pnl_pct = (
|
||||
((total_eval - purchase_total) / purchase_total * 100) if purchase_total > 0 else 0.0
|
||||
((total_eval - purchase_total) / purchase_total * 100)
|
||||
if purchase_total > 0
|
||||
else 0.0
|
||||
)
|
||||
portfolio_data = {
|
||||
"portfolio_pnl_pct": pnl_pct,
|
||||
"total_cash": total_cash,
|
||||
"total_eval": total_eval,
|
||||
}
|
||||
|
||||
# Execute decisions for each stock
|
||||
# Evaluate scenarios for each stock (local, no API calls)
|
||||
logger.info(
|
||||
"Evaluating %d stocks against playbook for %s",
|
||||
len(stocks_data), market.name,
|
||||
)
|
||||
for stock_data in stocks_data:
|
||||
stock_code = stock_data["stock_code"]
|
||||
decision = decisions.get(stock_code)
|
||||
|
||||
if not decision:
|
||||
logger.warning("No decision for %s — skipping", stock_code)
|
||||
continue
|
||||
match = scenario_engine.evaluate(
|
||||
playbook, stock_code, stock_data, portfolio_data,
|
||||
)
|
||||
decision = TradeDecision(
|
||||
action=match.action.value,
|
||||
confidence=match.confidence,
|
||||
rationale=match.rationale,
|
||||
)
|
||||
|
||||
logger.info(
|
||||
"Decision for %s (%s): %s (confidence=%d)",
|
||||
@@ -458,6 +558,7 @@ async def run_daily_session(
|
||||
"purchase_total": purchase_total,
|
||||
"pnl_pct": pnl_pct,
|
||||
},
|
||||
"scenario_match": match.match_details,
|
||||
}
|
||||
input_data = {
|
||||
"current_price": stock_data["current_price"],
|
||||
@@ -509,7 +610,9 @@ async def run_daily_session(
|
||||
threshold=exc.threshold,
|
||||
)
|
||||
except Exception as notify_exc:
|
||||
logger.warning("Circuit breaker notification failed: %s", notify_exc)
|
||||
logger.warning(
|
||||
"Circuit breaker notification failed: %s", notify_exc
|
||||
)
|
||||
raise
|
||||
|
||||
# Send order
|
||||
@@ -544,7 +647,9 @@ async def run_daily_session(
|
||||
except Exception as exc:
|
||||
logger.warning("Telegram notification failed: %s", exc)
|
||||
except Exception as exc:
|
||||
logger.error("Order execution failed for %s: %s", stock_code, exc)
|
||||
logger.error(
|
||||
"Order execution failed for %s: %s", stock_code, exc
|
||||
)
|
||||
continue
|
||||
|
||||
# Log trade
|
||||
@@ -571,6 +676,20 @@ async def run(settings: Settings) -> None:
|
||||
decision_logger = DecisionLogger(db_conn)
|
||||
context_store = ContextStore(db_conn)
|
||||
|
||||
# V2 proactive strategy components
|
||||
context_selector = ContextSelector(context_store)
|
||||
scenario_engine = ScenarioEngine()
|
||||
playbook_store = PlaybookStore(db_conn)
|
||||
pre_market_planner = PreMarketPlanner(
|
||||
gemini_client=brain,
|
||||
context_store=context_store,
|
||||
context_selector=context_selector,
|
||||
settings=settings,
|
||||
)
|
||||
|
||||
# Track playbooks per market (in-memory cache)
|
||||
playbooks: dict[str, DayPlaybook] = {}
|
||||
|
||||
# Initialize Telegram notifications
|
||||
telegram = TelegramClient(
|
||||
bot_token=settings.TELEGRAM_BOT_TOKEN,
|
||||
@@ -732,8 +851,8 @@ async def run(settings: Settings) -> None:
|
||||
settings=settings,
|
||||
)
|
||||
|
||||
# Track scan candidates for selection context logging
|
||||
scan_candidates: dict[str, ScanCandidate] = {} # stock_code -> candidate
|
||||
# Track scan candidates per market for selection context logging
|
||||
scan_candidates: dict[str, dict[str, ScanCandidate]] = {} # market -> {stock_code -> candidate}
|
||||
|
||||
# Active stocks per market (dynamically discovered by scanner)
|
||||
active_stocks: dict[str, list[str]] = {} # market_code -> [stock_codes]
|
||||
@@ -802,7 +921,9 @@ async def run(settings: Settings) -> None:
|
||||
await run_daily_session(
|
||||
broker,
|
||||
overseas_broker,
|
||||
brain,
|
||||
scenario_engine,
|
||||
playbook_store,
|
||||
pre_market_planner,
|
||||
risk,
|
||||
db_conn,
|
||||
decision_logger,
|
||||
@@ -850,6 +971,8 @@ async def run(settings: Settings) -> None:
|
||||
except Exception as exc:
|
||||
logger.warning("Market close notification failed: %s", exc)
|
||||
_market_states[market_code] = False
|
||||
# Clear playbook for closed market (new one generated next open)
|
||||
playbooks.pop(market_code, None)
|
||||
|
||||
# No markets open — wait until next market opens
|
||||
try:
|
||||
@@ -887,7 +1010,8 @@ async def run(settings: Settings) -> None:
|
||||
# Smart Scanner: dynamic stock discovery (no static watchlists)
|
||||
now_timestamp = asyncio.get_event_loop().time()
|
||||
last_scan = last_scan_time.get(market.code, 0.0)
|
||||
if now_timestamp - last_scan >= SCAN_INTERVAL_SECONDS:
|
||||
rescan_interval = settings.RESCAN_INTERVAL_SECONDS
|
||||
if now_timestamp - last_scan >= rescan_interval:
|
||||
try:
|
||||
logger.info("Smart Scanner: Scanning %s market", market.name)
|
||||
|
||||
@@ -899,9 +1023,10 @@ async def run(settings: Settings) -> None:
|
||||
candidates
|
||||
)
|
||||
|
||||
# Store candidates for selection context logging
|
||||
for candidate in candidates:
|
||||
scan_candidates[candidate.stock_code] = candidate
|
||||
# Store candidates per market for selection context logging
|
||||
scan_candidates[market.code] = {
|
||||
c.stock_code: c for c in candidates
|
||||
}
|
||||
|
||||
logger.info(
|
||||
"Smart Scanner: Found %d candidates for %s: %s",
|
||||
@@ -909,6 +1034,62 @@ async def run(settings: Settings) -> None:
|
||||
market.name,
|
||||
[f"{c.stock_code}(RSI={c.rsi:.0f})" for c in candidates],
|
||||
)
|
||||
|
||||
# Get market-local date for playbook keying
|
||||
market_today = datetime.now(
|
||||
market.timezone
|
||||
).date()
|
||||
|
||||
# Load or generate playbook (1 Gemini call per market per day)
|
||||
if market.code not in playbooks:
|
||||
# Try DB first (survives process restart)
|
||||
stored_pb = playbook_store.load(market_today, market.code)
|
||||
if stored_pb is not None:
|
||||
playbooks[market.code] = stored_pb
|
||||
logger.info(
|
||||
"Loaded existing playbook for %s from DB"
|
||||
" (%d stocks, %d scenarios)",
|
||||
market.code,
|
||||
stored_pb.stock_count,
|
||||
stored_pb.scenario_count,
|
||||
)
|
||||
else:
|
||||
try:
|
||||
pb = await pre_market_planner.generate_playbook(
|
||||
market=market.code,
|
||||
candidates=candidates,
|
||||
today=market_today,
|
||||
)
|
||||
playbook_store.save(pb)
|
||||
playbooks[market.code] = pb
|
||||
try:
|
||||
await telegram.notify_playbook_generated(
|
||||
market=market.code,
|
||||
stock_count=pb.stock_count,
|
||||
scenario_count=pb.scenario_count,
|
||||
token_count=pb.token_count,
|
||||
)
|
||||
except Exception as exc:
|
||||
logger.warning(
|
||||
"Playbook notification failed: %s", exc
|
||||
)
|
||||
except Exception as exc:
|
||||
logger.error(
|
||||
"Playbook generation failed for %s: %s",
|
||||
market.code, exc,
|
||||
)
|
||||
try:
|
||||
await telegram.notify_playbook_failed(
|
||||
market=market.code,
|
||||
reason=str(exc)[:200],
|
||||
)
|
||||
except Exception:
|
||||
pass
|
||||
playbooks[market.code] = (
|
||||
PreMarketPlanner._empty_playbook(
|
||||
market_today, market.code
|
||||
)
|
||||
)
|
||||
else:
|
||||
logger.info(
|
||||
"Smart Scanner: No candidates for %s — no trades", market.name
|
||||
@@ -933,13 +1114,22 @@ async def run(settings: Settings) -> None:
|
||||
if shutdown.is_set():
|
||||
break
|
||||
|
||||
# Get playbook for this market
|
||||
market_playbook = playbooks.get(
|
||||
market.code,
|
||||
PreMarketPlanner._empty_playbook(
|
||||
datetime.now(market.timezone).date(), market.code
|
||||
),
|
||||
)
|
||||
|
||||
# Retry logic for connection errors
|
||||
for attempt in range(1, MAX_CONNECTION_RETRIES + 1):
|
||||
try:
|
||||
await trading_cycle(
|
||||
broker,
|
||||
overseas_broker,
|
||||
brain,
|
||||
scenario_engine,
|
||||
market_playbook,
|
||||
risk,
|
||||
db_conn,
|
||||
decision_logger,
|
||||
@@ -988,7 +1178,8 @@ async def run(settings: Settings) -> None:
|
||||
metrics = await priority_queue.get_metrics()
|
||||
if metrics.total_enqueued > 0:
|
||||
logger.info(
|
||||
"Priority queue metrics: enqueued=%d, dequeued=%d, size=%d, timeouts=%d, errors=%d",
|
||||
"Priority queue metrics: enqueued=%d, dequeued=%d,"
|
||||
" size=%d, timeouts=%d, errors=%d",
|
||||
metrics.total_enqueued,
|
||||
metrics.total_dequeued,
|
||||
metrics.current_size,
|
||||
|
||||
@@ -1,12 +1,46 @@
|
||||
"""Tests for main trading loop telegram integration."""
|
||||
"""Tests for main trading loop integration."""
|
||||
|
||||
import asyncio
|
||||
from datetime import date
|
||||
from unittest.mock import AsyncMock, MagicMock, patch
|
||||
|
||||
import pytest
|
||||
|
||||
from src.core.risk_manager import CircuitBreakerTripped, FatFingerRejected
|
||||
from src.main import safe_float, trading_cycle
|
||||
from src.strategy.models import (
|
||||
DayPlaybook,
|
||||
ScenarioAction,
|
||||
StockCondition,
|
||||
StockScenario,
|
||||
)
|
||||
from src.strategy.scenario_engine import ScenarioEngine, ScenarioMatch
|
||||
|
||||
|
||||
def _make_playbook(market: str = "KR") -> DayPlaybook:
|
||||
"""Create a minimal empty playbook for testing."""
|
||||
return DayPlaybook(date=date(2026, 2, 8), market=market)
|
||||
|
||||
|
||||
def _make_buy_match(stock_code: str = "005930") -> ScenarioMatch:
|
||||
"""Create a ScenarioMatch that returns BUY."""
|
||||
return ScenarioMatch(
|
||||
stock_code=stock_code,
|
||||
matched_scenario=None,
|
||||
action=ScenarioAction.BUY,
|
||||
confidence=85,
|
||||
rationale="Test buy",
|
||||
)
|
||||
|
||||
|
||||
def _make_hold_match(stock_code: str = "005930") -> ScenarioMatch:
|
||||
"""Create a ScenarioMatch that returns HOLD."""
|
||||
return ScenarioMatch(
|
||||
stock_code=stock_code,
|
||||
matched_scenario=None,
|
||||
action=ScenarioAction.HOLD,
|
||||
confidence=0,
|
||||
rationale="No scenario conditions met",
|
||||
)
|
||||
|
||||
|
||||
class TestSafeFloat:
|
||||
@@ -81,15 +115,16 @@ class TestTradingCycleTelegramIntegration:
|
||||
return broker
|
||||
|
||||
@pytest.fixture
|
||||
def mock_brain(self) -> MagicMock:
|
||||
"""Create mock brain that decides to buy."""
|
||||
brain = MagicMock()
|
||||
decision = MagicMock()
|
||||
decision.action = "BUY"
|
||||
decision.confidence = 85
|
||||
decision.rationale = "Test buy"
|
||||
brain.decide = AsyncMock(return_value=decision)
|
||||
return brain
|
||||
def mock_scenario_engine(self) -> MagicMock:
|
||||
"""Create mock scenario engine that returns BUY."""
|
||||
engine = MagicMock(spec=ScenarioEngine)
|
||||
engine.evaluate = MagicMock(return_value=_make_buy_match())
|
||||
return engine
|
||||
|
||||
@pytest.fixture
|
||||
def mock_playbook(self) -> DayPlaybook:
|
||||
"""Create a minimal day playbook."""
|
||||
return _make_playbook()
|
||||
|
||||
@pytest.fixture
|
||||
def mock_risk(self) -> MagicMock:
|
||||
@@ -134,6 +169,7 @@ class TestTradingCycleTelegramIntegration:
|
||||
telegram.notify_trade_execution = AsyncMock()
|
||||
telegram.notify_fat_finger = AsyncMock()
|
||||
telegram.notify_circuit_breaker = AsyncMock()
|
||||
telegram.notify_scenario_matched = AsyncMock()
|
||||
return telegram
|
||||
|
||||
@pytest.fixture
|
||||
@@ -151,7 +187,8 @@ class TestTradingCycleTelegramIntegration:
|
||||
self,
|
||||
mock_broker: MagicMock,
|
||||
mock_overseas_broker: MagicMock,
|
||||
mock_brain: MagicMock,
|
||||
mock_scenario_engine: MagicMock,
|
||||
mock_playbook: DayPlaybook,
|
||||
mock_risk: MagicMock,
|
||||
mock_db: MagicMock,
|
||||
mock_decision_logger: MagicMock,
|
||||
@@ -165,7 +202,8 @@ class TestTradingCycleTelegramIntegration:
|
||||
await trading_cycle(
|
||||
broker=mock_broker,
|
||||
overseas_broker=mock_overseas_broker,
|
||||
brain=mock_brain,
|
||||
scenario_engine=mock_scenario_engine,
|
||||
playbook=mock_playbook,
|
||||
risk=mock_risk,
|
||||
db_conn=mock_db,
|
||||
decision_logger=mock_decision_logger,
|
||||
@@ -190,7 +228,8 @@ class TestTradingCycleTelegramIntegration:
|
||||
self,
|
||||
mock_broker: MagicMock,
|
||||
mock_overseas_broker: MagicMock,
|
||||
mock_brain: MagicMock,
|
||||
mock_scenario_engine: MagicMock,
|
||||
mock_playbook: DayPlaybook,
|
||||
mock_risk: MagicMock,
|
||||
mock_db: MagicMock,
|
||||
mock_decision_logger: MagicMock,
|
||||
@@ -208,7 +247,8 @@ class TestTradingCycleTelegramIntegration:
|
||||
await trading_cycle(
|
||||
broker=mock_broker,
|
||||
overseas_broker=mock_overseas_broker,
|
||||
brain=mock_brain,
|
||||
scenario_engine=mock_scenario_engine,
|
||||
playbook=mock_playbook,
|
||||
risk=mock_risk,
|
||||
db_conn=mock_db,
|
||||
decision_logger=mock_decision_logger,
|
||||
@@ -228,7 +268,8 @@ class TestTradingCycleTelegramIntegration:
|
||||
self,
|
||||
mock_broker: MagicMock,
|
||||
mock_overseas_broker: MagicMock,
|
||||
mock_brain: MagicMock,
|
||||
mock_scenario_engine: MagicMock,
|
||||
mock_playbook: DayPlaybook,
|
||||
mock_risk: MagicMock,
|
||||
mock_db: MagicMock,
|
||||
mock_decision_logger: MagicMock,
|
||||
@@ -250,7 +291,8 @@ class TestTradingCycleTelegramIntegration:
|
||||
await trading_cycle(
|
||||
broker=mock_broker,
|
||||
overseas_broker=mock_overseas_broker,
|
||||
brain=mock_brain,
|
||||
scenario_engine=mock_scenario_engine,
|
||||
playbook=mock_playbook,
|
||||
risk=mock_risk,
|
||||
db_conn=mock_db,
|
||||
decision_logger=mock_decision_logger,
|
||||
@@ -275,7 +317,8 @@ class TestTradingCycleTelegramIntegration:
|
||||
self,
|
||||
mock_broker: MagicMock,
|
||||
mock_overseas_broker: MagicMock,
|
||||
mock_brain: MagicMock,
|
||||
mock_scenario_engine: MagicMock,
|
||||
mock_playbook: DayPlaybook,
|
||||
mock_risk: MagicMock,
|
||||
mock_db: MagicMock,
|
||||
mock_decision_logger: MagicMock,
|
||||
@@ -299,7 +342,8 @@ class TestTradingCycleTelegramIntegration:
|
||||
await trading_cycle(
|
||||
broker=mock_broker,
|
||||
overseas_broker=mock_overseas_broker,
|
||||
brain=mock_brain,
|
||||
scenario_engine=mock_scenario_engine,
|
||||
playbook=mock_playbook,
|
||||
risk=mock_risk,
|
||||
db_conn=mock_db,
|
||||
decision_logger=mock_decision_logger,
|
||||
@@ -319,7 +363,8 @@ class TestTradingCycleTelegramIntegration:
|
||||
self,
|
||||
mock_broker: MagicMock,
|
||||
mock_overseas_broker: MagicMock,
|
||||
mock_brain: MagicMock,
|
||||
mock_scenario_engine: MagicMock,
|
||||
mock_playbook: DayPlaybook,
|
||||
mock_risk: MagicMock,
|
||||
mock_db: MagicMock,
|
||||
mock_decision_logger: MagicMock,
|
||||
@@ -329,18 +374,15 @@ class TestTradingCycleTelegramIntegration:
|
||||
mock_market: MagicMock,
|
||||
) -> None:
|
||||
"""Test no trade notification sent when decision is HOLD."""
|
||||
# Change brain decision to HOLD
|
||||
decision = MagicMock()
|
||||
decision.action = "HOLD"
|
||||
decision.confidence = 50
|
||||
decision.rationale = "Insufficient signal"
|
||||
mock_brain.decide = AsyncMock(return_value=decision)
|
||||
# Scenario engine returns HOLD
|
||||
mock_scenario_engine.evaluate = MagicMock(return_value=_make_hold_match())
|
||||
|
||||
with patch("src.main.log_trade"):
|
||||
await trading_cycle(
|
||||
broker=mock_broker,
|
||||
overseas_broker=mock_overseas_broker,
|
||||
brain=mock_brain,
|
||||
scenario_engine=mock_scenario_engine,
|
||||
playbook=mock_playbook,
|
||||
risk=mock_risk,
|
||||
db_conn=mock_db,
|
||||
decision_logger=mock_decision_logger,
|
||||
@@ -472,15 +514,16 @@ class TestOverseasBalanceParsing:
|
||||
return market
|
||||
|
||||
@pytest.fixture
|
||||
def mock_brain_hold(self) -> MagicMock:
|
||||
"""Create mock brain that always holds."""
|
||||
brain = MagicMock()
|
||||
decision = MagicMock()
|
||||
decision.action = "HOLD"
|
||||
decision.confidence = 50
|
||||
decision.rationale = "Testing balance parsing"
|
||||
brain.decide = AsyncMock(return_value=decision)
|
||||
return brain
|
||||
def mock_scenario_engine_hold(self) -> MagicMock:
|
||||
"""Create mock scenario engine that always returns HOLD."""
|
||||
engine = MagicMock(spec=ScenarioEngine)
|
||||
engine.evaluate = MagicMock(return_value=_make_hold_match("AAPL"))
|
||||
return engine
|
||||
|
||||
@pytest.fixture
|
||||
def mock_playbook(self) -> DayPlaybook:
|
||||
"""Create a minimal playbook."""
|
||||
return _make_playbook("US")
|
||||
|
||||
@pytest.fixture
|
||||
def mock_risk(self) -> MagicMock:
|
||||
@@ -517,14 +560,17 @@ class TestOverseasBalanceParsing:
|
||||
@pytest.fixture
|
||||
def mock_telegram(self) -> MagicMock:
|
||||
"""Create mock telegram client."""
|
||||
return MagicMock()
|
||||
telegram = MagicMock()
|
||||
telegram.notify_scenario_matched = AsyncMock()
|
||||
return telegram
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_overseas_balance_list_format(
|
||||
self,
|
||||
mock_domestic_broker: MagicMock,
|
||||
mock_overseas_broker_with_list: MagicMock,
|
||||
mock_brain_hold: MagicMock,
|
||||
mock_scenario_engine_hold: MagicMock,
|
||||
mock_playbook: DayPlaybook,
|
||||
mock_risk: MagicMock,
|
||||
mock_db: MagicMock,
|
||||
mock_decision_logger: MagicMock,
|
||||
@@ -539,7 +585,8 @@ class TestOverseasBalanceParsing:
|
||||
await trading_cycle(
|
||||
broker=mock_domestic_broker,
|
||||
overseas_broker=mock_overseas_broker_with_list,
|
||||
brain=mock_brain_hold,
|
||||
scenario_engine=mock_scenario_engine_hold,
|
||||
playbook=mock_playbook,
|
||||
risk=mock_risk,
|
||||
db_conn=mock_db,
|
||||
decision_logger=mock_decision_logger,
|
||||
@@ -559,7 +606,8 @@ class TestOverseasBalanceParsing:
|
||||
self,
|
||||
mock_domestic_broker: MagicMock,
|
||||
mock_overseas_broker_with_dict: MagicMock,
|
||||
mock_brain_hold: MagicMock,
|
||||
mock_scenario_engine_hold: MagicMock,
|
||||
mock_playbook: DayPlaybook,
|
||||
mock_risk: MagicMock,
|
||||
mock_db: MagicMock,
|
||||
mock_decision_logger: MagicMock,
|
||||
@@ -574,7 +622,8 @@ class TestOverseasBalanceParsing:
|
||||
await trading_cycle(
|
||||
broker=mock_domestic_broker,
|
||||
overseas_broker=mock_overseas_broker_with_dict,
|
||||
brain=mock_brain_hold,
|
||||
scenario_engine=mock_scenario_engine_hold,
|
||||
playbook=mock_playbook,
|
||||
risk=mock_risk,
|
||||
db_conn=mock_db,
|
||||
decision_logger=mock_decision_logger,
|
||||
@@ -594,7 +643,8 @@ class TestOverseasBalanceParsing:
|
||||
self,
|
||||
mock_domestic_broker: MagicMock,
|
||||
mock_overseas_broker_with_empty: MagicMock,
|
||||
mock_brain_hold: MagicMock,
|
||||
mock_scenario_engine_hold: MagicMock,
|
||||
mock_playbook: DayPlaybook,
|
||||
mock_risk: MagicMock,
|
||||
mock_db: MagicMock,
|
||||
mock_decision_logger: MagicMock,
|
||||
@@ -609,7 +659,8 @@ class TestOverseasBalanceParsing:
|
||||
await trading_cycle(
|
||||
broker=mock_domestic_broker,
|
||||
overseas_broker=mock_overseas_broker_with_empty,
|
||||
brain=mock_brain_hold,
|
||||
scenario_engine=mock_scenario_engine_hold,
|
||||
playbook=mock_playbook,
|
||||
risk=mock_risk,
|
||||
db_conn=mock_db,
|
||||
decision_logger=mock_decision_logger,
|
||||
@@ -629,7 +680,8 @@ class TestOverseasBalanceParsing:
|
||||
self,
|
||||
mock_domestic_broker: MagicMock,
|
||||
mock_overseas_broker_with_empty_price: MagicMock,
|
||||
mock_brain_hold: MagicMock,
|
||||
mock_scenario_engine_hold: MagicMock,
|
||||
mock_playbook: DayPlaybook,
|
||||
mock_risk: MagicMock,
|
||||
mock_db: MagicMock,
|
||||
mock_decision_logger: MagicMock,
|
||||
@@ -644,7 +696,8 @@ class TestOverseasBalanceParsing:
|
||||
await trading_cycle(
|
||||
broker=mock_domestic_broker,
|
||||
overseas_broker=mock_overseas_broker_with_empty_price,
|
||||
brain=mock_brain_hold,
|
||||
scenario_engine=mock_scenario_engine_hold,
|
||||
playbook=mock_playbook,
|
||||
risk=mock_risk,
|
||||
db_conn=mock_db,
|
||||
decision_logger=mock_decision_logger,
|
||||
@@ -658,3 +711,341 @@ class TestOverseasBalanceParsing:
|
||||
|
||||
# Verify price API was called
|
||||
mock_overseas_broker_with_empty_price.get_overseas_price.assert_called_once()
|
||||
|
||||
|
||||
class TestScenarioEngineIntegration:
|
||||
"""Test scenario engine integration in trading_cycle."""
|
||||
|
||||
@pytest.fixture
|
||||
def mock_broker(self) -> MagicMock:
|
||||
"""Create mock broker with standard domestic data."""
|
||||
broker = MagicMock()
|
||||
broker.get_orderbook = AsyncMock(
|
||||
return_value={
|
||||
"output1": {"stck_prpr": "50000", "frgn_ntby_qty": "100"}
|
||||
}
|
||||
)
|
||||
broker.get_balance = AsyncMock(
|
||||
return_value={
|
||||
"output2": [
|
||||
{
|
||||
"tot_evlu_amt": "10000000",
|
||||
"dnca_tot_amt": "5000000",
|
||||
"pchs_amt_smtl_amt": "9500000",
|
||||
}
|
||||
]
|
||||
}
|
||||
)
|
||||
broker.send_order = AsyncMock(return_value={"msg1": "OK"})
|
||||
return broker
|
||||
|
||||
@pytest.fixture
|
||||
def mock_market(self) -> MagicMock:
|
||||
"""Create mock KR market."""
|
||||
market = MagicMock()
|
||||
market.name = "Korea"
|
||||
market.code = "KR"
|
||||
market.exchange_code = "KRX"
|
||||
market.is_domestic = True
|
||||
return market
|
||||
|
||||
@pytest.fixture
|
||||
def mock_telegram(self) -> MagicMock:
|
||||
"""Create mock telegram with all notification methods."""
|
||||
telegram = MagicMock()
|
||||
telegram.notify_trade_execution = AsyncMock()
|
||||
telegram.notify_scenario_matched = AsyncMock()
|
||||
telegram.notify_fat_finger = AsyncMock()
|
||||
return telegram
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_scenario_engine_called_with_enriched_market_data(
|
||||
self, mock_broker: MagicMock, mock_market: MagicMock, mock_telegram: MagicMock,
|
||||
) -> None:
|
||||
"""Test scenario engine receives market_data enriched with scanner metrics."""
|
||||
from src.analysis.smart_scanner import ScanCandidate
|
||||
|
||||
engine = MagicMock(spec=ScenarioEngine)
|
||||
engine.evaluate = MagicMock(return_value=_make_hold_match())
|
||||
playbook = _make_playbook()
|
||||
|
||||
candidate = ScanCandidate(
|
||||
stock_code="005930", name="Samsung", price=50000,
|
||||
volume=1000000, volume_ratio=3.5, rsi=25.0,
|
||||
signal="oversold", score=85.0,
|
||||
)
|
||||
|
||||
with patch("src.main.log_trade"):
|
||||
await trading_cycle(
|
||||
broker=mock_broker,
|
||||
overseas_broker=MagicMock(),
|
||||
scenario_engine=engine,
|
||||
playbook=playbook,
|
||||
risk=MagicMock(),
|
||||
db_conn=MagicMock(),
|
||||
decision_logger=MagicMock(),
|
||||
context_store=MagicMock(get_latest_timeframe=MagicMock(return_value=None)),
|
||||
criticality_assessor=MagicMock(
|
||||
assess_market_conditions=MagicMock(return_value=MagicMock(value="NORMAL")),
|
||||
get_timeout=MagicMock(return_value=5.0),
|
||||
),
|
||||
telegram=mock_telegram,
|
||||
market=mock_market,
|
||||
stock_code="005930",
|
||||
scan_candidates={"KR": {"005930": candidate}},
|
||||
)
|
||||
|
||||
# Verify evaluate was called
|
||||
engine.evaluate.assert_called_once()
|
||||
call_args = engine.evaluate.call_args
|
||||
market_data = call_args[0][2] # 3rd positional arg
|
||||
portfolio_data = call_args[0][3] # 4th positional arg
|
||||
|
||||
# Scanner data should be enriched into market_data
|
||||
assert market_data["rsi"] == 25.0
|
||||
assert market_data["volume_ratio"] == 3.5
|
||||
assert market_data["current_price"] == 50000.0
|
||||
|
||||
# Portfolio data should include pnl
|
||||
assert "portfolio_pnl_pct" in portfolio_data
|
||||
assert "total_cash" in portfolio_data
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_scan_candidates_market_scoped(
|
||||
self, mock_broker: MagicMock, mock_market: MagicMock, mock_telegram: MagicMock,
|
||||
) -> None:
|
||||
"""Test scan_candidates uses market-scoped lookup, ignoring other markets."""
|
||||
from src.analysis.smart_scanner import ScanCandidate
|
||||
|
||||
engine = MagicMock(spec=ScenarioEngine)
|
||||
engine.evaluate = MagicMock(return_value=_make_hold_match())
|
||||
|
||||
# Candidate stored under US market — should NOT be found for KR market
|
||||
us_candidate = ScanCandidate(
|
||||
stock_code="005930", name="Overlap", price=100,
|
||||
volume=500000, volume_ratio=5.0, rsi=15.0,
|
||||
signal="oversold", score=90.0,
|
||||
)
|
||||
|
||||
with patch("src.main.log_trade"):
|
||||
await trading_cycle(
|
||||
broker=mock_broker,
|
||||
overseas_broker=MagicMock(),
|
||||
scenario_engine=engine,
|
||||
playbook=_make_playbook(),
|
||||
risk=MagicMock(),
|
||||
db_conn=MagicMock(),
|
||||
decision_logger=MagicMock(),
|
||||
context_store=MagicMock(get_latest_timeframe=MagicMock(return_value=None)),
|
||||
criticality_assessor=MagicMock(
|
||||
assess_market_conditions=MagicMock(return_value=MagicMock(value="NORMAL")),
|
||||
get_timeout=MagicMock(return_value=5.0),
|
||||
),
|
||||
telegram=mock_telegram,
|
||||
market=mock_market, # KR market
|
||||
stock_code="005930",
|
||||
scan_candidates={"US": {"005930": us_candidate}}, # Wrong market
|
||||
)
|
||||
|
||||
# Should NOT have rsi/volume_ratio because candidate is under US, not KR
|
||||
market_data = engine.evaluate.call_args[0][2]
|
||||
assert "rsi" not in market_data
|
||||
assert "volume_ratio" not in market_data
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_scenario_engine_called_without_scanner_data(
|
||||
self, mock_broker: MagicMock, mock_market: MagicMock, mock_telegram: MagicMock,
|
||||
) -> None:
|
||||
"""Test scenario engine works when stock has no scan candidate."""
|
||||
engine = MagicMock(spec=ScenarioEngine)
|
||||
engine.evaluate = MagicMock(return_value=_make_hold_match())
|
||||
playbook = _make_playbook()
|
||||
|
||||
with patch("src.main.log_trade"):
|
||||
await trading_cycle(
|
||||
broker=mock_broker,
|
||||
overseas_broker=MagicMock(),
|
||||
scenario_engine=engine,
|
||||
playbook=playbook,
|
||||
risk=MagicMock(),
|
||||
db_conn=MagicMock(),
|
||||
decision_logger=MagicMock(),
|
||||
context_store=MagicMock(get_latest_timeframe=MagicMock(return_value=None)),
|
||||
criticality_assessor=MagicMock(
|
||||
assess_market_conditions=MagicMock(return_value=MagicMock(value="NORMAL")),
|
||||
get_timeout=MagicMock(return_value=5.0),
|
||||
),
|
||||
telegram=mock_telegram,
|
||||
market=mock_market,
|
||||
stock_code="005930",
|
||||
scan_candidates={}, # No scanner data
|
||||
)
|
||||
|
||||
# Should still work, just without rsi/volume_ratio
|
||||
engine.evaluate.assert_called_once()
|
||||
market_data = engine.evaluate.call_args[0][2]
|
||||
assert "rsi" not in market_data
|
||||
assert "volume_ratio" not in market_data
|
||||
assert market_data["current_price"] == 50000.0
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_scenario_matched_notification_sent(
|
||||
self, mock_broker: MagicMock, mock_market: MagicMock, mock_telegram: MagicMock,
|
||||
) -> None:
|
||||
"""Test telegram notification sent when a scenario matches."""
|
||||
# Create a match with matched_scenario (not None)
|
||||
scenario = StockScenario(
|
||||
condition=StockCondition(rsi_below=30),
|
||||
action=ScenarioAction.BUY,
|
||||
confidence=88,
|
||||
rationale="RSI oversold bounce",
|
||||
)
|
||||
match = ScenarioMatch(
|
||||
stock_code="005930",
|
||||
matched_scenario=scenario,
|
||||
action=ScenarioAction.BUY,
|
||||
confidence=88,
|
||||
rationale="RSI oversold bounce",
|
||||
match_details={"rsi": 25.0},
|
||||
)
|
||||
engine = MagicMock(spec=ScenarioEngine)
|
||||
engine.evaluate = MagicMock(return_value=match)
|
||||
|
||||
with patch("src.main.log_trade"):
|
||||
await trading_cycle(
|
||||
broker=mock_broker,
|
||||
overseas_broker=MagicMock(),
|
||||
scenario_engine=engine,
|
||||
playbook=_make_playbook(),
|
||||
risk=MagicMock(),
|
||||
db_conn=MagicMock(),
|
||||
decision_logger=MagicMock(),
|
||||
context_store=MagicMock(get_latest_timeframe=MagicMock(return_value=None)),
|
||||
criticality_assessor=MagicMock(
|
||||
assess_market_conditions=MagicMock(return_value=MagicMock(value="NORMAL")),
|
||||
get_timeout=MagicMock(return_value=5.0),
|
||||
),
|
||||
telegram=mock_telegram,
|
||||
market=mock_market,
|
||||
stock_code="005930",
|
||||
scan_candidates={},
|
||||
)
|
||||
|
||||
# Scenario matched notification should be sent
|
||||
mock_telegram.notify_scenario_matched.assert_called_once()
|
||||
call_kwargs = mock_telegram.notify_scenario_matched.call_args.kwargs
|
||||
assert call_kwargs["stock_code"] == "005930"
|
||||
assert call_kwargs["action"] == "BUY"
|
||||
assert "rsi=25.0" in call_kwargs["condition_summary"]
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_no_scenario_matched_notification_on_default_hold(
|
||||
self, mock_broker: MagicMock, mock_market: MagicMock, mock_telegram: MagicMock,
|
||||
) -> None:
|
||||
"""Test no scenario notification when default HOLD is returned."""
|
||||
engine = MagicMock(spec=ScenarioEngine)
|
||||
engine.evaluate = MagicMock(return_value=_make_hold_match())
|
||||
|
||||
with patch("src.main.log_trade"):
|
||||
await trading_cycle(
|
||||
broker=mock_broker,
|
||||
overseas_broker=MagicMock(),
|
||||
scenario_engine=engine,
|
||||
playbook=_make_playbook(),
|
||||
risk=MagicMock(),
|
||||
db_conn=MagicMock(),
|
||||
decision_logger=MagicMock(),
|
||||
context_store=MagicMock(get_latest_timeframe=MagicMock(return_value=None)),
|
||||
criticality_assessor=MagicMock(
|
||||
assess_market_conditions=MagicMock(return_value=MagicMock(value="NORMAL")),
|
||||
get_timeout=MagicMock(return_value=5.0),
|
||||
),
|
||||
telegram=mock_telegram,
|
||||
market=mock_market,
|
||||
stock_code="005930",
|
||||
scan_candidates={},
|
||||
)
|
||||
|
||||
# No scenario matched notification for default HOLD
|
||||
mock_telegram.notify_scenario_matched.assert_not_called()
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_decision_logger_receives_scenario_match_details(
|
||||
self, mock_broker: MagicMock, mock_market: MagicMock, mock_telegram: MagicMock,
|
||||
) -> None:
|
||||
"""Test decision logger context includes scenario match details."""
|
||||
match = ScenarioMatch(
|
||||
stock_code="005930",
|
||||
matched_scenario=None,
|
||||
action=ScenarioAction.HOLD,
|
||||
confidence=0,
|
||||
rationale="No match",
|
||||
match_details={"rsi": 45.0, "volume_ratio": 1.2},
|
||||
)
|
||||
engine = MagicMock(spec=ScenarioEngine)
|
||||
engine.evaluate = MagicMock(return_value=match)
|
||||
decision_logger = MagicMock()
|
||||
|
||||
with patch("src.main.log_trade"):
|
||||
await trading_cycle(
|
||||
broker=mock_broker,
|
||||
overseas_broker=MagicMock(),
|
||||
scenario_engine=engine,
|
||||
playbook=_make_playbook(),
|
||||
risk=MagicMock(),
|
||||
db_conn=MagicMock(),
|
||||
decision_logger=decision_logger,
|
||||
context_store=MagicMock(get_latest_timeframe=MagicMock(return_value=None)),
|
||||
criticality_assessor=MagicMock(
|
||||
assess_market_conditions=MagicMock(return_value=MagicMock(value="NORMAL")),
|
||||
get_timeout=MagicMock(return_value=5.0),
|
||||
),
|
||||
telegram=mock_telegram,
|
||||
market=mock_market,
|
||||
stock_code="005930",
|
||||
scan_candidates={},
|
||||
)
|
||||
|
||||
decision_logger.log_decision.assert_called_once()
|
||||
call_kwargs = decision_logger.log_decision.call_args.kwargs
|
||||
assert "scenario_match" in call_kwargs["context_snapshot"]
|
||||
assert call_kwargs["context_snapshot"]["scenario_match"]["rsi"] == 45.0
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_reduce_all_does_not_execute_order(
|
||||
self, mock_broker: MagicMock, mock_market: MagicMock, mock_telegram: MagicMock,
|
||||
) -> None:
|
||||
"""Test REDUCE_ALL action does not trigger order execution."""
|
||||
match = ScenarioMatch(
|
||||
stock_code="005930",
|
||||
matched_scenario=None,
|
||||
action=ScenarioAction.REDUCE_ALL,
|
||||
confidence=100,
|
||||
rationale="Global rule: portfolio loss > 2%",
|
||||
)
|
||||
engine = MagicMock(spec=ScenarioEngine)
|
||||
engine.evaluate = MagicMock(return_value=match)
|
||||
|
||||
with patch("src.main.log_trade"):
|
||||
await trading_cycle(
|
||||
broker=mock_broker,
|
||||
overseas_broker=MagicMock(),
|
||||
scenario_engine=engine,
|
||||
playbook=_make_playbook(),
|
||||
risk=MagicMock(),
|
||||
db_conn=MagicMock(),
|
||||
decision_logger=MagicMock(),
|
||||
context_store=MagicMock(get_latest_timeframe=MagicMock(return_value=None)),
|
||||
criticality_assessor=MagicMock(
|
||||
assess_market_conditions=MagicMock(return_value=MagicMock(value="NORMAL")),
|
||||
get_timeout=MagicMock(return_value=5.0),
|
||||
),
|
||||
telegram=mock_telegram,
|
||||
market=mock_market,
|
||||
stock_code="005930",
|
||||
scan_candidates={},
|
||||
)
|
||||
|
||||
# REDUCE_ALL is not BUY or SELL — no order sent
|
||||
mock_broker.send_order.assert_not_called()
|
||||
mock_telegram.notify_trade_execution.assert_not_called()
|
||||
|
||||
Reference in New Issue
Block a user