Commit Graph
21 Commits
Author SHA1 Message Date
genichinandClaude Opus 4.8 b851b927ee fix(ledger): dcy-015 outbox 영속 오류 분류 + 백오프/dead-letter (BAT-019)
portfolio-hub 404(미등록 포트폴리오)가 retryable 로 분류되어 백오프·상한 없이
초당 영구 재시도되던 버그(issue #13, 실측 930회) 정정.

거버넌스: 전부 change-req (신규 REQ 없음, dcy-015 OQ-3; 백오프 env 는 REQ-019 흡수).
- REQ-017: 404/400/422 → LedgerPermanentError 분류 (409 conflict / 408·429·5xx·
  network retryable 불변). LedgerPort 에 permanent 타입 추가
- REQ-019: 영속 → 터미널 delivery_failed 즉시(재시도 0). retryable 지수 백오프
  +jitter+next_retry_at, max_attempts 초과 시 dead-letter delivery_failed.
  list_pending 이 백오프 행 스킵(head-of-line 보호). 마이그레이션 007(status enum
  +next_retry_at). 백오프 env(base/cap/max-attempts)
- REQ-018: 터미널/dead-letter WARN + WriterStats.delivery_failed 카운터 가시화
- REQ-020: runtime-flows/data-model/interface-contract 정합

단일 터미널 delivery_failed(permanent/dead-letter, last_outcome 로 사유 구분) —
conflict_blocked 패턴과 동형. dead-letter 는 운영자 재처리 가능 상태.

검증: 단위 311 passed/34 skipped, 인프라 게이트(PG+RabbitMQ) 345 passed, ruff clean.
신규 test_ledger_outbox_backoff + adapter/PG 테스트, 기존 retryable 백오프 정합.
BAT-019 release-candidate.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-08 16:12:17 +09:00
genichinandClaude Opus 4.8 0b7db66463 feat(runtime): dcy-014 trading_date KST 타임존 분리 + 일일 슬롯 preallocation (BAT-018)
데몬이 시스템 UTC 날짜로 슬롯을 1회만 preallocate 해 trading_date 가 KST 캘린더와
어긋나고 재시작 없이는 다음날 슬롯이 안 생기던 문제(issue #14) 정정.

거버넌스: 전부 change-req (신규 REQ 없음, dcy-014 OQ-3; 스케줄러는 REQ-030 패턴 준용).
- REQ-005: trading_date 를 설정 타임존(기본 Asia/Seoul) 기준 계산(timestamp 는 UTC
  유지). DailyPreallocationScheduler 신규 — TZ 일자당 1회 멱등 fire, 재시작 없이
  다음날 슬롯 자동 생성. setup() 배선 + trex-preallocation 스레드
- REQ-048: TOML trading_timezone/daily_preallocation_time 키 + from_config 전파
- REQ-009: runtime-flows/data-model 정합 + instance.example.toml

OQ-1 기본 00:00(TZ 자정, TOML 설정 가능) / OQ-2 매 캘린더일 생성(거래일 캘린더 후속).
멱등 preallocate 로 중복 tick·다중 인스턴스 중복 생성 0.

검증: 단위 302 passed/32 skipped, 인프라 게이트(PG+RabbitMQ) 334 passed, ruff clean.
신규 test_preallocation_scheduler(고정 clock) + test_instance_config 확장.
BAT-018 release-candidate.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-08 15:39:04 +09:00
genichinandClaude Opus 4.8 4746f74aae fix(broker): dcy-013 KIS 외부 취소 감지 + 부분 체결 회계 + stale 워치독 (BAT-017)
KIS 계좌에서 직접 매도 주문을 취소하면 poll_fill 이 cncl_yn 을 파싱하지 않아
슬롯이 pending_sell 에 영구히 갇히고 죽은 주문을 무한 폴링하던 버그(issue #9)
정정.

거버넌스: 전부 change-req (신규 REQ 없음, dcy-013 OQ-3).
- REQ-012: KIS poll_fill cncl_yn 파싱 → 외부 취소 감지. FillStatus.cancelled
  /is_partial_cancel 추가. 체결 0 취소→cancelled, 부분 체결 후 취소→partial+cancelled
- REQ-013: 집행 _on_partial_cancel — 체결분 원장 1회 + qty 차감(매도 잔여/매수 체결분)
  + holding 복귀 (OQ-1 A안). 전량 취소는 기존 롤백 경로 재사용
- REQ-018: 부분 체결분 원장 1회 기록 (_record_ledger 재사용)
- REQ-031: 잔여 포지션 정합 연계 (slot.qty ↔ 실 잔고, 과매도 방지)
- REQ-014: pending stale 워치독 Alert-only (TREX_PENDING_STALE_THRESHOLD,
  WARN+카운터, 상태 미변경) — 취소 감지 누락 backstop
- REQ-015: interface-contract/runtime-flows 정합

전량 취소는 holding 자동 복구(limit_price clear→재매도 시장가, dcy-012 연계).
KIS cncl_yn 실 규약은 운영 검증(미지 값 보수 처리 + 워치독 backstop).

검증: 단위 294 passed/32 skipped, 인프라 게이트(PG+RabbitMQ) 326 passed, ruff clean.
신규 test_kis_cancel(httpx stub) + test_execution_cancel(programmable broker).
BAT-017 release-candidate.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-08 14:11:49 +09:00
genichinandClaude Opus 4.8 cc28e1fba5 fix(execution): dcy-012 매도 가격 버그 정정 — slot entry_price→limit_price rename (BAT-016)
매도(sell) 주문이 현재가가 아닌 매수가(entry_price) 지정가로 나가던 버그(issue #10)
정정. 슬롯의 매수 전용 entry_price 를 side 무관 "현재 pending 주문 지정가"
limit_price 로 내부 rename 하고, pending 이탈(→holding/→closed) 시 clear 해
매수 지정가가 매도로 누수되는 경로를 구조적으로 제거.

거버넌스: 전부 change-req (신규 REQ 없음, dcy-012 OQ-3).
- REQ-002: Slot entry_price→limit_price rename + clear-on-exit 생애주기
- REQ-006: daily_slot 컬럼 rename 마이그레이션 006 (idempotent) + persistence 매핑
- REQ-013: _dispatch 가 slot.limit_price 단일 필드로 가격/주문유형 결정
- REQ-007: SellIntent/sell payload optional limit_price + _admit_sell 운반
- REQ-036: SlotService.sell optional 가격 → 자율 매도(safety-sell/kill-switch) 시장가
- REQ-003: 컬럼 계약/드리프트 가드 limit_price 반영
- REQ-015: runtime-flows/interface-contract/data-model 정합

매도 정책(OQ-1 하이브리드): 가격 지정→지정가("00"), 미지정→시장가("01").
자율 매도는 가격 미전달 → 항상 시장가 (손절 미체결 pending_sell 갇힘 방지, issue #9 연계).

검증: 단위 286 passed/32 skipped, 인프라 게이트(PG+RabbitMQ) 318 passed, ruff clean.
신규 test_sell_pricing(recording broker)로 지정가 매수 슬롯의 매도가 매수가
재사용 안 함을 직접 검증. BAT-016 release-candidate.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-08 13:41:46 +09:00
genichinandClaude Opus 4.8 aee3512383 feat(intents): dcy-011 buy intent portfolio_id 필수 계약 + ledger skip 가시화 (BAT-015)
dcy-011 전 주기 delivery. issue #6(buy 식별자 정정) + issue #11(ledger silent skip) 흡수.
거버넌스: 기존 Implemented REQ는 change-req, 신규 기능만 신규 REQ(OQ-3).

- REQ-007 (change-req): buy admission portfolio_id 필수화 + SnapshotStore symbol
  resolve + unknown_portfolio 거부 (hard cutover, OQ-1/OQ-2)
- REQ-021 (change-req): buy 메시지 payload symbol 제거
- REQ-018 (change-req): _record_ledger skip 가시화 (WARN + ledger_skipped 카운터)
- REQ-055 (신규): closed↔ledger_outbox anti-join 운영 점검 find_ledger_gap_slots

코드: intents.py(BuyIntent portfolio_id 필수 + resolve_symbol 주입 + _admit_buy
reject 분기), messaging_handlers.py, runtime/daemon.py(_resolve_symbol 배선),
execution.py(skip WARN 로그), adapters/persistence/ops_checks.py(신규 anti-join).

검증: 단위 282 passed/32 skipped, 인프라 게이트(PG+RabbitMQ) 314 passed, ruff clean.
BAT-015(standard) release-candidate. srs/index Next ID → REQ-056.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-08 12:57:51 +09:00
genichinandClaude Opus 4.8 565af6792c feat(safety-sell): dcy-010 BAT-014 — 영속 거부 재시도 쿨다운 (REQ-052~054)
dcy-010(safety-sell 영속 거부 재시도 백오프) 전체 SDLC: draft-req → confirm-req →
draft-batch(BAT-014 batch-lite) → run-batch-delivery → confirm-req-implemented.
2026-06-07 실 KIS LIVE SL/TP e2e에서 발견한 KIS 주문 spam 위험(영속 거부 시
매도 무한 재제출) 차단.

- REQ-052 (Integration): SafetySellService per-slot 인메모리 쿨다운
  (_in_cooldown/_attempt_sell/_prune_cooldown) — forced_close·stop-loss·take-profit
  트리거에 min_retry_interval_sec(600s) 이내 재시도 skip(가격 poll 포함) +
  max_retry_warn_count(3) 도달 시 WARNING + holding 이탈 정리. market_close 제외.
- REQ-053 (Configuration): SafetySellConfig min_retry_interval_sec/max_retry_warn_count
  + TREX_SAFETY_SELL_MIN_RETRY_INTERVAL/MAX_RETRY_WARN_COUNT env (additive)
- REQ-054 (Documentation): runtime-flows 쿨다운 flow + 가격체크 tr_id(FHKST01010100) 정정

인메모리 only(DB/스레드/인터페이스 변경 없음), default 미설정 시 기존 동작 불변.
검증: 신규 8 단위(쿨다운 6 + config 2), 전체 278 passed, ruff clean.
REQ-052~054 Implemented, BAT-014 release-candidate.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-08 00:35:11 +09:00
genichinandClaude Opus 4.8 9f83c64a40 feat(multi-instance): dcy-009 BAT-013 — RabbitMQ 다중 인스턴스 토폴로지 (REQ-048~051)
dcy-009(다중 인스턴스 RabbitMQ 토폴로지) 전체 SDLC: draft-req → confirm-req →
draft-batch(BAT-013) → run-batch-delivery → confirm-req-implemented.

- REQ-048 (Configuration): DaemonInstanceConfig.from_toml + DaemonConfig.from_config
  (env 병합, portfolio_hub_token 주입) + 데몬 --config 필수 argparse + .gitignore 인스턴스
  TOML + config/instance.example.toml
- REQ-049 (Interface): topology 동적 큐명 빌더 intent/snapshot/control_queue(instance_key)
  — default=리터럴 큐명(하위호환), named=`{base}.{instance_key}` + daemon per-instance 구독
- REQ-050 (Integration): trex.direct(topic)/trex.broadcast(fanout) exchange +
  declare_topology(instance_key) per-instance binding + control dual binding + direct_routing_key
- REQ-051 (Documentation): interface-contract/runtime-flows/project-structure 정합

하위호환: instance_key="default" → 기존 리터럴 큐명·RPC 경로 무회귀(FR-5).
split-brain 방지(NFR-3): 인스턴스 전용 큐. 민감값(NFR-4): TOML + gitignore.

검증: 신규 9 단위 + RabbitMQ 라우팅 2, 단위 271 passed,
인프라 게이트 301 passed(directed 격리 S-2 + broadcast dual S-3 + default 무회귀 S-4),
ruff clean. REQ-048~051 Implemented, BAT-013 release-candidate.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-07 15:10:36 +09:00
genichin 7e23648489 feat: add kill-switch control flow 2026-06-07 14:00:20 +09:00
genichinandClaude Opus 4.8 a30efa8f3b feat(safety-sell): dcy-007 core 자율 안전 매도 — BAT-010 기반 + BAT-011 트리거
dcy-007(core 자율 안전 매도) 전체 delivery. BAT-010 기반 인프라가 미커밋
상태였고 BAT-011 트리거 레이어와 공유 파일에서 섞여 있어 한 커밋으로 묶음.

BAT-010 (REQ-034/035/036/041, release-candidate):
- WatchdogStore(application/watchdog.py) + snapshot RPC 워치독 갱신
- Slot stop_loss_price/take_profit_price 필드 + 004_add_price_thresholds.sql
- SlotService.sell() 자율 매도 내부 경로 + idempotency
- SafetySellConfig env 5종 파싱

BAT-011 (REQ-037/038/039/040/042, release-candidate):
- SafetySellService(application/safety_sell.py): forced_close 만기 / stop-loss·
  take-profit KIS polling / 장 마감 강제 청산 + trex-safety-sell 7번째 스레드
- BrokerPort.get_current_price + KIS inquire-price(TTTC8397R/VTTC8397R, stck_prpr)
  + paper stub
- SlotService.update_thresholds() + intent RPC update_thresholds 액션
  (부분 갱신/null 해제, 워치독 last_seen_at 리셋)
- baseline 3문서 정합(interface-contract/runtime-flows/project-structure)

거버넌스: REQ-034~042 Implemented 동기화, BAT-010/011 release-candidate,
dcy-007 discovery 포함.

검증: 신규 38 테스트 green, 단위 236 passed/23 skipped,
인프라 게이트(PG+RabbitMQ) 259 passed(마이그레이션 001~004 round-trip 포함),
ruff clean.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-07 08:32:30 +09:00
genichinandClaude Opus 4.8 c8174e8bb6 docs(release): REL-001 released → feedback-captured (BAT-001~008 released)
운영자(Hermes Agent) 실 환경 배포 evidence(revision 71faa23, host 86ba33f2997a,
broker=kis LIVE, 2026-06-06T11:18:16Z) 반영:
- Release Execution Evidence 기재(배포 facts + app-service profile checks +
  동일 revision 71faa23 pre-deploy 게이트 재현 보강).
- Rollback Outcome 의미 정정: 운영자 보고 "rollback yes"는 배포 롤백이 아니라
  휴장일(현충일) KIS 거절에 따른 slot pending_buy→idle 정상 복구(매수 거부 경로 smoke 성공).
- capture-release-feedback: release feedback-captured 전환,
  포함 batch BAT-001~008 released 전환.
- 잔여(차기 거래일 실증): ledger flush(fill→ledger), 마이그레이션 명시 로그,
  포지션 정합/warning 마스킹 결과 로그 — Feedback Handoff 기록.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-07 00:26:08 +09:00
genichinandClaude Opus 4.8 2f57aad7b2 feat(config): expose paper broker tuning via env (REQ-010 AC-5, BAT-009)
Wire the existing PaperBroker scenario knobs (fill_after_polls /
qty_per_poll / reject_symbols) to daemon env so REQ-011 AC-2 e2e
scenarios (delayed / partial / reject) can be reproduced by a daemon
restart. Unset == prior PaperBroker() behaviour, byte-for-byte (no
regression).

- config.py: PaperBrokerConfig + from_env parsing (TREX_PAPER_*)
- daemon.py: _build_broker injects config.paper into PaperBroker
- .env.example: document the 3 paper tuning keys (paper-only, e2e)
- tests: paper broker defaults + env tuning parsing

SDLC: REQ-010 CHG-20260606-01 (AC-5 added) / -02 (Implemented);
BAT-009 batch-lite release-candidate; REL-001 feedback handoff records
pre-deployment e2e observations incl. event-publisher idle-drop and
paper correlation_id collision (GitHub #4, #5).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-06 16:49:51 +09:00
genichin ca757009f5 feat: add portfolio snapshot read layer 2026-06-06 09:33:00 +09:00
genichin 6c7a1cf6f6 feat: add messaging transport rollout 2026-06-06 00:02:52 +09:00
genichin aed54911d4 chore: commit current workspace changes 2026-06-04 23:38:42 +09:00
genichinandClaude Opus 4.8 01a51d0c16 Advance dcy-004 ledger slice: confirm Discovery, draft+approve REQ-017~020, draft BAT-006
- confirm-discovery dcy-004: draft→confirmed (Confirmed By: Justin Ko),
  OQ-1~3 resolved, scope/R-1 accepted, portfolio-hub contract deferred to REQ/delivery
- draft-req: REQ-017 (LedgerPort + portfolio-hub adapter + stub), REQ-018
  (fill→ledger recording + FillStatus exec_id/fee), REQ-019 (ledger outbox
  durability + correlation_id idempotency), REQ-020 (baseline alignment);
  NFR-1~3 absorbed
- confirm-req: REQ-017~020 Proposed→Approved
- draft-batch: BAT-006 (standard) bundling REQ-017~020 from dcy-004
- Update SRS/Discovery/batches indexes

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-04 22:32:27 +09:00
genichinandClaude Opus 4.8 ac6061f510 Deliver BAT-005: KIS token Redis cache (REQ-016, minor-change)
Port the trex-orbit Redis token-cache pattern so the KIS access token
survives daemon restarts and is shared across processes, avoiding KIS
token-issuance rate limits. KISBroker._ensure_token now resolves L1
in-memory -> L2 cache -> issue, storing issued tokens back to the cache.

- adapters/broker/token_cache.py: TokenCache protocol + RedisTokenCache
  (key trex:broker:kis:{account_ref}:token, token+expiry+app_key suffix
  JSON, ex=ttl, best-effort with graceful fallback) + InMemoryTokenCache.
- kis.py: optional token_cache injection.
- runtime config/daemon: TREX_KIS_REDIS_URL wires a RedisTokenCache.
- pyproject: redis>=5.

Verification: 96 passed, ruff clean — cache hit avoids issuance, cache is
shared across instances (restart/multi-process survival), backend errors
fall back to issuance (no regression), and app_key-suffix mismatch
invalidates a stale cache. Tested deterministically with an in-memory
double (no Redis server required). Processed via the minor-change fast
path: REQ-016 -> Implemented, BAT-005 (batch-lite) -> release-candidate.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-04 18:44:09 +09:00
genichinandClaude Opus 4.8 97d58318db Deliver BAT-004: broker execution + KIS adapter + durable handoff
Implement REQ-011~015 (dcy-003): the order-execution layer on top of the
BAT-002 slot-core daemon, connecting intents to real orders.
- application: BrokerPort (submit/poll_fill/cancel) + ExecutionService
  (claim -> submit -> poll -> transition -> rollback -> reconcile)
- adapters/broker: PaperBroker (deterministic async double) and KISBroker
  (KIS Open API: token, order-cash, inquire-daily-ccld, cancel; VTS sandbox
  tr_ids; live account gated behind allow_live)
- persistence: option-A execution queue (find_pending_for_dispatch FOR
  UPDATE SKIP LOCKED, mark_dispatched idempotency guard, clear_dispatch)
- runtime: background execution loop in the daemon lifespan + startup
  reconcile; config selects paper|kis

Verification (real KIS VTS 모의투자 + PostgreSQL): 92 passed, ruff clean,
domain import sweep shows zero broker/web/db deps, buy->holding /
sell->closed / reject->idle, idempotent dispatch (exactly one order),
reconcile after crash creates no duplicate order, KIS sandbox token+submit
round-trip, and a daemon E2E through the background loop. Align the four
baseline docs. BAT-004 -> release-candidate; REQ-011~015 -> Implemented.

KIS live-account go-live, partial-fill accrual, option-B outbox, connection
pooling, and the safety floor remain deferred to follow-up Discoveries.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-04 17:46:17 +09:00
genichinandClaude Opus 4.8 575ba9f513 Deliver BAT-003: daemon .env loading (REQ-010, minor-change)
DaemonConfig.from_env() now best-effort loads a .env file (find_dotenv
usecwd, load_dotenv override=False) so operators can manage daemon env
vars via .env, with real OS env taking precedence. Add python-dotenv
dependency, gitignore .env, ship .env.example, and add config tests
(load / OS-env precedence / absence regression / missing-DSN guard).

Verification: 78 passed, ruff clean, .env ignored + .env.example tracked,
and a .env-only DaemonConfig smoke. Processed via the minor-change fast
path: REQ-010 (Configuration) -> Implemented, BAT-003 (batch-lite,
no design) -> release-candidate.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-04 14:56:42 +09:00
genichinandClaude Opus 4.8 a4e263c8e7 Deliver BAT-002: slot-core daemon + HTTP intent API + PostgreSQL
Implement REQ-005~009 (dcy-002 slice 1) as an onion stack on top of the
BAT-001 pure domain:
- application: IntentService (admission + single-writer transition +
  intent_id idempotency via durable intent_log)
- adapters/persistence: PostgreSQLSlotRepository + PostgreSQLIntentStore +
  trex-core-owned migrations (daily_slot per Appendix A, daily_symbol_entries,
  slot_transition_log, intent_log) + in-memory intent store
- adapters/api: FastAPI /v1 (healthz, intents/buy|sell|{id}, slots, capacity)
- runtime: trex-core-daemon entrypoint with lifespan migrate+preallocate

Verification (against local PostgreSQL 18): 74 passed, ruff clean, domain
import sweep shows zero web/db deps (NFR-1), 20-thread concurrency admits
exactly the idle count (NFR-2), idempotent resubmit causes no double
transition (NFR-3), EXPECTED_SLOT_FIELDS<->information_schema drift test
(realizes dcy-001 OQ-2's deferred DB-side check), and a daemon E2E smoke.

Align all four baseline docs to the new daemon/API/persistence layers.
BAT-002 -> release-candidate; REQ-005~009 -> Implemented. Full async outbox,
broker execution, safety floor, and strategy daemons remain deferred to a
follow-up Discovery.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-04 13:46:08 +09:00
genichinandClaude Opus 4.8 20498aa4ce Deliver BAT-001: bootstrap trex-core + port slot pure domain
Implement REQ-001~004 (dcy-001): scaffold the trex-core package
(pyproject, trex_core/, tests/), port the slot pure domain into
trex_core.domain.slot with import rewrites, add the EXPECTED_SLOT_FIELDS
contract guard (test_contract), and port the state/service unit tests.

Verification: 56 passed, ruff clean, import sweep shows zero external
deps, public __all__ exposes no infra symbols, contract negative case
reproduced. Align project-structure/runtime-flows baselines to the real
tree. Sync SDLC tracking: REQ-001~004 → Implemented, BAT-001 →
release-candidate.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-04 00:16:07 +09:00
genichinandClaude Opus 4.8 b40e569537 Bootstrap StagePilot baseline for slot domain migration
Initialize baseline seed, four baseline docs (project-structure,
runtime-flows, interface-contract, data-model), and the active
discovery/srs/batches/releases indexes. Content derived from the
trex-orbit slot pure-domain migration plan (trex-core as a new base,
pure domain only).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-03 09:47:14 +00:00