Sydnee nightly — config_drift audit — 2026-05-31 P0 findings: 0 P1 findings: 4 Risks: 8 - Area: Config Drift (Sunday theme) - Bugs found (P0 / P1): 0 / 4 (0 new, 4 carried) - Risks noted: 9 (1 new, 8 carried) Cross-check: `git log --oneline -30` and `docs/strategy_decisions.md` reviewed before filing. `docs/audit_2026-05-25_config_drift.md` (last config-drift) and `docs/audit_2026-05-31_weekly.md` reviewed. No P1 or RISK items from previous audits were resolved in the intervening commits (last code commit `0f470d7` feat(scalp); all items confirmed still present in current code). --- Full report (dev branch): https://github.com/kanex1/sydnee.signals/blob/dev/docs/audit_2026-05-31_config_drift.md Reply FROM [email protected] to [email protected] to request fixes, e.g.: "code_task on sydnee-signals-dev: apply fix for the P0 about RVOL threshold in bot.py" Sydnee Agent will propose + you APPROVE (or plain 'approve') + auto-push to dev. --- Full audit below (first 12 KB) --- # Nightly Audit 2026-05-31 — CONFIG DRIFT ## Summary - Area: Config Drift (Sunday theme) - Bugs found (P0 / P1): 0 / 4 (0 new, 4 carried) - Risks noted: 9 (1 new, 8 carried) Cross-check: `git log --oneline -30` and `docs/strategy_decisions.md` reviewed before filing. `docs/audit_2026-05-25_config_drift.md` (last config-drift) and `docs/audit_2026-05-31_weekly.md` reviewed. No P1 or RISK items from previous audits were resolved in the intervening commits (last code commit `0f470d7` feat(scalp); all items confirmed still present in current code). --- ## Fixed Since Last Config-Drift Audit (2026-05-25) *(No config-drift items resolved. 36 code commits landed since 2026-05-25 — all feature or bug-fix work, none touching the carried config issues below.)* --- ## Findings ### BUG [P1] (carried from 2026-05-24): `TV_BXT_MAX_AGE_SEC` env var documented but never implemented **File:** `bot.py:3681` **Evidence:** ```python # bot.py:3681 — _tv_bxt_signal() docstring only """... 0 (none). Optionally ignore stale signals via TV_BXT_MAX_AGE_SEC.""" # grep "TV_BXT_MAX_AGE_SEC" bot.py → single hit, docstring only — no os.environ.get call ``` `tv_bxt_signal:{sym}` Redis key has no TTL. A stale Friday-close signal persists through the weekend. On Monday's first loop tick the bot acts on it before the first real webhook arrives. **Impact:** TV-BXt model (SNDK, real money since 2026-05-01) can enter based on a signal up to 65+ hours stale, potentially in the wrong direction vs. the Monday-gap open. **Fix:** Add age check in `_tv_bxt_signal()` using `tv_bxt_signal_ts:{sym}` Redis key, or set `EX` TTL on the webhook write at `bot.py:11832–11843`. --- ### BUG [P1] (carried from 2026-05-17): `REVERSE_TIME_DECAY` uses `== "1"` — ignores `true`/`yes` **File:** `bot.py:5749` **Evidence:** ```python # bot.py:5749 — sole check point for the flag if os.environ.get("REVERSE_TIME_DECAY", "0") == "1" and _in_rth: ``` Every other boolean env flag in bot.py uses `.lower() in ("1", "true", "yes")`. `REVERSE_TIME_DECAY=true` (YAML-style compose) silently has no effect. **Impact:** If dev docker-compose is edited to `REVERSE_TIME_DECAY: "true"`, the morning size ramp is silently disabled with no diagnostic. **Fix:** Change to `os.environ.get("REVERSE_TIME_DECAY", "").lower() in ("1", "true", "yes")`. --- ### BUG [P1] (carried from 2026-05-17): `daily_loss_limit` equals `capital` — DLL disabled **File:** `config.json:3,6`; `core/risk.py:105` **Evidence:** ```json { "capital": 50000, "daily_loss_limit": 50000 } ``` ```python # core/risk.py:105 — can_trade() gate if self.state.realized_pnl <= -self.cfg.daily_loss_limit: return False, ... ``` DLL fires only after a full $50,000 intraday loss — 100% drawdown. Bot is live on real money. The DLL is also the backstop that the TV-BXt SHIELD exemption (`8535f8e`) relies on. **Fix:** `"daily_loss_limit": 1500` (3% of $50K). Also set `core/risk.py:RiskConfig.daily_loss_limit` dataclass default to match (currently `1_500.0`, which would be correct — just fix config.json). --- ### BUG [P1] (carried from 2026-05-31 weekly): `TV_BXT_TRAIL_AGAINST` default mismatch — live=`1.0` vs history endpoint=`1.5` **File:** `bot.py:3750` vs `bot.py:14062` **Evidence:** ```python # bot.py:3750 — _obv_trail_stop_hit() LIVE TRADING path: agnst = float(os.environ.get("TV_BXT_TRAIL_AGAINST", "1.0")) # bot.py:14062 — api_trade_obv_trail() history REPLAY endpoint: agnst = float(os.environ.get("TV_BXT_TRAIL_AGAINST", "1.5")) ``` When `TV_BXT_TRAIL_AGAINST` is unset (the default path), the live engine trails at 1.0×ATR against-trend, but the OBV-trail history endpoint reconstructs at 1.5×ATR. Every closed-trade stop timeline shown on the dashboard reflects a 50% wider trail than what actually ran. **Impact:** All historical stop-analysis on trade cards is wrong when the env var is unset. Post-trade reviews ("was the stop too tight?") draw wrong conclusions. **Fix:** Change `bot.py:14062` from `"1.5"` to `"1.0"`. One-line fix. --- ### RISK (new): `TV_BXT_NLV_PCT = 50.0` sizing bypass absent from `strategy_decisions.md` **File:** `bot.py:6885–6898`; `docs/strategy_decisions.md` **Evidence:** ```python # bot.py:6885-6898 — TV-BXt sizing override # Bypasses MAX_POSITION_PCT and all earlier multipliers. _tv_pct = float(os.environ.get("TV_BXT_NLV_PCT", "50.0")) _tv_shares = int((nlv * _tv_pct / 100.0) / price) ``` `TV_BXT_NLV_PCT = 50.0` (default) sizes each TV-BXt entry to 50% of NLV — $25,000 per trade on a $50K account. This explicitly bypasses `MAX_POSITION_PCT` (10% cap), ATR-risk sizing, and all conviction multipliers. Only the `buying_power_pct` margin cap (bot.py:6913) and the DLL SHIELD remain active. No entry exists in `docs/strategy_decisions.md` for this sizing decision. The quarantine guard was raised 25%→55% (`a4a5277`) to accommodate it, but the underlying `TV_BXT_NLV_PCT=50.0` decision — bypassing the normal 10% NLV position cap — has no structured hypothesis, backtest, or validation record. **Impact:** If `TV_BXT_NLV_PCT` is accidentally left at default in a fresh staging reprovisioning, or if SNDK adds a second symbol to `TV_BXT_SYMBOLS`, the bot can open two $25K positions simultaneously on a $50K account. No post-trade review record exists to catch if this sizing is underperforming vs. the ATR-risk approach. **Fix:** Add a `strategy_decisions.md` entry for the TV-BXt 50% NLV sizing: hypothesis, approval date, expected outcome, and validation criteria. --- ### RISK (carried from 2026-05-25): `config.json:max_concurrent_positions: 50` disables concurrent-position gate **File:** `config.json:24`; `core/risk.py:114`; `bot.py:6781` **Evidence:** ```json "max_concurrent_positions": 50 ``` ```python # bot.py:6781 — gate in live path allowed, reason = self.risk.can_trade() # core/risk.py:114 — fires only when open_positions >= 50 ``` With 8 symbols in the watchlist, maximum realistic concurrent positions = 8–12. Gate never fires. The concurrent-position circuit breaker is effectively off on real money. **Fix:** `"max_concurrent_positions": 5` (or 8 at most for an 8-symbol book). --- ### RISK (carried from 2026-05-25): `config.json:risk_per_trade_pct: 2.0` bypassed by env-var sizing path **File:** `config.json:7`; `core/risk.py:19`; `bot.py:6179–6210` **Evidence:** ```json "risk_per_trade_pct": 2.0 ``` ```python # bot.py:6179-6210 — live sizing path (not config.json): sym_risk_pct = max(1.0, min(3.0, float(os.environ.get("RISK_ATR_BUDGET", "12.0")) / daily_atr_pct)) risk_amount = min(risk_amount, nlv * float(os.environ.get("MAX_RISK_PCT", "0.015"))) ``` Config.json value is read only by the backtester (`core/risk.py:position_size()`). Live sizing reads `RISK_ATR_BUDGET` and `MAX_RISK_PCT` env vars — not `risk_per_trade_pct`. Three sources, three values: `config.json` = 2.0%, `core/risk.py` default = 1.5%, live `MAX_RISK_PCT` default = 1.5%, dev override = 3.0%. **Impact:** An operator reading `config.json` to understand per-trade risk sees 2.0% — but actual live cap is the `MAX_RISK_PCT` env var (default 1.5%). **Fix:** Either (a) add a comment to `config.json` noting `risk_per_trade_pct` is backtester-only and live sizing is env-driven, or (b) wire live sizing to read this value as the base. --- ### RISK (carried from 2026-05-24): Oracle `HEALTH_PORT` and bot `DASH_PORT` both default to `8083` **Files:** `oracle/config.py:56`; `bot.py:184`; `Dockerfile.oracle:12`; `Dockerfile:20` ```python # oracle/config.py:56 HEALTH_PORT = int(os.environ.get("ORACLE_HEALTH_PORT", "8083")) # bot.py:184 DASH_PORT = int(os.environ.get("DASH_PORT", "8083")) ``` Both services `EXPOSE 8083`. Silent dashboard outage if oracle gets the host-8083 binding first on a compose-with-ports deploy. **Fix:** Change `ORACLE_HEALTH_PORT` default to `8084` in `oracle/config.py:56` and `Dockerfile.oracle:12`. --- ### RISK (carried from 2026-05-24): `config.json:atr_stop_multiplier` (1.5) diverges from code fallback (2.0) **Files:** `config.json:8`; `bot.py:750` ```python # bot.py:750 self.atr_stop_mult = self.cfg.get("atr_stop_multiplier", 2.0) ``` Removing or misspelling the config key silently widens stops by 33% with no log or alert. **Fix:** Add an inline comment to `config.json` documenting that the calibrated value is 1.5 and the code fallback is 2.0, to flag the divergence clearly on review. --- ### RISK (carried from 2026-05-25): `pages.py:1448` hardcodes `"TradingView webhook (SNDK only)"` model label **File:** `pages.py:1448` ```javascript var nm = { ..., tv: 'TradingView webhook (SNDK only)' }[m]; ``` `TV_BXT_SYMBOLS` env controls the actual symbol list; the dashboard label is static. If a second symbol is added to `TV_BXT_SYMBOLS`, the label goes stale immediately. **Fix:** Pass `TV_BXT_SYMBOLS` value through a Jinja2 template variable into the JS map. --- ### RISK (carried from 2026-05-17): `sentiment.cashtags` seed mismatches trading `symbols` **File:** `config.json:2,37` ```json "symbols": ["WDC", "STX", "MU", "SNDK", "APP", "META", "TSLA", "PLTR"], "cashtags": ["APP", "KORU", "MU", "SNDK", "STX", "WDC"] ``` `META`, `TSLA`, `PLTR` are traded but absent from the cashtag seed — sentiment bypassed for ~15 min post-startup on those symbols. `KORU` is in the seed but not traded. **Fix:** ```json "cashtags": ["APP", "META", "MU", "PLTR", "SNDK", "STX", "TSLA", "WDC"] ``` --- ### RISK (carried from 2026-05-25): ~40 undocumented feature-flag env vars **Files:** `bot.py` (multiple); `oracle/config.py`; no `.env.example` 30+ env vars flagged in 2026-05-25 audit have grown by 9 more this week (`TV_BXT_DISABLE_ALGO_STOPS`, `TV_BXT_FLOW_CONFIRM`, `TV_BXT_FLOW_RATIO_MIN`, `TV_BXT_FLOW_MIN`, `TV_BXT_VOL_MIN`, `TV_WEBHOOK_ALLOWED_IPS`, `SCALP_SIGNAL_LOG`, `TAPE_SCANNER_SYMBOLS`, `SCALP_SIGNAL_LOG_SEC`). None are in `.env.example` (no such file exists). A fresh staging reprovision would silently default all of these, breaking the configured dev A/B stack. `TV_BXT_DISABLE_ALGO_STOPS` (a stop-loss kill switch) is particularly dangerous if accidentally left set — no operator reading CLAUDE.md or a `.env.example` would know it exists. **Fix:** Add `.env.example` listing all strategy flags with expected dev/prod values. Highlight `TV_BXT_DISABLE_ALGO_STOPS` as a kill switch, not a tuning parameter. --- ## OK (checked, working or recently fixed) - **RVOL blowoff ceiling:** CLAUDE.md RVOL table correctly shows `> 5.0x` ceiling for RSI-extreme skip (updated after 2026-04-15 strategy decision; code at `bot.py:6027` matches). ✓ - **CLAUDE.md Entry Time Blockers table:** Correctly reflects close-hour BXt hard block and off-hours RSI-extreme toggle (default OFF). Both entries now present since 2026-05-19 correction. ✓ - **Off-hours RSI extreme toggle Redis default:** `_is_rsi_extreme_offhours_enabled()` returns `False` when Redis key absent (`bot.py:10527`). Correct fail-safe behavior. ✓ - **`TV_BXT_DISABLE_ALGO_STOPS` boolean handling:** Uses `.lower() in ("1", "true", "yes")` (`bot.py:3967`). Consistent with project convention. ✓ - **New scalp/tape scanner env vars:** `SCALP_SIGNAL_LOG` and related flags use `.lower() not in ("1", "true", "yes")` / `.lower() in (...)` patterns correctly. ✓ - **`TV_BXT_FLOW_CONFIRM` boolean handling:** Uses `.lower() in ("1", "true", "yes")` (`bot.py:4000`). Consistent. ✓ - **`TV_BXT_NLV_PCT` buying_power_pct backstop:** After the sizing override at `bot.py:6891–6898`, the margin check at `bot.py:6913` (`max_cost = bp * self._buying_power_pct`) still caps shares. TV-BXt sizing is bounded by available buying power; not a fully unconstrained path. ✓ - **Oracle `REDIS_MIRROR` guard:** Gated on non-empty var; no crash when unset. ✓ - **`signal_tf: 5` / `bias_tfs: [15]`:** Match bot.py defaults. ✓ - **`rsi_upper: 65` / `rsi_lower: 35`:** Match bot.py defaults. ✓ - **`TV_BXT_SECRET` reject-empty guard:** `if not expected: abort(403)` (`bot.py:11839`). ✓ - **`BXT_USE_15M_BIAS` default `"true"`:** Matches strategy