fix: resolve fast-unit-tests CI collection errors and stale config assertions

.gitignore

@@ -636,3 +636,5 @@ chronos2_long
 profile_output/
 deployments/
 binanceneural/profile_trace.json
+AGENTS.md
+.tmp_eval/

README.md

@@ -1,4 +1,8 @@
 # stock-prediction
+this is a collection of experiments training AI to trade.
+
+i reccomend you use my stats dashboard, forecasting platform and API [bitbank.nz](https://bitbank.nz) for cryptocurrency trading.
+
 
 A production RL trading system for crypto and equities. The core loop is:
 fast C market environment + PPO policy training + Chronos2 time-series
@@ -7,7 +11,7 @@ forecasting + live execution on Binance (FDUSD/USDT) and Alpaca (US stocks).
 ## Architecture
 
 ```
-Raw OHLCV data (Binance / Alpaca / yfinance)
+Raw OHLCV data (Binance / Alpaca)
         |
         v
   Binary export (export_data_*.py)

RLgpt/__init__.py

@@ -0,0 +1,20 @@
+"""RLgpt daily-plan trading experiment."""
+
+from .config import DailyPlanDataConfig, PlannerConfig, SimulatorConfig, TrainingConfig
+from .data import DailyPlanTensorDataset, DailyPlanTensors, TensorNormalizer, prepare_daily_plan_tensors
+from .model import CrossAssetDailyPlanner
+from .simulator import compute_trading_objective, simulate_daily_plans
+
+__all__ = [
+    "CrossAssetDailyPlanner",
+    "DailyPlanDataConfig",
+    "DailyPlanTensorDataset",
+    "DailyPlanTensors",
+    "PlannerConfig",
+    "SimulatorConfig",
+    "TensorNormalizer",
+    "TrainingConfig",
+    "compute_trading_objective",
+    "prepare_daily_plan_tensors",
+    "simulate_daily_plans",
+]

RLgpt/config.py

@@ -0,0 +1,66 @@
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from pathlib import Path
+
+
+@dataclass(frozen=True)
+class DailyPlanDataConfig:
+    symbols: tuple[str, ...]
+    data_root: Path = Path("trainingdatahourly") / "crypto"
+    forecast_cache_root: Path = Path("binanceneural") / "forecast_cache"
+    forecast_horizons: tuple[int, ...] = (1, 24)
+    sequence_length: int = 72
+    min_history_hours: int = 24 * 45
+    max_feature_lookback_hours: int = 24 * 7
+    min_bars_per_day: int = 1
+    validation_days: int = 30
+    cache_only: bool = True
+
+
+@dataclass(frozen=True)
+class PlannerConfig:
+    hidden_dim: int = 128
+    depth: int = 3
+    heads: int = 4
+    dropout: float = 0.1
+    max_center_offset_bps: float = 250.0
+    min_half_spread_bps: float = 5.0
+    max_half_spread_bps: float = 300.0
+
+
+@dataclass(frozen=True)
+class SimulatorConfig:
+    initial_cash: float = 100_000.0
+    shared_unit_budget: float = 20.0
+    max_units_per_asset: float = 10.0
+    maker_fee_bps: float = 10.0
+    slippage_bps: float = 5.0
+    fill_buffer_bps: float = 5.0
+    fill_temperature_bps: float = 8.0
+    carry_inventory: bool = False
+    min_trade_units: float = 0.0
+    raw_pnl_weight: float = 1.0
+    smooth_pnl_weight: float = 0.05
+    downside_penalty: float = 0.25
+    turnover_penalty: float = 1e-5
+    inventory_penalty: float = 1e-3
+
+
+@dataclass(frozen=True)
+class TrainingConfig:
+    data: DailyPlanDataConfig
+    planner: PlannerConfig = field(default_factory=PlannerConfig)
+    simulator: SimulatorConfig = field(default_factory=SimulatorConfig)
+    epochs: int = 25
+    batch_size: int = 16
+    learning_rate: float = 3e-4
+    weight_decay: float = 1e-4
+    grad_clip: float = 1.0
+    seed: int = 1337
+    device: str | None = None
+    num_workers: int = 0
+    max_train_days: int | None = None
+    max_val_days: int | None = None
+    run_name: str | None = None
+    output_root: Path = Path("experiments") / "RLgpt"