malkhut(T3): planner + adversarial counterparties

Planner: Decoupled UCB/UCT simultaneous-move MCTS (sm_mcts.py),
compact action space (action_menu.py), planner alternatives (alternatives.py).
Counterparties: 4+ adversarial agent ecology — ToxicTaker, LatencyArb,
MarketMaker, NoiseTrader + extended: LiquidationFlow, WhaleOrder,
MomentumFollower, SpoofDetector, QueueChaser.

MALKHUT/malkhut/planner/action_menu.py

@@ -0,0 +1,129 @@
+"""
+Action menu builder — reduces impossible action space to compact meaningful set.
+
+Menu size target:
+    our actions: 8-24
+    each counterparty role: 3-12
+    depth: 2-4
+"""
+from __future__ import annotations
+
+from typing import List, Optional, Tuple
+
+from malkhut.state import (
+    ActionKind,
+    FulfilmentPolicyParams,
+    IntentKind,
+    MarketWorldState,
+    OpenOrderState,
+    OrderType,
+    Side,
+    TradePathState,
+)
+from malkhut.actions import FulfilmentAction
+
+
+def _side_for_intent(intent_kind: IntentKind) -> Side:
+    if intent_kind in (IntentKind.ENTER_LONG, IntentKind.ADD_LONG, IntentKind.REDUCE_SHORT, IntentKind.EXIT_SHORT):
+        return Side.BUY
+    return Side.SELL
+
+
+def _exit_side_for_position(state: MarketWorldState) -> Optional[Side]:
+    path = state.trade_path
+    if path is None:
+        return None
+    return Side.SELL if path.side == Side.BUY else Side.BUY
+
+
+def _path_risk_says_exit(state: MarketWorldState, params: FulfilmentPolicyParams) -> bool:
+    path = state.trade_path
+    if path is None:
+        return False
+
+    if abs(path.mae_bps) >= params.mae_tail_cut_bps:
+        if path.recovery_velocity_bps_per_s < params.recovery_velocity_min_bps_per_s:
+            return True
+
+    if path.time_in_loss_s > params.max_time_in_loss_s:
+        return True
+
+    if path.failed_recovery_count >= params.failed_recovery_cut_count:
+        return True
+
+    if path.mfe_bps > 0:
+        giveback = path.distance_from_mfe_bps / max(path.mfe_bps, 1e-12)
+        if giveback >= params.mfe_giveback_cut_fraction:
+            return True
+
+    return False
+
+
+def build_our_actions(
+    state: MarketWorldState,
+    params: FulfilmentPolicyParams,
+) -> Tuple[FulfilmentAction, ...]:
+    intent = state.intent
+    if intent is None:
+        return (FulfilmentAction(ActionKind.NOOP, None, None, 0, 0.0, 0),)
+
+    side = _side_for_intent(intent)
+    actions: List[FulfilmentAction] = []
+
+    # Always allow no-op
+    actions.append(FulfilmentAction(
+        kind=ActionKind.NOOP, side=None, order_type=None,
+        price_ticks_from_best=0, qty_fraction=0.0, ttl_ms=100,
+    ))
+
+    # Existing order management
+    for oo in state.open_orders:
+        if oo.symbol != intent.symbol:
+            continue
+
+        actions.append(FulfilmentAction(
+            kind=ActionKind.CANCEL, side=oo.side, order_type=None,
+            price_ticks_from_best=0, qty_fraction=0.0, ttl_ms=0,
+            cancel_order_id=oo.client_order_id,
+        ))
+
+        for offset in params.quote_offsets_ticks:
+            actions.append(FulfilmentAction(
+                kind=ActionKind.CANCEL_REPLACE, side=side,
+                order_type=OrderType.POST_ONLY if intent.prefer_maker else OrderType.LIMIT,
+                price_ticks_from_best=offset, qty_fraction=0.25,
+                ttl_ms=params.passive_ttl_ms, cancel_order_id=oo.client_order_id,
+                post_only=intent.prefer_maker, reduce_only=intent.reduce_only,
+            ))
+
+    # Passive quote placements
+    for offset in params.quote_offsets_ticks:
+        for frac in params.quote_size_fractions:
+            actions.append(FulfilmentAction(
+                kind=ActionKind.PLACE, side=side,
+                order_type=OrderType.POST_ONLY if intent.prefer_maker else OrderType.LIMIT,
+                price_ticks_from_best=offset, qty_fraction=frac,
+                ttl_ms=params.passive_ttl_ms,
+                post_only=intent.prefer_maker, reduce_only=intent.reduce_only,
+            ))
+
+    # Aggressive crossing if urgency allows
+    if intent.urgency > 0.65:
+        for frac in (0.05, 0.10, 0.25):
+            actions.append(FulfilmentAction(
+                kind=ActionKind.CROSS_SPREAD, side=side,
+                order_type=OrderType.IOC, price_ticks_from_best=0,
+                qty_fraction=frac, ttl_ms=params.aggressive_ttl_ms,
+                reduce_only=intent.reduce_only,
+            ))
+
+    # Path-risk exits
+    if _path_risk_says_exit(state, params):
+        actions.append(FulfilmentAction(
+            kind=ActionKind.FULL_EXIT, side=_exit_side_for_position(state),
+            order_type=OrderType.REDUCE_ONLY_MARKET, price_ticks_from_best=0,
+            qty_fraction=1.0, ttl_ms=0, reduce_only=True,
+            metadata={"reason": "path_risk_exit"},
+        ))
+
+    return tuple(actions)