File size: 24,646 Bytes
51aafb0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
"""
simulation.py β€” NegotiateAI: Adversarial Procurement Arena
Market dynamics engine:
- Price volatility and market benchmarks
- Demand pattern simulation (weekly, seasonal)
- Disruption event scheduling
- Rival buyer pressure propagation
- Market signal generation
"""

from __future__ import annotations

import math
import random
import uuid
from typing import Any

from models import (
    DisruptionEvent,
    DisruptionType,
    ItemCategory,
    MarketSignal,
    SupplierReputation,
    SupplierView,
)


# ─────────────────────────────────────────────────────────────
# MARKET PRICE DYNAMICS
# ─────────────────────────────────────────────────────────────

# Base price volatility per category (weekly Οƒ as fraction of price)
PRICE_VOLATILITY: dict[ItemCategory, float] = {
    ItemCategory.SOFTWARE:  0.03,   # stable β€” subscription models
    ItemCategory.HARDWARE:  0.08,   # volatile β€” supply chain sensitive
    ItemCategory.SERVICES:  0.05,   # moderate β€” labour market
}

# Seasonal price multipliers by month (1=Jan … 12=Dec)
# Maps to fiscal weeks: week 1-4 = Q1, 5-8 = Q2, 9-12 = Q3
SEASONAL_MULTIPLIERS: dict[ItemCategory, dict[int, float]] = {
    ItemCategory.SOFTWARE: {
        1: 0.95,   # Q1: post-renewal discounts
        2: 0.95,
        3: 1.00,
        4: 1.02,   # Q2: mid-year renewals
        5: 1.02,
        6: 1.00,
        7: 0.98,   # Q3: summer slow
        8: 0.98,
        9: 1.05,   # Q4: year-end budget flush
        10: 1.08,
        11: 1.10,
        12: 1.05,
    },
    ItemCategory.HARDWARE: {
        1: 0.92,   # post-holiday stock clearance
        2: 0.90,
        3: 0.95,
        4: 0.98,
        5: 1.00,
        6: 1.02,
        7: 0.97,   # summer lull
        8: 0.98,
        9: 1.05,   # back-to-school / enterprise refresh
        10: 1.08,
        11: 1.12,  # pre-holiday demand spike
        12: 1.10,
    },
    ItemCategory.SERVICES: {
        1: 1.05,   # new year headcount push
        2: 1.02,
        3: 1.00,
        4: 1.00,
        5: 0.97,
        6: 0.95,   # summer holidays
        7: 0.93,
        8: 0.95,
        9: 1.02,
        10: 1.05,
        11: 1.08,  # year-end projects
        12: 1.10,
    },
}

# Day-of-week demand multipliers (week index 1-12 maps to simulated months)
DOW_MULTIPLIERS = [1.0, 1.05, 1.10, 1.08, 1.15, 1.20, 0.90]
# Mon  Tue   Wed   Thu   Fri   Sat   Sun
# (Sun = low demand, Fri-Sat = high demand in B2B simulated sense)


class MarketSimulator:
    """
    Simulates realistic tech procurement market dynamics.
    Provides price benchmarks, demand signals, and volatility.
    """

    def __init__(self, seed: int | None = None):
        self.rng = random.Random(seed)
        self._price_history: dict[str, list[float]] = {}  # supplier_id β†’ prices

    # ── Price benchmarks ──────────────────────────────────────

    def get_market_price(
        self,
        category: ItemCategory,
        base_price: float,
        week: int,
        noise: bool = True,
    ) -> float:
        """
        Compute current market price for a category.
        Applies seasonal multiplier + Gaussian noise.
        """
        # Map week (1-12) to month (1-12)
        month = min(12, max(1, week))
        seasonal = SEASONAL_MULTIPLIERS[category].get(month, 1.0)
        price = base_price * seasonal

        if noise:
            sigma = PRICE_VOLATILITY[category]
            noise_factor = self.rng.gauss(1.0, sigma)
            price *= max(0.7, min(1.3, noise_factor))   # cap at Β±30%

        return round(price, 2)

    def compute_benchmark(
        self,
        suppliers: list[SupplierView],
        quantity: int,
        week: int,
    ) -> float:
        """
        Market benchmark = average market price across active suppliers Γ— quantity.
        Used by graders to compute cost savings.
        """
        if not suppliers:
            return 0.0
        active = [s for s in suppliers if s.is_active]
        if not active:
            active = suppliers

        prices = [
            self.get_market_price(s.category, s.base_price, week, noise=False)
            for s in active
        ]
        avg_price = sum(prices) / len(prices)
        return round(avg_price * quantity, 2)

    def price_trend(
        self,
        supplier_id: str,
        category: ItemCategory,
        base_price: float,
        weeks: int = 4,
        seed: int | None = None,
    ) -> list[float]:
        """Generate a price trend for a supplier over N weeks."""
        rng = random.Random(seed)
        prices = []
        current = base_price
        for w in range(1, weeks + 1):
            sigma = PRICE_VOLATILITY[category]
            drift = rng.gauss(0, sigma * current)
            current = max(base_price * 0.7, current + drift)
            prices.append(round(current, 2))
        return prices


# ─────────────────────────────────────────────────────────────
# DEMAND ENGINE
# ─────────────────────────────────────────────────────────────

class DemandEngine:
    """
    Stochastic demand simulation for tech procurement items.
    Models urgency spikes, seasonal patterns, and category behaviour.
    """

    # Category-specific demand elasticity to discount
    DISCOUNT_ELASTICITY: dict[ItemCategory, float] = {
        ItemCategory.SOFTWARE:  0.10,   # +10% demand per 10% discount
        ItemCategory.HARDWARE:  0.15,   # +15%
        ItemCategory.SERVICES:  0.08,   # +8% (less elastic)
    }

    # Urgency decay: demand drops as deadline approaches without action
    URGENCY_DECAY_WEEKS = 2   # last N weeks before deadline

    def __init__(self, seed: int | None = None):
        self.rng = random.Random(seed)

    def daily_demand(
        self,
        base_quantity: int,
        category: ItemCategory,
        week: int,
        discount_pct: float = 0.0,
        weeks_to_deadline: int = 99,
    ) -> int:
        """
        Compute stochastic demand for a given week.
        Applies: seasonal Γ— DOW Γ— discount elasticity Γ— urgency Γ— noise
        """
        # Seasonal factor
        month = min(12, max(1, week))
        seasonal = SEASONAL_MULTIPLIERS[category].get(month, 1.0)

        # Simulated day-of-week (week number maps to a day pattern)
        dow_idx = week % 7
        dow = DOW_MULTIPLIERS[dow_idx]

        # Discount elasticity
        elasticity = self.DISCOUNT_ELASTICITY[category]
        discount_boost = 1.0 + (discount_pct / 10.0) * elasticity

        # Urgency decay (demand drops last 2 days before deadline)
        urgency = 1.0
        if weeks_to_deadline <= self.URGENCY_DECAY_WEEKS:
            urgency = max(0.5, weeks_to_deadline / self.URGENCY_DECAY_WEEKS)

        # Gaussian noise (Οƒ=15%)
        noise = self.rng.gauss(1.0, 0.15)
        noise = max(0.5, min(1.5, noise))

        demand = base_quantity * seasonal * dow * discount_boost * urgency * noise
        return max(1, int(round(demand)))

    def demand_spike(
        self,
        base_quantity: int,
        multiplier: float = 2.0,
        seed: int | None = None,
    ) -> int:
        """Simulate an unexpected demand spike (e.g. new hire batch)."""
        rng = random.Random(seed)
        spike = base_quantity * multiplier * rng.uniform(0.8, 1.2)
        return max(1, int(round(spike)))

    def expected_weekly_demand(
        self,
        base_quantity: int,
        category: ItemCategory,
        total_weeks: int,
    ) -> float:
        """Average expected demand per week over the episode."""
        total = sum(
            SEASONAL_MULTIPLIERS[category].get(w, 1.0)
            for w in range(1, total_weeks + 1)
        )
        return (base_quantity * total) / total_weeks


# ─────────────────────────────────────────────────────────────
# DISRUPTION SCHEDULER
# ─────────────────────────────────────────────────────────────

class DisruptionScheduler:
    """
    Schedules and tracks market disruption events.
    Supports both pre-configured (task-level) and dynamic disruptions.
    """

    def __init__(
        self,
        disruption_config: list[dict[str, Any]],
        total_weeks: int,
        seed: int | None = None,
    ):
        self.rng = random.Random(seed)
        self.total_weeks = total_weeks
        self._schedule: list[dict[str, Any]] = list(disruption_config)
        self._triggered: list[DisruptionEvent] = []

    def get_scheduled_weeks(self) -> list[int]:
        """Return all weeks that have disruptions scheduled."""
        return [cfg.get("week", 0) for cfg in self._schedule]

    def pop_events_for_week(self, week: int) -> list[dict[str, Any]]:
        """Return and remove all events scheduled for this week."""
        due = [cfg for cfg in self._schedule if cfg.get("week") == week]
        self._schedule = [cfg for cfg in self._schedule if cfg.get("week") != week]
        return due

    def inject_dynamic_disruption(
        self,
        dtype: DisruptionType,
        week: int,
        **kwargs: Any,
    ) -> None:
        """Add a runtime disruption (e.g. triggered by rival activity)."""
        self._schedule.append({"type": dtype, "week": week, **kwargs})

    def record_triggered(self, event: DisruptionEvent) -> None:
        self._triggered.append(event)

    def triggered_count(self) -> int:
        return len(self._triggered)

    def all_triggered(self) -> list[DisruptionEvent]:
        return list(self._triggered)

    def remaining_count(self) -> int:
        return len(self._schedule)


# ─────────────────────────────────────────────────────────────
# RIVAL PRESSURE PROPAGATOR
# ─────────────────────────────────────────────────────────────

class RivalPressurePropagator:
    """
    Models how rival buyer activity spreads across the supplier network.
    When rival targets a supplier, pressure bleeds to related suppliers.
    Creates realistic market tension that the buyer agent must navigate.
    """

    # How much pressure bleeds to same-category suppliers
    CATEGORY_BLEED_FACTOR = 0.3

    # Pressure decay per week (rival pressure fades if not reinforced)
    DECAY_RATE = 0.15

    def __init__(self, seed: int | None = None):
        self.rng = random.Random(seed)

    def propagate(
        self,
        suppliers: dict[str, SupplierView],
        rival_target_id: str,
        pressure_added: float = 0.3,
    ) -> dict[str, float]:
        """
        When rival targets a supplier, compute pressure updates
        for all suppliers in the same category.
        Returns dict of supplier_id β†’ new pressure.
        """
        updates: dict[str, float] = {}

        if rival_target_id not in suppliers:
            return updates

        target = suppliers[rival_target_id]

        # Direct pressure on targeted supplier
        current = target.rival_pressure
        updates[rival_target_id] = min(1.0, current + pressure_added)

        # Bleed to same-category suppliers
        same_cat = [
            (sid, sv) for sid, sv in suppliers.items()
            if sv.category == target.category and sid != rival_target_id
        ]
        for sid, sv in same_cat:
            bleed = pressure_added * self.CATEGORY_BLEED_FACTOR
            bleed *= self.rng.uniform(0.5, 1.0)   # randomise bleed
            updates[sid] = min(1.0, sv.rival_pressure + bleed)

        return updates

    def decay_all(
        self,
        suppliers: dict[str, SupplierView],
    ) -> dict[str, float]:
        """
        Apply weekly pressure decay to all suppliers.
        Returns dict of supplier_id β†’ decayed pressure.
        """
        updates: dict[str, float] = {}
        for sid, sv in suppliers.items():
            decayed = sv.rival_pressure * (1.0 - self.DECAY_RATE)
            updates[sid] = max(0.0, round(decayed, 3))
        return updates


# ─────────────────────────────────────────────────────────────
# MARKET SIGNAL GENERATOR
# ─────────────────────────────────────────────────────────────

class MarketSignalGenerator:
    """
    Generates realistic market intelligence signals.
    Some signals are genuine, some are noise β€” agent must learn to filter.
    """

    # Probability that a signal about a deceptive supplier is a true warning
    TRUE_SIGNAL_RATE: dict[str, float] = {
        "cooperative":  0.3,    # rarely generate warnings
        "aggressive":   0.5,    # sometimes leak pressure signals
        "deceptive":    0.7,    # often generate quality/capacity alerts
        "distressed":   0.8,    # frequently leak financial stress signals
    }

    def __init__(self, seed: int | None = None):
        self.rng = random.Random(seed)

    def generate_weekly_signals(
        self,
        suppliers: list[SupplierView],
        reputations: list[SupplierReputation],
        week: int,
        rival_active: bool = False,
    ) -> list[MarketSignal]:
        """
        Generate market intelligence signals for this week.
        Mix of genuine warnings and noise.
        """
        signals: list[MarketSignal] = []
        rep_map = {r.supplier_id: r for r in reputations}

        for sv in suppliers:
            if not sv.is_active:
                continue

            rep = rep_map.get(sv.supplier_id)
            if not rep:
                continue

            # Community reputation signal
            if rep.community_rating < 0.65:
                signals.append(MarketSignal(
                    signal_id=str(uuid.uuid4()),
                    week=week,
                    signal_type="quality_alert",
                    supplier_id=sv.supplier_id,
                    description=(
                        f"{sv.name} community rating dropped to "
                        f"{rep.community_rating:.2f}. "
                        f"Known issues: {', '.join(rep.known_issues) or 'none reported'}."
                    ),
                    severity=round(1.0 - rep.community_rating, 2),
                ))

            # Capacity warning β€” high rival pressure
            if sv.rival_pressure > 0.6:
                signals.append(MarketSignal(
                    signal_id=str(uuid.uuid4()),
                    week=week,
                    signal_type="capacity_warning",
                    supplier_id=sv.supplier_id,
                    description=(
                        f"High buyer interest at {sv.name}. "
                        f"Available capacity may be limited this week."
                    ),
                    severity=round(sv.rival_pressure, 2),
                ))

            # Price shift signal (random, category-level)
            if self.rng.random() < PRICE_VOLATILITY[sv.category] * 2:
                direction = self.rng.choice(["rising", "falling"])
                pct = self.rng.uniform(3, 12)
                signals.append(MarketSignal(
                    signal_id=str(uuid.uuid4()),
                    week=week,
                    signal_type="price_shift",
                    supplier_id=sv.supplier_id,
                    category=sv.category,
                    description=(
                        f"{sv.category.value.title()} market prices {direction} "
                        f"~{pct:.0f}% this week. "
                        f"Check {sv.name} quotes."
                    ),
                    severity=round(pct / 100, 2),
                ))

        # Rival activity signal
        if rival_active and self.rng.random() < 0.4:
            high_pressure = [s for s in suppliers if s.rival_pressure > 0.5]
            if high_pressure:
                target = self.rng.choice(high_pressure)
                signals.append(MarketSignal(
                    signal_id=str(uuid.uuid4()),
                    week=week,
                    signal_type="rival_activity",
                    supplier_id=target.supplier_id,
                    description=(
                        f"Competitor activity detected at {target.name}. "
                        f"They may be negotiating for similar capacity."
                    ),
                    severity=round(target.rival_pressure, 2),
                ))

        return signals

    def generate_disruption_warning(
        self,
        week: int,
        disruption_week: int,
        dtype: DisruptionType,
        severity: float = 0.7,
    ) -> MarketSignal | None:
        """
        Generate an early warning 1-2 weeks before a disruption.
        Gives the agent a chance to hedge proactively.
        """
        weeks_ahead = disruption_week - week
        if weeks_ahead not in (1, 2):
            return None

        # Less certain warning 2 weeks out
        if weeks_ahead == 2 and self.rng.random() > 0.5:
            return None

        descriptions = {
            DisruptionType.SUPPLIER_DARK: (
                "Industry reports of supply chain stress. "
                "Some suppliers may face operational issues soon."
            ),
            DisruptionType.BUDGET_CUT: (
                "Finance team reviewing Q3 procurement spend. "
                "Budget adjustments possible next week."
            ),
            DisruptionType.RIVAL_LOCKOUT: (
                "Competitor procurement activity increasing. "
                "Key suppliers may have reduced availability soon."
            ),
            DisruptionType.QUALITY_SCANDAL: (
                "Industry watchdog investigating supplier quality claims. "
                "Announcements expected soon."
            ),
            DisruptionType.DEMAND_SPIKE: (
                "Headcount projections revised upward. "
                "Additional procurement requirements likely."
            ),
        }

        return MarketSignal(
            signal_id=str(uuid.uuid4()),
            week=week,
            signal_type="disruption_warning",
            description=descriptions.get(dtype, "Market disruption signal detected."),
            severity=severity * (0.5 if weeks_ahead == 2 else 1.0),
        )


# ─────────────────────────────────────────────────────────────
# MARKET BENCHMARK CALCULATOR
# ─────────────────────────────────────────────────────────────

def compute_episode_benchmark(
    suppliers: list[SupplierView],
    requirements_config: list[dict[str, Any]],
    week: int = 1,
    simulator: MarketSimulator | None = None,
) -> float:
    """
    Compute the total market benchmark for an episode.
    This is the 'fair market price' a naive buyer would pay β€”
    used by graders to measure cost savings.
    """
    if simulator is None:
        simulator = MarketSimulator()

    total = 0.0
    for req in requirements_config:
        category = req["category"]
        quantity = req["quantity"]

        # Find matching suppliers
        matching = [s for s in suppliers if s.category == category and s.is_active]
        if not matching:
            # Fallback: use any supplier
            matching = [s for s in suppliers if s.is_active] or suppliers

        if not matching:
            continue

        # Benchmark = average market price across matching suppliers
        avg_base = sum(s.base_price for s in matching) / len(matching)
        market_price = simulator.get_market_price(category, avg_base, week, noise=False)
        total += market_price * quantity

    return round(total, 2)


# ─────────────────────────────────────────────────────────────
# STRESS TEST UTILITY
# ─────────────────────────────────────────────────────────────

def stress_test_simulation(
    n_episodes: int = 100,
    seed: int = 0,
) -> dict[str, Any]:
    """
    Run N random simulations and check for crashes / degenerate rewards.
    Used to validate env before onsite training.
    """
    import sys, unittest.mock as mock
    # Proper mock: return valid string content from LLM
    mock_client = mock.MagicMock()
    mock_choice = mock.MagicMock()
    mock_choice.message.content = "Our price is $1000/unit with 7-day delivery. Ready to proceed."
    mock_client.return_value.chat.completions.create.return_value.choices = [mock_choice]
    sys.modules['openai'] = mock_client

    from env import NegotiateAIEnv
    from models import ActionType, ProcurementAction

    results = {
        "episodes": n_episodes,
        "crashes":  0,
        "nan_rewards": 0,
        "zero_rewards": 0,
        "scores": [],
        "errors": [],
    }

    rng = random.Random(seed)

    for i in range(n_episodes):
        try:
            task = rng.choice(["easy_negotiation", "medium_adversarial"])
            env = NegotiateAIEnv()
            obs = env.reset(task_id=task, seed=i)

            for _ in range(20):   # max 20 steps per episode
                if env.done:
                    break

                # Random valid action
                if not obs.suppliers:
                    break
                sup  = rng.choice([s for s in obs.suppliers if s.is_active] or obs.suppliers)
                req  = rng.choice(obs.requirements)
                atype = rng.choice([
                    ActionType.NEGOTIATE,
                    ActionType.RAISE_PR,
                    ActionType.AWARD_CONTRACT,
                    ActionType.DEFER,
                ])

                action = ProcurementAction(
                    action_type=atype,
                    supplier_id=sup.supplier_id,
                    item_id=req.item_id,
                    message="Random test action",
                    proposed_price=rng.uniform(500, 1500),
                    proposed_quantity=rng.randint(1, 50),
                )

                obs, reward, done, info = env.step(action)

                import math as _math
                if _math.isnan(reward.total) or _math.isinf(reward.total):
                    results["nan_rewards"] += 1
                if reward.total <= 0:
                    results["zero_rewards"] += 1

            result = env.get_episode_result()
            results["scores"].append(result.total_score)

        except Exception as e:
            results["crashes"] += 1
            results["errors"].append(f"Episode {i}: {str(e)[:80]}")

    if results["scores"]:
        results["mean_score"]  = round(sum(results["scores"]) / len(results["scores"]), 4)
        results["min_score"]   = round(min(results["scores"]), 4)
        results["max_score"]   = round(max(results["scores"]), 4)

    return results


if __name__ == "__main__":
    print("Running simulation stress test (100 episodes)...")
    result = stress_test_simulation(n_episodes=100, seed=42)
    print(f"Episodes:     {result['episodes']}")
    print(f"Crashes:      {result['crashes']}")
    print(f"NaN rewards:  {result['nan_rewards']}")
    print(f"Mean score:   {result.get('mean_score', 'N/A')}")
    print(f"Score range:  {result.get('min_score', 'N/A')} – {result.get('max_score', 'N/A')}")
    if result["errors"]:
        print(f"Errors:       {result['errors'][:3]}")
    else:
        print("βœ… No crashes or errors detected")