File size: 16,341 Bytes
698f4d8
 
 
f33ad7b
 
 
 
 
 
698f4d8
 
 
 
 
 
 
 
7183e08
698f4d8
df724f2
 
 
698f4d8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2568517
 
 
 
 
698f4d8
 
 
 
14577ec
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2568517
14577ec
 
 
 
 
 
df724f2
698f4d8
 
2568517
 
698f4d8
 
2568517
 
698f4d8
 
 
 
2568517
 
 
698f4d8
 
 
 
 
 
 
 
 
2568517
698f4d8
 
 
2568517
 
 
 
698f4d8
 
 
 
 
2568517
698f4d8
 
 
 
 
 
 
 
 
2568517
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
df724f2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
698f4d8
 
 
 
 
 
 
2568517
 
 
698f4d8
 
 
 
 
2568517
 
 
 
 
698f4d8
 
 
 
 
 
 
 
 
2568517
698f4d8
 
 
 
 
 
 
2568517
 
698f4d8
 
 
2568517
 
 
 
 
 
 
698f4d8
 
 
 
2568517
698f4d8
 
 
 
 
 
 
 
 
 
 
2568517
698f4d8
2568517
698f4d8
2568517
698f4d8
2568517
698f4d8
2568517
698f4d8
2568517
698f4d8
 
 
 
 
 
 
 
2568517
14577ec
 
 
 
 
698f4d8
 
 
 
 
 
2568517
 
698f4d8
 
2568517
698f4d8
 
 
2568517
 
698f4d8
2568517
698f4d8
2568517
 
698f4d8
 
 
 
 
 
 
 
 
 
 
 
 
2568517
 
698f4d8
df724f2
 
 
 
698f4d8
2568517
 
 
698f4d8
 
2568517
 
698f4d8
 
 
 
df724f2
 
 
698f4d8
 
 
2568517
698f4d8
2568517
698f4d8
 
 
 
2568517
698f4d8
 
 
 
 
 
df724f2
 
 
 
 
 
 
698f4d8
 
2568517
 
 
 
 
 
 
 
 
698f4d8
df724f2
 
698f4d8
2568517
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
698f4d8
2568517
 
 
 
 
 
 
df724f2
 
2568517
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
698f4d8
2568517
 
 
 
 
 
 
 
 
698f4d8
df724f2
 
2568517
698f4d8
 
2568517
 
698f4d8
 
 
df724f2
698f4d8
 
2568517
 
df724f2
 
 
 
698f4d8
 
2568517
 
 
698f4d8
 
df724f2
7183e08
 
 
 
 
 
 
2568517
7183e08
 
2568517
 
 
 
 
 
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
"""
Parlay OpenEnv WebSocket server.
Implements the standard reset/step/state protocol.

OpenEnv WebSocket URL (this module only; default port from ENV_PORT or --port, default 8001):
  ws://localhost:8001/env/ws
  (path is /env/ws because the router uses prefix="/env" and defines @router.websocket("/ws").)
When the same router is mounted on main:app (uvicorn main:app --port 8000):
  ws://localhost:8000/env/ws
"""
import json
import logging
import os
import uuid
from typing import Any

import numpy as np
from fastapi import APIRouter, FastAPI, WebSocket, WebSocketDisconnect

from agent.tom_tracker import ToMTracker
from game.scenarios import get_scenario

from .exceptions import (
    EpisodeAlreadyDoneError,
    InvalidActionError,
    InvalidScenarioError,
    SessionNotFoundError,
)
from .game_theory import compute_nash_bargaining_solution, compute_zopa
from .grader import compute_step_reward
from .models import (
    BeliefState,
    HiddenState,
    ParlayAction,
    ParlayObservation,
    ParlayState,
    PersonaType,
    TacticalMove,
)
from .reward import (
    ZOPA_EROSION_CONSECUTIVE_TURNS,
    ZOPA_EROSION_RATE,
    ZOPA_EROSION_TENSION_THRESHOLD,
)

logger = logging.getLogger(__name__)
router = APIRouter(prefix="/env", tags=["OpenEnv"])

_FALLBACK_BELIEF = BeliefState(
    est_budget=0.0,
    est_walk_away=0.0,
    est_urgency=0.5,
    est_has_alternative=False,
    confidence=0.1,
)
FALLBACK_OBSERVATION = ParlayObservation(
    step_count=0,
    episode_done=False,
    current_offer=0.0,
    opponent_offer=0.0,
    zopa_lower=0.0,
    zopa_upper=0.0,
    nash_point=0.0,
    tension_score=0.0,
    belief_state=_FALLBACK_BELIEF,
    last_utterance="[Connection issue - AI is thinking]",
    available_moves=list(TacticalMove),
    credibility_points=100,
    reward=0.0,
    cumulative_reward=0.0,
)

_sessions: dict[str, dict[str, Any]] = {}

MAX_TURNS = int(os.getenv("MAX_TURNS_PER_EPISODE", "20"))
CP_START = int(os.getenv("CREDIBILITY_POINTS_START", "100"))
CP_REGEN = int(os.getenv("CREDIBILITY_REGEN_PER_TURN", "5"))

_SCENARIO_DEFAULTS: dict[str, dict[str, Any]] = {
    "saas_enterprise": dict(budget=165_000, walk=125_000, urgency=0.55, alt=True),
    "hiring_package": dict(budget=230_000, walk=195_000, urgency=0.60, alt=False),
    "acquisition_term_sheet": dict(budget=16_000_000, walk=10_500_000, urgency=0.65, alt=True),
}

_CP_COSTS: dict[TacticalMove, int] = {
    TacticalMove.ANCHOR_HIGH: 0,
    TacticalMove.BATNA_REVEAL: 20,
    TacticalMove.SILENCE: 5,
}


def _get_scenario_hidden_state(scenario_id: str, rng_seed: int = 42) -> HiddenState:
    """Return a HiddenState for the given scenario with slight randomisation."""
    if scenario_id not in _SCENARIO_DEFAULTS:
        raise InvalidScenarioError(f"Unknown scenario: {scenario_id}")

    rng = np.random.default_rng(rng_seed)
    defaults = _SCENARIO_DEFAULTS[scenario_id]
    noise = rng.uniform(0.95, 1.05)

    return HiddenState(
        budget_ceiling=round(defaults["budget"] * noise, 2),
        walk_away_price=round(defaults["walk"] * noise, 2),
        urgency_score=float(np.clip(defaults["urgency"] + rng.uniform(-0.1, 0.1), 0.0, 1.0)),
        has_alternative=bool(defaults["alt"]),
        persona_drifted=False,
    )


def _initial_belief(hidden: HiddenState) -> BeliefState:
    """Initial belief state - intentionally imprecise."""
    return BeliefState(
        est_budget=hidden.budget_ceiling * 0.80,
        est_walk_away=hidden.walk_away_price * 1.15,
        est_urgency=0.50,
        est_has_alternative=False,
        confidence=0.30,
    )


def _get_cp_cost(move: TacticalMove | None) -> int:
    if move is None:
        return 0
    return _CP_COSTS.get(move, 0)


def _compute_tension(state: ParlayState, action: ParlayAction) -> float:
    """Compute the current tension score for the turn."""
    base = 20.0 + ((state.step_count + 1) / MAX_TURNS) * 60.0
    if action.tactical_move == TacticalMove.ANCHOR_HIGH:
        base += 15.0
    elif action.tactical_move == TacticalMove.BATNA_REVEAL:
        base += 10.0
    elif action.tactical_move == TacticalMove.SILENCE:
        base += 5.0
    return float(max(0.0, min(100.0, base)))


def _apply_drift_event(state: ParlayState, tom: ToMTracker) -> str | None:
    """Apply scenario drift event at the current turn, if any."""
    try:
        scenario = get_scenario(state.scenario_id)
    except Exception:
        return None

    for event in scenario.drift_events:
        if event.trigger_turn == state.step_count:
            state.drift_events_fired += 1
            state.hidden_state.persona_drifted = True
            tom.drift_event(
                event.effect_on_urgency,
                event.effect_on_has_alternative,
                event_description=event.event,
            )
            state.belief_history = list(tom.history)
            return event.event
    return None


def _make_observation(
    state: ParlayState,
    reward: float,
    utterance: str,
    drift_event: str | None = None,
) -> ParlayObservation:
    """Build a ParlayObservation from the current state."""
    zopa = compute_zopa(state.hidden_state.budget_ceiling, state.hidden_state.walk_away_price)
    zopa_lower = zopa[0] if zopa else state.hidden_state.walk_away_price
    zopa_upper = zopa[1] if zopa else state.hidden_state.budget_ceiling
    nash = compute_nash_bargaining_solution(
        state.hidden_state.budget_ceiling,
        state.hidden_state.walk_away_price,
    )
    current_offer = state.offer_history[-1] if state.offer_history else 0.0
    belief = state.belief_history[-1] if state.belief_history else _initial_belief(state.hidden_state)

    original_zopa = state.original_zopa_width
    current_zopa = max(0.0, state.hidden_state.budget_ceiling - state.hidden_state.walk_away_price)
    width_pct = current_zopa / original_zopa if original_zopa > 0 else 0.0

    return ParlayObservation(
        step_count=state.step_count,
        episode_done=state.episode_done,
        current_offer=current_offer,
        opponent_offer=zopa_upper * 0.9,
        zopa_lower=zopa_lower,
        zopa_upper=zopa_upper,
        nash_point=nash,
        tension_score=state.tension_score,
        belief_state=belief,
        last_utterance=utterance,
        available_moves=list(TacticalMove),
        credibility_points=state.credibility_points,
        reward=reward,
        cumulative_reward=state.cumulative_reward,
        drift_event=drift_event,
        zopa_erosion_ticks=state.zopa_erosion_ticks,
        zopa_width_pct_remaining=width_pct,
    )


def _coerce_message_params(msg: dict[str, Any]) -> tuple[str, dict[str, Any]]:
    """Accept both {cmd, ...} and {method, params} envelope formats."""
    if "method" in msg:
        return str(msg["method"]), dict(msg.get("params", {}))
    command = str(msg.get("cmd", ""))
    params = {k: v for k, v in msg.items() if k != "cmd"}
    return command, params


@router.websocket("/ws")
async def env_websocket(websocket: WebSocket) -> None:
    """OpenEnv WebSocket endpoint."""
    await websocket.accept()
    logger.info("OpenEnv WebSocket client connected")
    try:
        while True:
            raw = await websocket.receive_text()
            try:
                msg = json.loads(raw)
            except json.JSONDecodeError:
                await websocket.send_json({"error": "Invalid JSON"})
                continue

            command, params = _coerce_message_params(msg)
            try:
                match command:
                    case "reset":
                        result = await _handle_reset(params)
                    case "step":
                        result = await _handle_step(params)
                    case "state":
                        result = await _handle_state(params)
                    case _:
                        result = {"error": f"Unknown command: {command}"}
            except (
                InvalidActionError,
                SessionNotFoundError,
                EpisodeAlreadyDoneError,
                InvalidScenarioError,
            ) as exc:
                result = {"error": str(exc)}
            except Exception:
                logger.exception("Unhandled error in env WebSocket - returning fallback observation")
                result = {
                    "observation": FALLBACK_OBSERVATION.model_dump(),
                    "done": False,
                    "_fallback": True,
                }

            await websocket.send_json(result)
    except WebSocketDisconnect:
        logger.info("OpenEnv WebSocket client disconnected")


async def _handle_reset(msg: dict[str, Any]) -> dict:
    """Handle reset: create a fresh episode."""
    scenario_id = msg.get("scenario_id", "saas_enterprise")
    persona_str = msg.get("persona", "shark")
    seed = int(msg.get("seed", 42))

    try:
        persona = PersonaType(persona_str)
    except ValueError as exc:
        raise InvalidScenarioError(f"Unknown persona: {persona_str}") from exc

    hidden = _get_scenario_hidden_state(scenario_id, seed)
    initial_belief = _initial_belief(hidden)
    session_id = str(uuid.uuid4())
    original_zopa_width = hidden.budget_ceiling - hidden.walk_away_price

    state = ParlayState(
        session_id=session_id,
        scenario_id=scenario_id,
        persona=persona,
        step_count=0,
        cumulative_reward=0.0,
        hidden_state=hidden,
        belief_history=[initial_belief],
        offer_history=[],
        drift_events_fired=0,
        episode_done=False,
        credibility_points=CP_START,
        original_zopa_width=original_zopa_width,
        zopa_width_pct_remaining=1.0,
    )
    _sessions[session_id] = {
        "state": state,
        "tom_tracker": ToMTracker(initial_belief, persona),
    }

    observation = _make_observation(state, 0.0, "Negotiation started. Make your opening move.")
    logger.info("Reset: session=%s, scenario=%s, persona=%s", session_id, scenario_id, persona_str)
    return {"session_id": session_id, "observation": observation.model_dump(), "done": False}


async def _handle_step(msg: dict[str, Any]) -> dict:
    """Advance the episode by one action."""
    session_id = msg.get("session_id")
    if not session_id or session_id not in _sessions:
        raise SessionNotFoundError(f"Session {session_id} not found")

    session = _sessions[session_id]
    state: ParlayState = session["state"]
    tom: ToMTracker = session["tom_tracker"]
    if state.episode_done:
        raise EpisodeAlreadyDoneError(f"Episode {session_id} is already done")

    action_payload = msg.get("action", msg)
    try:
        action = ParlayAction.model_validate(action_payload)
    except Exception as exc:
        raise InvalidActionError(f"Invalid action: {exc}") from exc

    cp_cost = _get_cp_cost(action.tactical_move)
    if state.credibility_points < cp_cost:
        raise InvalidActionError("Insufficient credibility points for that move")

    new_offers = list(state.offer_history)
    if action.offer_amount is not None:
        new_offers.append(action.offer_amount)

    tom.update(
        observed_offer=action.offer_amount,
        observed_move=action.tactical_move,
        utterance=action.utterance,
        turn=state.step_count + 1,
    )
    new_beliefs = list(tom.history)

    next_state = ParlayState(
        **{
            **state.model_dump(),
            "step_count": state.step_count + 1,
            "offer_history": new_offers,
            "belief_history": new_beliefs,
            "credibility_points": min(CP_START, state.credibility_points + CP_REGEN - cp_cost),
            "tension_score": _compute_tension(state, action),
            "hidden_state": HiddenState(**state.hidden_state.model_dump()),
        }
    )
    # Keep belief history aligned with ToM tracker history (single source of truth).
    next_state.belief_history = new_beliefs

    if action.tactical_move == TacticalMove.BATNA_REVEAL:
        revealed = action.offer_amount if action.offer_amount is not None else next_state.hidden_state.walk_away_price
        next_state.hidden_state.last_stated_batna = float(revealed)

    if next_state.tension_score >= ZOPA_EROSION_TENSION_THRESHOLD:
        next_state.high_tension_streak += 1
    else:
        next_state.high_tension_streak = 0

    if next_state.high_tension_streak >= ZOPA_EROSION_CONSECUTIVE_TURNS:
        zopa_width = next_state.hidden_state.budget_ceiling - next_state.hidden_state.walk_away_price
        base_width = next_state.original_zopa_width or zopa_width
        shift = base_width * ZOPA_EROSION_RATE
        next_state.hidden_state.budget_ceiling -= shift
        next_state.hidden_state.walk_away_price += shift
        next_state.zopa_erosion_ticks += 1
        next_state.high_tension_streak = 0

        if next_state.hidden_state.budget_ceiling <= next_state.hidden_state.walk_away_price:
            next_state.walk_away = True
            next_state.termination_reason = "zopa_collapsed"

    current_zopa = max(0.0, next_state.hidden_state.budget_ceiling - next_state.hidden_state.walk_away_price)
    next_state.zopa_width_pct_remaining = (
        current_zopa / next_state.original_zopa_width if next_state.original_zopa_width > 0 else 0.0
    )

    if action.offer_amount is not None:
        next_state.deal_reached = (
            next_state.hidden_state.walk_away_price
            <= action.offer_amount
            <= next_state.hidden_state.budget_ceiling
        )

    drift_event = _apply_drift_event(next_state, tom)
    step_reward = compute_step_reward(state, action, next_state, drift_event=drift_event)
    next_state.cumulative_reward = state.cumulative_reward + step_reward

    if step_reward >= 0.0 and action.tactical_move is None and state.hidden_state.last_stated_batna is not None:
        from .grader import detect_bluff_challenge  # noqa: PLC0415

        if detect_bluff_challenge(
            utterance=action.utterance,
            opponent_stated_batna=state.hidden_state.last_stated_batna,
            opponent_true_batna=state.hidden_state.budget_ceiling,
        ):
            next_state.bluffs_caught = state.bluffs_caught + 1

    next_state.episode_done = (
        next_state.step_count >= MAX_TURNS
        or step_reward < -100.0
        or next_state.deal_reached
        or next_state.walk_away
    )
    if next_state.episode_done and next_state.termination_reason is None:
        if next_state.deal_reached:
            next_state.termination_reason = "deal_reached"
        elif step_reward < -100.0:
            next_state.termination_reason = "reward_floor"
        elif next_state.walk_away:
            next_state.termination_reason = "walk_away"
        else:
            next_state.termination_reason = "max_turns"

    _sessions[session_id] = {"state": next_state, "tom_tracker": tom}
    observation = _make_observation(next_state, step_reward, action.utterance, drift_event=drift_event)
    return {"observation": observation.model_dump(), "done": next_state.episode_done}


async def _handle_state(msg: dict[str, Any]) -> dict:
    """Return raw session state."""
    session_id = msg.get("session_id")
    if not session_id or session_id not in _sessions:
        raise SessionNotFoundError(f"Session {session_id} not found")
    return {"state": _sessions[session_id]["state"].model_dump()}


def get_session_state(session_id: str) -> ParlayState | None:
    """Return the in-memory session state for SSE and tests."""
    session = _sessions.get(session_id)
    if not session:
        return None
    return session["state"]


@router.get("/sessions/{session_id}")
async def get_session(session_id: str) -> dict:
    """Get session state via REST."""
    if session_id not in _sessions:
        raise SessionNotFoundError(f"Session {session_id} not found")
    return {"state": _sessions[session_id]["state"].model_dump()}


_env_app = FastAPI(title="Parlay OpenEnv", version="1.0.0")
_env_app.include_router(router)


if __name__ == "__main__":
    import argparse
    import uvicorn

    parser = argparse.ArgumentParser(description="Run the Parlay OpenEnv server")
    parser.add_argument("--port", type=int, default=int(os.getenv("ENV_PORT", "8001")))
    args = parser.parse_args()

    port = int(args.port)
    uvicorn.run(_env_app, host="0.0.0.0", port=port)