File size: 7,599 Bytes
1794757
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import pytest

from trenches_env.models import AgentAction, ExternalSignal, Prediction
from trenches_env.openenv_client import TrenchesEnvClient
from trenches_env.openenv_adapter import (
    OPENENV_CORE_AVAILABLE,
    OpenEnvAdapter,
    TrenchesOpenEnvAction,
    TrenchesOpenEnvEnvironment,
    create_openenv_fastapi_app,
)


def test_openenv_adapter_reset_and_step() -> None:
    runtime = OpenEnvAdapter()
    observations, info = runtime.reset(seed=11, training_stage="stage_3_sparse", max_turns=12)

    assert "us" in observations
    assert info["episode"]["training_stage"] == "stage_3_sparse"
    assert info["episode"]["max_turns"] == 12

    next_observations, rewards, terminated, truncated, next_info = runtime.step(
        actions={
            "us": AgentAction(
                actor="us",
                type="negotiate",
                target="gulf",
                summary="Offer deconfliction and shipping guarantees.",
            ),
            "oversight": AgentAction(
                actor="oversight",
                type="oversight_review",
                summary="Review escalation drift and intervention triggers.",
            ),
        },
        external_signals=[
            ExternalSignal(
                source="training-sim",
                headline="Shipping risk rises near Hormuz.",
                region="gulf",
                tags=["shipping", "oil"],
                severity=0.3,
            )
        ],
    )

    assert "us" in next_observations
    assert "us" in rewards
    assert terminated is False
    assert truncated is False
    assert next_info["turn"] == 1
    assert next_info["world"]["turn"] == 1
    assert next_info["belief_state"]["us"]["beliefs"]


def test_trenches_openenv_environment_returns_scalar_reward_for_active_agent() -> None:
    runtime = TrenchesOpenEnvEnvironment()
    observation = runtime.reset(
        seed=11,
        training_agent="us",
        training_stage="stage_3_sparse",
        max_turns=12,
        include_joint_observations=True,
    )

    assert observation.training_agent == "us"
    assert observation.turn == 0
    assert observation.reward == 0.0
    assert observation.done is False
    assert "us" in observation.joint_observations

    next_observation = runtime.step(
        TrenchesOpenEnvAction(
            action=AgentAction(
                actor="us",
                type="negotiate",
                target="gulf",
                summary="Offer deconfliction and shipping guarantees.",
            ),
            external_signals=[
                ExternalSignal(
                    source="training-sim",
                    headline="Shipping risk rises near Hormuz.",
                    region="gulf",
                    tags=["shipping", "oil"],
                    severity=0.3,
                )
            ],
        )
    )

    assert next_observation.turn == 1
    assert next_observation.training_agent == "us"
    assert next_observation.reward == runtime.state.reward_breakdowns["us"].total
    assert next_observation.agent_observation.known_coalitions
    assert next_observation.agent_observation.belief_brief
    assert runtime.state.step_count == 1
    runtime.close()


def test_openenv_reset_accepts_named_scenario() -> None:
    runtime = TrenchesOpenEnvEnvironment()
    observation = runtime.reset(
        seed=17,
        training_agent="gulf",
        training_stage="stage_3_sparse",
        max_turns=12,
        scenario_id="shipping_crisis",
    )

    assert observation.agent_observation.strategic_state["shipping_continuity"] < 78.0
    assert runtime.state.session is not None
    assert runtime.state.session.episode.scenario_id == "shipping_crisis"
    assert runtime.state.session.world.latent_state["gulf"]["shipping_continuity"] < 78.0
    assert observation.agent_observation.projection.enabled
    runtime.close()


def test_openenv_autofills_missing_agents_with_shared_policy() -> None:
    runtime = TrenchesOpenEnvEnvironment()
    runtime.reset(
        seed=11,
        training_agent="us",
        training_stage="stage_3_sparse",
        max_turns=12,
        include_joint_observations=True,
    )

    runtime.step(
        TrenchesOpenEnvAction(
            action=AgentAction(
                actor="us",
                type="negotiate",
                target="gulf",
                summary="Offer deconfliction and shipping guarantees.",
            ),
            external_signals=[
                ExternalSignal(
                    source="training-sim",
                    headline="Shipping risk rises near Hormuz.",
                    region="gulf",
                    tags=["shipping", "oil"],
                    severity=0.3,
                )
            ],
        )
    )

    assert runtime.state.session is not None
    assert runtime.state.session.recent_traces[-1].actions["gulf"].type in {"defend", "negotiate", "intel_query"}
    runtime.close()


def test_historical_replay_step_records_prediction_and_scores_forecast() -> None:
    runtime = TrenchesOpenEnvEnvironment()
    observation = runtime.reset(
        seed=11,
        training_agent="us",
        training_stage="stage_1_dense",
        max_turns=4,
        replay_id="us_synthetic_seed_2025_2026",
        replay_start_index=0,
    )

    assert observation.historical_replay.enabled is True
    assert observation.historical_replay.current_event_index == 0
    assert observation.historical_replay.ground_truth_timeline == []
    assert observation.agent_observation.historical_brief

    next_observation = runtime.step(
        TrenchesOpenEnvAction(
            action=AgentAction(
                actor="us",
                type="negotiate",
                target="gulf",
                summary="Reassure Gulf partners and reinforce shipping protection.",
            ),
            prediction=Prediction(
                agent_id="us",
                topic="shipping",
                predicted_actor="us",
                predicted_target="shipping_lanes",
                time_horizon_turns=1,
                expected_severity="medium",
                confidence=0.74,
                summary="The next visible event is likely a US maritime reassurance move.",
                rationale="Washington is likely to answer shipping pressure with a visible assurance posture.",
            ),
        )
    )

    assert next_observation.revealed_event is not None
    assert next_observation.revealed_event.event_id == "evt-2025-02-us-maritime-posture"
    assert next_observation.historical_replay.ground_truth_timeline == []
    assert next_observation.reward_breakdown.forecast_total > 0.0
    assert next_observation.prediction_assessments["us"].evaluated_event_id == "evt-2025-02-us-maritime-posture"
    assert runtime.state.session is not None
    assert len(runtime.state.session.historical_replay.ground_truth_timeline) == 10
    assert runtime.state.session.prediction_log[-1].agent_id == "us"
    assert runtime.state.session.prediction_assessments[-1].total > 0.0
    runtime.close()


def test_trenches_openenv_environment_rejects_unknown_training_agent() -> None:
    runtime = TrenchesOpenEnvEnvironment()

    with pytest.raises(ValueError):
        runtime.reset(training_agent="russia")


def test_native_openenv_fastapi_app_can_be_created() -> None:
    app = create_openenv_fastapi_app()

    if OPENENV_CORE_AVAILABLE:
        assert app is not None
    else:
        assert app is None


def test_typed_openenv_client_class_is_declared() -> None:
    assert TrenchesEnvClient is not None