syllogym-env / __init__.py
farffadet's picture
refactor: drivers→generators, v2_reward→reward + cleanup
41814f2 verified
"""
SylloGym — Multi-turn Legal Reasoning Environment.
JudgeEnv: the agent plays a judge who receives case facts turn by turn
and must revise their ruling as the case evolves. Twelve domains of US
law, procedurally generated episodes, deterministic Python verifiers.
Example (standalone):
>>> from syllogym_env import SylloGymEnv
>>> env = SylloGymEnv()
>>> obs = env.reset()
>>> result = env.review_document("arrest_report")
>>> result = env.conclude("Yes")
>>> print(env.reward)
Example (MCP via OpenEnv client):
>>> from openenv.core.mcp_client import MCPToolClient
>>> client = MCPToolClient("http://localhost:8000")
>>> tools = client.list_tools()
>>> result = client.call_tool("review_document", name="arrest_report")
"""
from .server.core.investigation_env import SylloGymEnv
from .server.core.judge_environment import JudgeAction, JudgeObservation
from .judge_env import JudgeEnv, JudgeObs
from .models import SylloState
__all__ = ["SylloGymEnv", "JudgeEnv", "JudgeObs", "JudgeAction", "JudgeObservation", "SylloState"]