Spaces:
Configuration error
Configuration error
ychanakya12
Triple-layer clamp on all graders + server endpoint to guarantee strictly (0.0,1.0)
2f6851a Download src/server.py from MetaStableMinds/business-sim-openenv: direct link, hf CLI and curl.
- Browser
- Download file 5.29 kB
-
https://huggingface.co/spaces/MetaStableMinds/business-sim-openenv/resolve/819f3dc6f471a94de8b068108285f5ef67a50bb6/src/server.py
- Command line
-
hf download hf://spaces/MetaStableMinds/business-sim-openenv@819f3dc6f471a94de8b068108285f5ef67a50bb6/src/server.py
-
curl -L -o server.py https://huggingface.co/spaces/MetaStableMinds/business-sim-openenv/resolve/819f3dc6f471a94de8b068108285f5ef67a50bb6/src/server.py
5.29 kB
| """ | |
| server.py β FastAPI server exposing the OpenEnv standard API. | |
| Endpoints (all required by OpenEnv spec): | |
| GET / β health check (judges ping this) | |
| GET /health β health check | |
| GET /tasks β list all tasks | |
| POST /reset β start new episode, returns CompanyObservation | |
| POST /step β take action, returns StepResult | |
| GET /state β full internal state + history | |
| GET /grade β current task score (0.0β1.0) | |
| """ | |
| from fastapi import FastAPI, HTTPException | |
| from fastapi.responses import JSONResponse | |
| from src.models import CEOAction, StepResult, FullState | |
| from src.environment.company_env import CompanyEnv | |
| from src.tasks import task_easy, task_medium, task_hard | |
| app = FastAPI( | |
| title = "Business Simulation OpenEnv", | |
| description = ( | |
| "Multi-agent business simulation where an AI CEO manages a software " | |
| "company across quarters under adversarial conditions." | |
| ), | |
| version = "0.1.0", | |
| ) | |
| # ββ In-memory session store βββββββββββββββββββββββββββββββββββββββββββββββββββ | |
| _sessions: dict[str, CompanyEnv] = {} | |
| TASK_CONFIG: dict[str, dict] = { | |
| "single_quarter_survival": dict( | |
| max_quarters = 1, | |
| difficulty = "easy", | |
| grader = task_easy, | |
| description = "End the quarter with positive cash flow.", | |
| ), | |
| "four_quarter_growth": dict( | |
| max_quarters = 4, | |
| difficulty = "medium", | |
| grader = task_medium, | |
| description = "Grow revenue 30% over 4 quarters amid market shifts.", | |
| ), | |
| "adversarial_resilience": dict( | |
| max_quarters = 8, | |
| difficulty = "hard", | |
| grader = task_hard, | |
| description = "Survive 8 quarters with reputation >= 0.6 under shocks.", | |
| ), | |
| } | |
| # ββ Health ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ | |
| def root(): | |
| """Root health check β judges automated ping.""" | |
| return {"status": "ok", "env": "business-sim-env", "version": "0.1.0"} | |
| def health(): | |
| return {"status": "ok"} | |
| # ββ Task listing ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ | |
| def list_tasks(): | |
| """OpenEnv spec: enumerate available tasks.""" | |
| return [ | |
| { | |
| "id": tid, | |
| "difficulty": cfg["difficulty"], | |
| "max_quarters": cfg["max_quarters"], | |
| "description": cfg["description"], | |
| } | |
| for tid, cfg in TASK_CONFIG.items() | |
| ] | |
| # ββ Core OpenEnv endpoints ββββββββββββββββββββββββββββββββββββββββββββββββββββ | |
| def reset(task_id: str = "single_quarter_survival"): | |
| """ | |
| POST /reset β start a new episode. | |
| Returns CompanyObservation (includes session_id for subsequent calls). | |
| """ | |
| if task_id not in TASK_CONFIG: | |
| raise HTTPException( | |
| 400, | |
| f"Unknown task_id '{task_id}'. " | |
| f"Valid options: {list(TASK_CONFIG.keys())}" | |
| ) | |
| cfg = TASK_CONFIG[task_id] | |
| env = CompanyEnv( | |
| task_id = task_id, | |
| max_quarters = cfg["max_quarters"], | |
| difficulty = cfg["difficulty"], | |
| ) | |
| result = env.reset() | |
| _sessions[env.session_id] = env | |
| return JSONResponse(result.observation.model_dump()) | |
| def step(action: CEOAction, session_id: str): | |
| """POST /step β submit CEO action, receive observation + reward.""" | |
| env = _require_session(session_id) | |
| if env.done: | |
| raise HTTPException(400, "Episode is done. Call /reset to start a new one.") | |
| return env.step(action) | |
| def state(session_id: str): | |
| """GET /state β full internal state including history and counterfactuals.""" | |
| return _require_session(session_id).get_full_state() | |
| def grade(session_id: str): | |
| """GET /grade β current episode score in [0.0, 1.0].""" | |
| env = _require_session(session_id) | |
| cfg = TASK_CONFIG[env.task_id] | |
| raw = cfg["grader"].grade(env) | |
| # OpenEnv requirement: score must be strictly in (0.0, 1.0) | |
| score = min(max(float(raw), 0.01), 0.99) | |
| return { | |
| "task_id": env.task_id, | |
| "score": round(score, 4), | |
| "quarter": env.quarter, | |
| "done": env.done, | |
| } | |
| # ββ Helper ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ | |
| def _require_session(session_id: str) -> CompanyEnv: | |
| if session_id not in _sessions: | |
| raise HTTPException(404, "Session not found. Call /reset first.") | |
| return _sessions[session_id] | |