File size: 13,468 Bytes
ef737d3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# main.py β€” OpenEnv-Compliant FastAPI Server
# Autonomy Calibration Environment v1.0
# OpenEnv India Hackathon 2026 β€” by Rhythm

from __future__ import annotations
import os
from collections import Counter
from fastapi import FastAPI, HTTPException, BackgroundTasks
from fastapi.middleware.cors import CORSMiddleware
from fastapi.staticfiles import StaticFiles
from fastapi.responses import FileResponse
import subprocess
import sys
from pydantic import BaseModel

from models import Action, Observation, Reward, StepResult, ResetRequest
from environment.scenarios import SCENARIOS

def _build_registry():
    from tasks.email_triage import EmailTriageTask
    from tasks.devops_incident import DevOpsIncidentTask
    from tasks.financial_request import FinancialRequestTask
    return {
        "email_triage": EmailTriageTask,
        "devops_incident": DevOpsIncidentTask,
        "financial_request": FinancialRequestTask,
    }

TASK_REGISTRY = _build_registry()

# ─── App ─────────────────────────────────────────────────────────────────────

app = FastAPI(
    title="Autonomy Calibration Environment",
    description=(
        "OpenEnv-compliant RL environment training agents to calibrate autonomy "
        "across Email Triage, DevOps Incident Response, and Financial Request Handling."
    ),
    version="2.0.0",
)

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# ─── Session State ────────────────────────────────────────────────────────────
# Single-session in-memory state. Sufficient for hackathon + HF Spaces.

_session: dict = {
    "task_name": None,
    "task": None,
    "step": 0,
    "history": [],
    "done": True,
    "seed": None,
    "episode_id": None,
}

# Global episode log for /api/history
_episode_log: list[dict] = []


# ─── Helpers ──────────────────────────────────────────────────────────────────

def _get_task(name: str):
    if name not in TASK_REGISTRY:
        raise HTTPException(
            status_code=400,
            detail=f"Unknown task '{name}'. Valid: {list(TASK_REGISTRY.keys())}"
        )
    return TASK_REGISTRY[name]()


# ─── API: Reset ───────────────────────────────────────────────────────────────

@app.post("/reset")
@app.post("/api/reset", response_model=Observation)
def reset(body: ResetRequest = ResetRequest()):
    try:
        task = _get_task(body.task)
        obs = task.reset(seed=body.seed)
        # Store seed in session and observation
        obs.seed = body.seed
        _session["task_name"] = body.task
        _session["task"] = task
        _session["step"] = 0
        _session["history"] = []
        _session["done"] = False
        _session["seed"] = body.seed
        # Create DB episode and store ID
        import database as db
        _session["episode_id"] = db.create_episode(body.task, body.seed)
        return obs
    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))


# ─── API: Step ───────────────────────────────────────────────────────────────

@app.post("/step")
@app.post("/api/step", response_model=StepResult)
async def step_env(action: Action):
    task = _session.get("task")
    if task is None or _session.get("done"):
        raise HTTPException(status_code=400, detail="No active episode. Call /api/reset first.")
    try:
        obs, reward, done, info = task.step(action)
        step_idx = _session["step"]
        _session["step"] += 1
        _session["done"] = done
        step_entry = {
            "step": step_idx,
            "action": action.type,
            "reward": reward.value,
            "done": done,
        }
        _session["history"].append(step_entry)
        # Persist step to SQLite
        import database as db
        db.log_step(
            episode_id=_session["episode_id"],
            step_index=step_idx,
            decision=action.type,
            reward=reward.value,
            done=done,
        )
        if done:
            episode_score = info.get("episode_score")
            db.close_episode(_session["episode_id"], episode_score or 0.0)
            _episode_log.append({
                "episode_id": _session["episode_id"],
                "task": _session["task_name"],
                "seed": _session["seed"],
                "episode_score": episode_score,
                "steps": _session["step"],
                "history": list(_session["history"]),
            })
        return StepResult(observation=obs, reward=reward, done=done, info=info)
    except RuntimeError as e:
        raise HTTPException(status_code=400, detail=str(e))
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))


# ─── API: State ───────────────────────────────────────────────────────────────

@app.get("/api/state")
def state():
    task = _session.get("task")
    if task is None:
        return {"status": "not_started"}
    return {
        "status": "done" if _session["done"] else "active",
        "task": _session["task_name"],
        "step": _session["step"],
        **task.state(),
    }

# ─── API: Training ────────────────────────────────────────────────────────────

def run_training():
    """Runs the training script in a separate process and pipes output to logs."""
    try:
        print("πŸš€ GRPO TRAINING CORE: Initializing...")
        # Redirect stdout and stderr to the main process ones so they appear in HF Logs
        process = subprocess.Popen(
            [sys.executable, "train_rl.py"],
            stdout=sys.stdout, 
            stderr=sys.stderr,
            bufsize=1, # Line buffered
            universal_newlines=True
        )
        print(f"βœ… Background process PID {process.pid} spawned.")
    except Exception as e:
        print(f"❌ Error during background training: {e}")

@app.post("/api/train")
def start_training(background_tasks: BackgroundTasks):
    # Basic check for GPU presence (useful for logs)
    try:
        import torch
        has_gpu = torch.cuda.is_available()
        device_name = torch.cuda.get_device_name(0) if has_gpu else "CPU"
    except ImportError:
        has_gpu = False
        device_name = "CPU"
    
    background_tasks.add_task(run_training)
    
    return {
        "status": "started",
        "message": "GRPO Training started in background.",
        "using_gpu": has_gpu,
        "device": device_name
    }

@app.post("/api/upload")
def upload_to_hub(repo_id: str = "JOY0021/autonomy-agent-v2"):
    """Pushes the trained folder to the HF Hub model repo, creating it if needed."""
    try:
        import os
        from huggingface_hub import HfApi, create_repo
        
        token = os.getenv("HF_TOKEN")
        api = HfApi(token=token)
        
        # 1. Create repo if it doesn't exist
        print(f"πŸ“¦ Ensuring repo {repo_id} exists...")
        create_repo(repo_id=repo_id, repo_type="model", exist_ok=True, token=token)
        
        # 2. Upload the folder
        print(f"πŸ“‘ Uploading autonomy-agent-v2 to {repo_id}...")
        api.upload_folder(
            folder_path="autonomy-agent-v2",
            repo_id=repo_id,
            repo_type="model",
        )
        return {"status": "success", "message": f"Model live at https://huggingface.co/{repo_id}"}
    except Exception as e:
        print(f"❌ Upload Error: {e}")
        return {"status": "error", "message": str(e)}


# ─── API: Health ─────────────────────────────────────────────────────────────

@app.get("/api/health")
def health():
    difficulty_dist = Counter(s["difficulty"] for s in SCENARIOS)
    decision_dist = Counter(s["best_decision"] for s in SCENARIOS)
    return {
        "status": "ok",
        "environment": "autonomy-calibration-env",
        "version": "2.0.0",
        "tasks": list(TASK_REGISTRY.keys()),
        "autonomy_action_space": ["ACT", "ASK", "STOP", "RECOVER"],
        "autonomy_scenarios": len(SCENARIOS),
        "autonomy_difficulty_distribution": dict(difficulty_dist),
        "autonomy_decision_distribution": dict(decision_dist),
        "reward_range": [0.01, 0.99],
    }


# ─── API: History ─────────────────────────────────────────────────────────────

@app.get("/api/history")
def history():
    total = len(_episode_log)
    scores = [e["episode_score"] for e in _episode_log if e.get("episode_score") is not None]
    return {
        "total_episodes": total,
        "avg_score": round(sum(scores) / len(scores), 4) if scores else 0.0,
        "episodes": _episode_log,
    }


@app.delete("/api/history")
def clear_history():
    _episode_log.clear()
    return {"status": "cleared"}


# ─── API: Observability (Step 4) ─────────────────────────────────────────────

@app.get("/api/episodes")
def episodes(limit: int = 20):
    """List recent episodes from SQLite with metadata."""
    import database as db
    try:
        rows = db.list_episodes(limit=limit)
        return {"episodes": rows, "count": len(rows)}
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))


@app.get("/api/replay/{episode_id}")
def replay(episode_id: int):
    """
    Return the full step history for a past episode.
    Can be used to reproduce the episode by feeding steps back into reset + step.
    """
    import database as db
    try:
        data = db.get_episode(episode_id)
        return {
            "episode_id": episode_id,
            "episode": data["episode"],
            "steps": data["steps"],
            "total_steps": len(data["steps"]),
        }
    except ValueError as e:
        raise HTTPException(status_code=404, detail=str(e))
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))


@app.get("/api/grade")
def grade_current():
    """Run deterministic grader on the current session's episode history."""
    task = _session.get("task")
    if task is None:
        raise HTTPException(status_code=400, detail="No active episode.")
    score = task.grade_episode(_session["history"])
    return {
        "task": _session["task_name"],
        "seed": _session["seed"],
        "episode_id": _session["episode_id"],
        "score": score,
        "steps_completed": _session["step"],
        "done": _session["done"],
    }


@app.get("/api/grade/{episode_id}")
def grade_episode(episode_id: int):
    """Run deterministic grader on a completed historical episode."""
    import database as db
    try:
        data = db.get_episode(episode_id)
        ep = data["episode"]
        steps = data["steps"]
        # Reconstruct history format expected by grade_episode()
        history = [
            {"step": s["step_index"], "action": s["decision"],
             "reward": {"value": s["reward"]}}
            for s in steps
        ]
        total_reward = sum(s["reward"] for s in steps)
        from utils import clamp
        score = clamp(total_reward)
        return {
            "episode_id": episode_id,
            "task": ep["task"],
            "seed": ep["seed"],
            "score": score,
            "total_steps": len(steps),
            "started_at": ep["started_at"],
            "ended_at": ep["ended_at"],
            "steps": steps,
        }
    except ValueError as e:
        raise HTTPException(status_code=404, detail=str(e))
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))



# ─── Static UI ────────────────────────────────────────────────────────────────

if os.path.exists("static"):
    app.mount("/static", StaticFiles(directory="static"), name="static")

    @app.get("/")
    def serve_ui():
        return FileResponse("static/index.html")
else:
    @app.get("/")
    def serve_fallback():
        return {
            "message": "Autonomy Calibration Environment API v1.0",
            "docs": "/docs",
            "tasks": list(TASK_REGISTRY.keys()),
        }
def main():
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=7860)

if __name__ == "__main__":
    main()