Spaces:
Sleeping
Sleeping
File size: 15,583 Bytes
dd87944 | 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 | #!/usr/bin/env python3
"""
Boucle autonome de test des capacités Émo (think/todo/plan + exécution agent).
Usage:
py -3.11 scripts/emo_capability_loop.py
py -3.11 scripts/emo_capability_loop.py --project task_platform --round 2
py -3.11 scripts/emo_capability_loop.py --fix-report-only
Écrit les rapports dans scripts/capability_runs/<run_id>/
"""
from __future__ import annotations
import argparse
import json
import os
import re
import sys
import time
import uuid
from datetime import datetime, timezone
from pathlib import Path
from typing import Any
import httpx
ROOT = Path(__file__).resolve().parents[1]
# Windows/Python often lacks CA bundle — allow opt-out via EMO_SSL_VERIFY=1
_SSL_VERIFY = os.environ.get("EMO_SSL_VERIFY", "0").strip().lower() in ("1", "true", "yes")
def _http_client(**kwargs) -> httpx.Client:
"""httpx client — Windows often fails CA verify; default to False unless EMO_SSL_VERIFY=1."""
verify: bool | str = _SSL_VERIFY
if verify is True:
try:
import certifi
verify = certifi.where()
except ImportError:
verify = True
return httpx.Client(verify=verify, follow_redirects=True, **kwargs)
sys.path.insert(0, str(ROOT / "emo" / "backend"))
try:
import ssl_fix # noqa: F401
except ImportError:
pass
CONFIG_PATH = ROOT / "scripts" / "emo_capability_projects.json"
RUNS_DIR = ROOT / "scripts" / "capability_runs"
def _load_config() -> dict:
return json.loads(CONFIG_PATH.read_text(encoding="utf-8"))
def _run_id() -> str:
return datetime.now(timezone.utc).strftime("%Y%m%d_%H%M%S")
def login(client: httpx.Client, base: str, email: str, password: str) -> str:
r = client.post(f"{base}/api/auth/login", json={"email": email, "password": password}, timeout=120)
r.raise_for_status()
data = r.json()
token = data.get("session_token") or client.cookies.get("session_token")
if not token:
# Bearer from body if API returns it
token = data.get("token")
if not token:
raise RuntimeError("Login OK but no session token")
return token
def agent_status(client: httpx.Client, base: str, token: str) -> dict:
r = client.get(
f"{base}/api/agent/status",
headers={"Authorization": f"Bearer {token}", "X-Emo-Session": token},
timeout=30,
)
if r.status_code == 200:
return r.json()
return {"online": False, "error": r.text[:200]}
def create_conversation(client: httpx.Client, base: str, token: str, title: str) -> str:
r = client.post(
f"{base}/api/conversations",
json={"title": title[:80], "mode": "tech"},
headers={"Authorization": f"Bearer {token}", "X-Emo-Session": token},
timeout=60,
)
r.raise_for_status()
return r.json()["conversation_id"]
def stream_chat(
client: httpx.Client,
base: str,
token: str,
*,
conversation_id: str,
content: str,
project_path: str,
timeout_sec: int,
) -> dict[str, Any]:
headers = {
"Authorization": f"Bearer {token}",
"X-Emo-Session": token,
"Accept": "text/event-stream",
"Content-Type": "application/json",
}
body = {
"conversation_id": conversation_id,
"content": content,
"mode": "tech",
"model_preference": "auto",
"use_agent_tools": True,
"agent_project_path": project_path,
}
metrics: dict[str, Any] = {
"events": [],
"tools": [],
"thinks": [],
"todos": [],
"errors": [],
"gates": [],
"done": False,
"assistant_preview": "",
}
start = time.time()
with client.stream(
"POST",
f"{base}/api/chat/stream",
json=body,
headers=headers,
timeout=httpx.Timeout(timeout_sec, connect=120),
) as resp:
if resp.status_code != 200:
metrics["errors"].append(f"HTTP {resp.status_code}: {resp.read().decode()[:500]}")
return metrics
buf = ""
for chunk in resp.iter_text():
if time.time() - start > timeout_sec:
metrics["errors"].append("stream_timeout")
break
buf += chunk
while "\n\n" in buf:
block, buf = buf.split("\n\n", 1)
for line in block.split("\n"):
if not line.startswith("data:"):
continue
raw = line[5:].strip()
if not raw:
continue
try:
evt = json.loads(raw)
except json.JSONDecodeError:
continue
metrics["events"].append(evt.get("type"))
t = evt.get("type")
if t == "delta":
metrics["assistant_preview"] += evt.get("content") or ""
elif t == "tool_result":
tool = {
"name": evt.get("name"),
"ok": evt.get("result", {}).get("ok"),
"error": evt.get("result", {}).get("error"),
}
metrics["tools"].append(tool)
res = evt.get("result") or {}
if res.get("planning_gate") or res.get("think_gate"):
metrics["gates"].append({"tool": evt.get("name"), **res})
if res.get("ok") is False:
metrics["errors"].append(f"{evt.get('name')}: {res.get('error', '')[:200]}")
elif t == "think":
metrics["thinks"].append(evt)
elif t == "todo_update":
metrics["todos"].append(evt)
elif t == "error":
metrics["errors"].append(evt.get("content", "error"))
elif t == "done":
metrics["done"] = True
metrics["message_id"] = evt.get("message_id")
metrics["assistant_preview"] = metrics["assistant_preview"][:4000]
metrics["elapsed_sec"] = round(time.time() - start, 1)
return metrics
def score_project(metrics: dict, criteria: dict, project_path: Path, agent_online: bool) -> dict:
thinks = len(metrics.get("thinks") or [])
tools_ok = sum(1 for t in metrics.get("tools") or [] if t.get("ok") is not False)
tools_fail = sum(1 for t in metrics.get("tools") or [] if t.get("ok") is False)
todo_events = metrics.get("todos") or []
max_todos = max((len(e.get("todos") or []) for e in todo_events), default=0)
finalized = any(e.get("planning_complete") for e in todo_events)
set_plan = any(e.get("action") == "set_plan" for e in todo_events)
points = 0
max_pts = 100
notes: list[str] = []
if thinks >= criteria.get("min_thinks", 2):
points += 20
else:
notes.append(f"Pas assez de emo_think ({thinks}/{criteria.get('min_thinks', 2)})")
if max_todos >= criteria.get("min_todos_set", 5):
points += 20
else:
notes.append(f"Todo plan insuffisant ({max_todos} tâches)")
if not criteria.get("require_finalize_plan") or finalized:
points += 15
else:
notes.append("finalize_plan jamais appelé")
if set_plan:
points += 10
else:
notes.append("set_plan jamais appelé")
if tools_ok >= criteria.get("min_successful_tools", 8):
points += 15
else:
notes.append(f"Peu d'outils OK ({tools_ok})")
if metrics.get("done"):
points += 10
else:
notes.append("Stream non terminé (done absent)")
if not metrics.get("errors"):
points += 10
else:
notes.append(f"{len(metrics['errors'])} erreur(s) tool/stream")
files_found: list[str] = []
if project_path.is_dir():
for p in project_path.rglob("*"):
if p.is_file():
rel = p.relative_to(project_path).as_posix()
files_found.append(rel)
expected = criteria.get("expected_files_any") or []
if expected:
hit = any(
any(rel.endswith(ex) or ex in rel for rel in files_found)
for ex in expected
)
if hit:
points += 10
elif agent_online:
notes.append(f"Fichiers attendus absents: {expected}")
else:
notes.append("Agent offline — fichiers non vérifiables (+0)")
else:
points += 10
gate_after_finalize = finalized and any(
g.get("planning_gate") for g in metrics.get("gates") or []
)
if criteria.get("require_no_planning_gate_after_finalize") and gate_after_finalize:
points -= 10
notes.append("Planning gate encore actif après finalize")
return {
"score": max(0, min(max_pts, points)),
"notes": notes,
"thinks": thinks,
"tools_ok": tools_ok,
"tools_fail": tools_fail,
"max_todos": max_todos,
"finalized": finalized,
"files_count": len(files_found),
"files_sample": files_found[:25],
}
def follow_up_message(score: dict, round_idx: int) -> str:
notes = score.get("notes") or []
if round_idx == 1:
return (
"Continue le projet. Reprends PROJECT.md et la todo list. "
"emo_think avant chaque write_file/exec_shell. Complete les tâches restantes."
)
issues = "; ".join(notes[:5]) if notes else "qualité insuffisante"
return (
f"Round {round_idx}: corrige ces problèmes — {issues}. "
"Reprends le plan, mets à jour emo_todo, exécute sans skip think."
)
def run_project(
cfg: dict,
project: dict,
run_dir: Path,
*,
max_rounds: int,
timeout_sec: int,
min_score: int,
) -> dict:
defaults = cfg["defaults"]
base = defaults["backend_url"].rstrip("/")
email = defaults["email"]
password = defaults["password"]
root = Path(defaults["project_root"])
project_path = root / project["folder"]
project_path.mkdir(parents=True, exist_ok=True)
out: dict[str, Any] = {
"project_id": project["id"],
"name": project["name"],
"project_path": str(project_path),
"rounds": [],
"passed": False,
}
with _http_client() as client:
token = login(client, base, email, password)
client.headers.update({"Authorization": f"Bearer {token}", "X-Emo-Session": token})
ag = agent_status(client, base, token)
out["agent_online"] = bool(ag.get("online"))
conv_id = create_conversation(client, base, token, f"[TEST] {project['name']}")
out["conversation_id"] = conv_id
content = project["prompt"]
for rnd in range(1, max_rounds + 1):
print(f"\n=== {project['id']} round {rnd}/{max_rounds} ===", flush=True)
metrics = stream_chat(
client, base, token,
conversation_id=conv_id,
content=content,
project_path=str(project_path),
timeout_sec=timeout_sec,
)
score = score_project(metrics, project.get("criteria", {}), project_path, out["agent_online"])
round_doc = {
"round": rnd,
"prompt": content[:500],
"metrics_summary": {
"elapsed_sec": metrics.get("elapsed_sec"),
"done": metrics.get("done"),
"tools": len(metrics.get("tools") or []),
"thinks": len(metrics.get("thinks") or []),
"errors": metrics.get("errors")[:10],
},
"score": score,
}
out["rounds"].append(round_doc)
(run_dir / f"{project['id']}_round{rnd}.json").write_text(
json.dumps({"metrics": metrics, "score": score}, indent=2, ensure_ascii=False),
encoding="utf-8",
)
print(f"Score: {score['score']}/100 — {score.get('notes')}", flush=True)
if score["score"] >= min_score:
out["passed"] = True
out["final_score"] = score["score"]
break
content = follow_up_message(score, rnd)
return out
def suggest_code_fixes(results: list[dict]) -> list[str]:
"""Heuristiques pour corrections Émo si scores bas."""
fixes: list[str] = []
for r in results:
if r.get("passed"):
continue
for rnd in r.get("rounds") or []:
sc = rnd.get("score") or {}
for note in sc.get("notes") or []:
if "emo_think" in note:
fixes.append("Renforcer prompt AGENT_COGNITION + gate think")
if "finalize_plan" in note:
fixes.append("Auto-inject set_plan skeleton on mega project start")
if "Agent offline" in note:
fixes.append("WARN: agent local requis — boucle ne peut pas valider fichiers")
if "Stream non terminé" in note:
fixes.append("Augmenter timeout frontend ou max_agent_rounds backend")
return list(dict.fromkeys(fixes))
def main() -> int:
parser = argparse.ArgumentParser(description="Boucle test capacités Émo")
parser.add_argument("--project", help="ID projet (sinon tous)")
parser.add_argument("--round", type=int, default=0, help="Max rounds override")
parser.add_argument("--fix-report-only", action="store_true")
args = parser.parse_args()
cfg = _load_config()
defaults = cfg["defaults"]
run_id = _run_id()
run_dir = RUNS_DIR / run_id
run_dir.mkdir(parents=True, exist_ok=True)
projects = cfg["projects"]
if args.project:
projects = [p for p in projects if p["id"] == args.project]
if not projects:
print(f"Projet inconnu: {args.project}", file=sys.stderr)
return 1
max_rounds = args.round or defaults["max_rounds_per_project"]
results: list[dict] = []
if not args.fix_report_only:
for proj in projects:
try:
results.append(
run_project(
cfg, proj, run_dir,
max_rounds=max_rounds,
timeout_sec=defaults["stream_timeout_sec"],
min_score=defaults["min_score_to_pass"],
)
)
except Exception as e:
results.append({
"project_id": proj["id"],
"passed": False,
"fatal_error": str(e),
})
report = {
"run_id": run_id,
"started_at": datetime.now(timezone.utc).isoformat(),
"results": results,
"suggested_fixes": suggest_code_fixes(results),
"passed_count": sum(1 for r in results if r.get("passed")),
"total": len(projects) if not args.fix_report_only else 0,
}
report_path = run_dir / "report.json"
report_path.write_text(json.dumps(report, indent=2, ensure_ascii=False), encoding="utf-8")
print(f"\nRapport: {report_path}", flush=True)
print(f"Passés: {report['passed_count']}/{len(projects)}", flush=True)
if report["suggested_fixes"]:
print("Fixes suggérés:", report["suggested_fixes"], flush=True)
return 0 if report["passed_count"] == len(projects) else 1
if __name__ == "__main__":
raise SystemExit(main())
|