Spaces:
Running
Running
File size: 40,771 Bytes
1ccf652 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 | """
aureon_code_integrator.py — The organism writes its own source, under review.
When the Code Architect authors a change that targets existing aureon source
(not a new skill in the library, but an edit to a .py file already in the
tree), the change does not land directly. It lands *here*:
1. PROPOSE the change. We locate `old_text` inside `target_path`,
build the would-be resulting file in memory, and run a
static syntax check (ast.parse + compile) against it.
A pending record is written to
state/integrations_pending/<id>.json
with the full before-snapshot and the new text.
2. CONFIRM the change by id. We take a backup of the current file
under state/code_integrator_backups/<path>/<ts>.bak,
write the new content, run one more compile on disk,
and append an audit line to
state/integrations_applied.jsonl
with the unified diff. If the on-disk compile fails we
restore from backup and record the failure.
3. REJECT the change by id. The pending record is moved to
state/integrations_rejected.jsonl. No file is touched.
4. LIST pending edits.
5. HISTORY recent applied / rejected edits (audit log).
The module never applies a change without an explicit `confirm_edit`
call. It never writes outside the aureon-trading repository root. It
refuses to touch files under `state/` (runtime scratch) or anywhere
outside `aureon/`, `scripts/`, or repo-root `.py` files by default.
Typical flow from outside:
ci = get_code_integrator()
prop = ci.propose_edit(
target_path="aureon/core/aureon_lambda_engine.py",
old_text="PHI = 1.618",
new_text="PHI = 1.6180339887498948",
rationale="tighten constant to 16 digits",
)
# inspect prop["before_snapshot"], prop["after_preview"], prop["syntax_ok"]
ci.confirm_edit(prop["pending_id"])
# the file is now modified and the change is in state/integrations_applied.jsonl
"""
from __future__ import annotations
import ast
import difflib
import hashlib
import json
import logging
import py_compile
import shutil
import threading
import time
import uuid
from dataclasses import dataclass, field
from pathlib import Path
from typing import Any, Dict, List, Optional, Tuple
logger = logging.getLogger("aureon.core.code_integrator")
_REPO_ROOT = Path(__file__).resolve().parents[2]
_STATE_ROOT = _REPO_ROOT / "state"
_PENDING_DIR = _STATE_ROOT / "integrations_pending"
_APPLIED_LOG = _STATE_ROOT / "integrations_applied.jsonl"
_REJECTED_LOG = _STATE_ROOT / "integrations_rejected.jsonl"
_BACKUP_ROOT = _STATE_ROOT / "code_integrator_backups"
_IMPROVEMENT_LOG = _STATE_ROOT / "improvement_log.jsonl"
# Paths the integrator is permitted to touch, relative to repo root. These
# are prefixes — `aureon/` covers every module under aureon-trading/aureon/.
_DEFAULT_ALLOWED_PREFIXES: Tuple[str, ...] = (
"aureon/",
"scripts/",
)
# Paths the integrator refuses to touch even if an allowed prefix also matches.
_DEFAULT_DENIED_PREFIXES: Tuple[str, ...] = (
"state/",
".git/",
"__pycache__/",
)
# ─────────────────────────────────────────────────────────────────────────────
# Helpers
# ─────────────────────────────────────────────────────────────────────────────
def _now() -> float:
return time.time()
def _relpath(p: Path) -> str:
try:
return str(p.resolve().relative_to(_REPO_ROOT)).replace("\\", "/")
except Exception:
return str(p)
def _resolve_repo_path(target_path: str) -> Path:
"""Resolve `target_path` to an absolute path inside the repo, or raise."""
candidate = Path(target_path)
if not candidate.is_absolute():
candidate = _REPO_ROOT / candidate
candidate = candidate.resolve()
try:
candidate.relative_to(_REPO_ROOT)
except ValueError:
raise ValueError(f"path escapes repo root: {target_path}")
return candidate
def _check_allowed(target_path: str,
allowed: Tuple[str, ...],
denied: Tuple[str, ...]) -> None:
rel = target_path.replace("\\", "/").lstrip("./")
for bad in denied:
if rel.startswith(bad):
raise PermissionError(f"path under denied prefix {bad!r}: {rel}")
# Root-level .py files are permitted even without a prefix match.
if "/" not in rel and rel.endswith(".py"):
return
for good in allowed:
if rel.startswith(good):
return
raise PermissionError(
f"path {rel!r} is outside allowed prefixes {allowed!r}"
)
def _sha256(text: str) -> str:
return hashlib.sha256(text.encode("utf-8", errors="replace")).hexdigest()
def _unified_diff(before: str, after: str, rel: str) -> str:
diff = difflib.unified_diff(
before.splitlines(keepends=True),
after.splitlines(keepends=True),
fromfile=f"a/{rel}",
tofile=f"b/{rel}",
n=3,
)
return "".join(diff)
# ─────────────────────────────────────────────────────────────────────────────
# Dataclasses
# ─────────────────────────────────────────────────────────────────────────────
@dataclass
class PendingEdit:
pending_id: str
target_path: str # repo-relative
rationale: str
before_text: str
after_text: str
before_sha: str
after_sha: str
syntax_ok: bool
syntax_error: str
proposed_at: float
match_count: int # how many times old_text appeared in before_text
def preview(self, lines: int = 12) -> str:
diff = _unified_diff(self.before_text, self.after_text, self.target_path)
snippet = diff.splitlines()[:lines]
return "\n".join(snippet)
def to_dict(self) -> Dict[str, Any]:
return {
"pending_id": self.pending_id,
"target_path": self.target_path,
"rationale": self.rationale,
"before_sha": self.before_sha,
"after_sha": self.after_sha,
"syntax_ok": self.syntax_ok,
"syntax_error": self.syntax_error,
"proposed_at": self.proposed_at,
"match_count": self.match_count,
}
# ─────────────────────────────────────────────────────────────────────────────
# The integrator
# ─────────────────────────────────────────────────────────────────────────────
class CodeIntegrator:
"""
Holds a pending queue of proposed source edits, syntax-checks them,
and applies them only on explicit confirm.
"""
def __init__(
self,
allowed_prefixes: Tuple[str, ...] = _DEFAULT_ALLOWED_PREFIXES,
denied_prefixes: Tuple[str, ...] = _DEFAULT_DENIED_PREFIXES,
) -> None:
self.allowed_prefixes = tuple(allowed_prefixes)
self.denied_prefixes = tuple(denied_prefixes)
self._lock = threading.Lock()
_PENDING_DIR.mkdir(parents=True, exist_ok=True)
_BACKUP_ROOT.mkdir(parents=True, exist_ok=True)
# ------------------------------------------------------------------
# Propose
# ------------------------------------------------------------------
def propose_edit(
self,
target_path: str,
old_text: str,
new_text: str,
rationale: str = "",
) -> Dict[str, Any]:
"""
Build a pending edit. Does NOT touch the file on disk.
Returns:
{
"ok": bool,
"pending_id": "...",
"target_path": "aureon/...",
"match_count": int,
"before_sha": "...",
"after_sha": "...",
"syntax_ok": bool,
"syntax_error": str,
"preview": "<unified diff, trimmed>"
}
"""
try:
_check_allowed(target_path, self.allowed_prefixes, self.denied_prefixes)
abs_path = _resolve_repo_path(target_path)
except (PermissionError, ValueError) as e:
return {"ok": False, "error": str(e)}
if not abs_path.exists():
return {"ok": False, "error": f"file not found: {target_path}"}
try:
before_text = abs_path.read_text(encoding="utf-8")
except Exception as e:
return {"ok": False, "error": f"read failed: {e}"}
if not old_text:
return {"ok": False, "error": "old_text cannot be empty"}
match_count = before_text.count(old_text)
if match_count == 0:
return {
"ok": False,
"error": "old_text not found in target file",
"match_count": 0,
}
if match_count > 1:
return {
"ok": False,
"error": (
f"old_text matches {match_count} times — refusing "
"non-unique edit. Widen old_text to include surrounding "
"context so the match becomes unique."
),
"match_count": match_count,
}
# Build the would-be file.
after_text = before_text.replace(old_text, new_text, 1)
# Syntax check the *would-be* content.
syntax_ok, syntax_err = self._syntax_check(after_text, target_path)
pending_id = f"pe_{int(_now() * 1000)}_{uuid.uuid4().hex[:8]}"
pending = PendingEdit(
pending_id=pending_id,
target_path=_relpath(abs_path),
rationale=rationale or "",
before_text=before_text,
after_text=after_text,
before_sha=_sha256(before_text),
after_sha=_sha256(after_text),
syntax_ok=syntax_ok,
syntax_error=syntax_err,
proposed_at=_now(),
match_count=match_count,
)
with self._lock:
self._write_pending(pending)
return {
"ok": True,
"pending_id": pending_id,
"target_path": pending.target_path,
"match_count": match_count,
"before_sha": pending.before_sha,
"after_sha": pending.after_sha,
"syntax_ok": syntax_ok,
"syntax_error": syntax_err,
"preview": pending.preview(lines=24),
}
# ------------------------------------------------------------------
# Confirm / reject
# ------------------------------------------------------------------
def confirm_edit(self, pending_id: str) -> Dict[str, Any]:
"""Apply the pending edit after backing up the current file."""
with self._lock:
pending = self._load_pending(pending_id)
if pending is None:
return {"ok": False, "error": f"no such pending edit: {pending_id}"}
if not pending.syntax_ok:
return {
"ok": False,
"error": "refusing to apply: syntax check failed",
"syntax_error": pending.syntax_error,
}
try:
abs_path = _resolve_repo_path(pending.target_path)
except Exception as e:
return {"ok": False, "error": f"path resolve: {e}"}
if not abs_path.exists():
return {"ok": False, "error": "target file disappeared"}
# Guard: before_sha must still match current disk contents.
try:
current = abs_path.read_text(encoding="utf-8")
except Exception as e:
return {"ok": False, "error": f"re-read failed: {e}"}
if _sha256(current) != pending.before_sha:
return {
"ok": False,
"error": (
"target file changed since the edit was proposed — "
"re-propose against the new file state."
),
}
# Backup.
backup_path = self._backup(abs_path, current)
# Capture metrics BEFORE the write so we can measure positive
# improvement after the apply. Improvement is the organism's
# signal that the edit actually did something.
metrics_before = self._capture_metrics()
# Write and compile.
try:
abs_path.write_text(pending.after_text, encoding="utf-8")
except Exception as e:
return {"ok": False, "error": f"write failed: {e}"}
compile_ok, compile_err = self._disk_compile(abs_path)
if not compile_ok:
# Roll back.
try:
abs_path.write_text(current, encoding="utf-8")
except Exception:
pass
return {
"ok": False,
"error": "applied file failed on-disk compile — rolled back",
"compile_error": compile_err,
"backup_path": str(backup_path),
}
metrics_after = self._capture_metrics()
improvement = self._metrics_delta(metrics_before, metrics_after)
# File-level deltas from the pending text itself — sensitive to
# edits that activate dormant modules (new imports, new calls)
# even when the tree-wide scan shows no change.
file_delta = self._file_delta(pending.before_text, pending.after_text)
improvement["file"] = file_delta
# Upgrade verdict if file-level change is clearly positive.
if improvement.get("verdict") == "no-op" and (
file_delta.get("lines_added", 0) > 0
or file_delta.get("imports_added", 0) > 0
or file_delta.get("calls_added", 0) > 0
):
improvement["verdict"] = "positive"
improvement["verdict_source"] = "file_delta"
# If this pending was part of a comparison, auto-reject the sibling.
comparison_id, variant, sibling_id = self._read_comparison_meta(pending_id)
# Audit.
diff = _unified_diff(pending.before_text, pending.after_text, pending.target_path)
record = {
"applied_at": _now(),
"pending_id": pending.pending_id,
"target_path": pending.target_path,
"rationale": pending.rationale,
"before_sha": pending.before_sha,
"after_sha": pending.after_sha,
"backup_path": _relpath(backup_path),
"diff": diff,
"metrics_before": metrics_before,
"metrics_after": metrics_after,
"improvement": improvement,
}
if comparison_id:
record["comparison_id"] = comparison_id
record["variant"] = variant
record["sibling_pending_id"] = sibling_id
self._append_log(_APPLIED_LOG, record)
# Dedicated improvement log — tails here to see organism progress.
self._append_log(_IMPROVEMENT_LOG, {
"applied_at": record["applied_at"],
"pending_id": pending_id,
"target_path": pending.target_path,
"improvement": improvement,
"comparison_id": comparison_id or None,
"variant": variant or None,
})
self._delete_pending(pending_id)
sibling_rejection: Optional[Dict[str, Any]] = None
if comparison_id and sibling_id:
# Don't hold _lock recursively — sibling rejection doesn't
# touch disk state beyond our own log files.
sibling = self._load_pending(sibling_id)
if sibling is not None:
sib_record = {
"rejected_at": _now(),
"pending_id": sibling.pending_id,
"target_path": sibling.target_path,
"rationale": sibling.rationale,
"reason": f"lost comparison {comparison_id} to variant={variant}",
"comparison_id": comparison_id,
"sibling_pending_id": pending_id,
"before_sha": sibling.before_sha,
"after_sha": sibling.after_sha,
}
self._append_log(_REJECTED_LOG, sib_record)
self._delete_pending(sibling_id)
sibling_rejection = {
"sibling_pending_id": sibling_id,
"reason": sib_record["reason"],
}
result = {
"ok": True,
"applied_at": record["applied_at"],
"pending_id": pending_id,
"target_path": pending.target_path,
"backup_path": _relpath(backup_path),
"diff": diff,
"improvement": improvement,
}
if comparison_id:
result["comparison_id"] = comparison_id
result["variant"] = variant
if sibling_rejection:
result["sibling_rejected"] = sibling_rejection
# Self-report: if no positive improvement, the loop is not doing
# useful work — flag loudly so the reviewer sees it.
verdict = improvement.get("verdict", "unknown")
if verdict in ("no-op", "regression"):
logger.warning(
"edit applied but improvement verdict=%s on %s — loop may not be working",
verdict, pending.target_path,
)
else:
logger.info(
"edit applied: %s verdict=%s cls_delta=%s fn_delta=%s dp_delta=%s",
pending.target_path,
verdict,
improvement.get("classes_delta"),
improvement.get("functions_delta"),
improvement.get("decision_points_delta"),
)
# Continuity: hot-swap the module without the organism
# perceiving a restart. Phi-calendar counts + singletons are
# preserved across importlib.reload. Subjective time unbroken.
try:
from aureon.core.aureon_continuity_engine import get_continuity_engine
ce = get_continuity_engine()
swap = ce.apply_with_continuity(
target_path=pending.target_path,
pending_id=pending.pending_id,
triggered_by="integrator.confirm",
)
result["continuity"] = {
"ok": swap.ok,
"sandbox_ok": swap.sandbox_ok,
"sandbox_detail": swap.sandbox_detail,
"reload_ok": swap.reload_ok,
"reload_error": swap.reload_error,
"continuity_preserved": swap.continuity_preserved,
"elapsed_s": swap.elapsed_s,
}
except Exception as e:
# Continuity is best-effort. The file is already on disk;
# fresh imports will pick it up regardless. Log softly.
logger.debug("continuity engine unavailable: %s", e)
return result
def propose_comparison(
self,
target_path: str,
old_text: str,
aureon_new_text: str,
human_new_text: str,
rationale: str = "",
) -> Dict[str, Any]:
"""
Propose two competing edits for the same site so a reviewer can see
whether Aureon's version beats a human baseline.
Both variants are stored as independent pending edits that share
a `comparison_id`. When one is confirmed, the other is auto-rejected
with reason "lost comparison <id>" so the applied audit line can
always be traced back to what the organism chose over what a human
would have written.
Returns:
{
"ok": bool,
"comparison_id": "cmp_...",
"aureon": { <propose_edit result> },
"human": { <propose_edit result> },
"side_by_side": "<aureon diff>\\n---\\n<human diff>"
}
"""
comparison_id = f"cmp_{int(_now() * 1000)}_{uuid.uuid4().hex[:8]}"
aureon_res = self.propose_edit(
target_path=target_path,
old_text=old_text,
new_text=aureon_new_text,
rationale=f"[comparison {comparison_id} variant=aureon] {rationale}",
)
human_res = self.propose_edit(
target_path=target_path,
old_text=old_text,
new_text=human_new_text,
rationale=f"[comparison {comparison_id} variant=human] {rationale}",
)
# Tag each pending record with its sibling + comparison id so the
# audit trail can find the pair later.
for res, variant, sibling in (
(aureon_res, "aureon", human_res),
(human_res, "human", aureon_res),
):
pid = res.get("pending_id")
if not pid:
continue
try:
fp = self._pending_file(pid)
d = json.loads(fp.read_text(encoding="utf-8"))
d["comparison_id"] = comparison_id
d["variant"] = variant
d["sibling_pending_id"] = sibling.get("pending_id")
fp.write_text(json.dumps(d, indent=2), encoding="utf-8")
except Exception as e:
logger.debug("comparison tag failed: %s", e)
# Side-by-side preview (trimmed).
aureon_preview = aureon_res.get("preview", "") or ""
human_preview = human_res.get("preview", "") or ""
side_by_side = (
f"=== AUREON ({aureon_res.get('pending_id', '?')}) ===\n"
f"syntax_ok={aureon_res.get('syntax_ok')} "
f"err={aureon_res.get('syntax_error') or ''}\n"
f"{aureon_preview}\n"
f"\n=== HUMAN ({human_res.get('pending_id', '?')}) ===\n"
f"syntax_ok={human_res.get('syntax_ok')} "
f"err={human_res.get('syntax_error') or ''}\n"
f"{human_preview}\n"
)
return {
"ok": bool(aureon_res.get("ok") or human_res.get("ok")),
"comparison_id": comparison_id,
"aureon": aureon_res,
"human": human_res,
"side_by_side": side_by_side,
}
def reject_edit(self, pending_id: str, reason: str = "") -> Dict[str, Any]:
"""Discard a pending edit; write a rejection audit line."""
with self._lock:
pending = self._load_pending(pending_id)
if pending is None:
return {"ok": False, "error": f"no such pending edit: {pending_id}"}
record = {
"rejected_at": _now(),
"pending_id": pending.pending_id,
"target_path": pending.target_path,
"rationale": pending.rationale,
"reason": reason or "",
"before_sha": pending.before_sha,
"after_sha": pending.after_sha,
}
self._append_log(_REJECTED_LOG, record)
self._delete_pending(pending_id)
return {"ok": True, "pending_id": pending_id}
# ------------------------------------------------------------------
# Listing / history
# ------------------------------------------------------------------
def list_pending(self) -> List[Dict[str, Any]]:
result: List[Dict[str, Any]] = []
if not _PENDING_DIR.exists():
return result
for p in sorted(_PENDING_DIR.glob("*.json")):
try:
data = json.loads(p.read_text(encoding="utf-8"))
# Strip bulk text from the listing; caller can fetch full
# via inspect_pending(id).
light = {
k: v
for k, v in data.items()
if k not in ("before_text", "after_text")
}
result.append(light)
except Exception:
continue
return result
def inspect_pending(self, pending_id: str) -> Optional[Dict[str, Any]]:
pending = self._load_pending(pending_id)
if pending is None:
return None
d = pending.to_dict()
d["preview"] = pending.preview(lines=40)
return d
def history(self, limit: int = 20, kind: str = "applied") -> List[Dict[str, Any]]:
path = _APPLIED_LOG if kind == "applied" else _REJECTED_LOG
if not path.exists():
return []
try:
lines = path.read_text(encoding="utf-8").splitlines()
except Exception:
return []
recent = lines[-limit:] if limit > 0 else lines
out: List[Dict[str, Any]] = []
for ln in recent:
ln = ln.strip()
if not ln:
continue
try:
out.append(json.loads(ln))
except Exception:
continue
return out
# ------------------------------------------------------------------
# Internals
# ------------------------------------------------------------------
# ------------------------------------------------------------------
# Improvement capture — the "is this actually working?" signal
# ------------------------------------------------------------------
def _capture_metrics(self) -> Dict[str, Any]:
"""
Snapshot measurable state that can change from an edit. Returns a
compact dict. Safe to call on a live tree.
"""
metrics: Dict[str, Any] = {
"taken_at": _now(),
"modules_total": 0,
"decision_points_total": 0,
"decision_points_by_kind": {},
"classes_total": 0,
"functions_total": 0,
"loaded_aureon_modules": 0,
}
try:
from aureon.core.aureon_self_introspection import get_self_introspection
intro = get_self_introspection()
snap = intro.scan(fast=False)
by_kind: Dict[str, int] = {}
for dp in snap.decision_points:
by_kind[dp.kind] = by_kind.get(dp.kind, 0) + 1
cls_total = 0
fn_total = 0
for fp in snap.fingerprints:
cls_total += len(fp.classes)
fn_total += len(fp.functions)
metrics.update({
"modules_total": snap.modules_total,
"decision_points_total": len(snap.decision_points),
"decision_points_by_kind": by_kind,
"classes_total": cls_total,
"functions_total": fn_total,
"loaded_aureon_modules": len(intro.loaded_modules()),
})
except Exception as e:
metrics["error"] = f"capture failed: {e}"
return metrics
@staticmethod
def _file_delta(before_text: str, after_text: str) -> Dict[str, Any]:
"""
File-level delta between the before and after text of a single
pending edit. Counts lines, bytes, imports, and call expressions —
so an edit that activates dormant modules (new imports, new calls)
registers as positive even when the tree-wide scan shows no change.
"""
def _count_ast(text: str) -> Dict[str, int]:
out = {"imports": 0, "calls": 0, "classes": 0, "functions": 0}
try:
tree = ast.parse(text)
except Exception:
return out
for node in ast.walk(tree):
if isinstance(node, (ast.Import, ast.ImportFrom)):
# Count every imported name (import a, b → 2).
out["imports"] += len(getattr(node, "names", []) or [])
elif isinstance(node, ast.Call):
out["calls"] += 1
elif isinstance(node, ast.ClassDef):
out["classes"] += 1
elif isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):
out["functions"] += 1
return out
before_counts = _count_ast(before_text)
after_counts = _count_ast(after_text)
return {
"lines_added": after_text.count("\n") - before_text.count("\n"),
"bytes_added": len(after_text) - len(before_text),
"imports_added": after_counts["imports"] - before_counts["imports"],
"calls_added": after_counts["calls"] - before_counts["calls"],
"classes_added": after_counts["classes"] - before_counts["classes"],
"functions_added": after_counts["functions"] - before_counts["functions"],
}
@staticmethod
def _metrics_delta(before: Dict[str, Any], after: Dict[str, Any]) -> Dict[str, Any]:
"""
Compute a compact delta. Positive deltas on function/class counts
indicate new code was added. Changes to decision-point kinds tell
us whether gates/flags moved.
"""
def _num(d: Dict[str, Any], k: str) -> int:
v = d.get(k, 0)
try:
return int(v)
except Exception:
return 0
kinds_before = dict(before.get("decision_points_by_kind") or {})
kinds_after = dict(after.get("decision_points_by_kind") or {})
all_kinds = set(kinds_before) | set(kinds_after)
kinds_delta = {
k: int(kinds_after.get(k, 0)) - int(kinds_before.get(k, 0))
for k in sorted(all_kinds)
}
delta = {
"modules_total_delta": _num(after, "modules_total") - _num(before, "modules_total"),
"decision_points_delta": (
_num(after, "decision_points_total") - _num(before, "decision_points_total")
),
"decision_points_by_kind_delta": kinds_delta,
"classes_delta": _num(after, "classes_total") - _num(before, "classes_total"),
"functions_delta": _num(after, "functions_total") - _num(before, "functions_total"),
"loaded_aureon_modules_delta": (
_num(after, "loaded_aureon_modules") - _num(before, "loaded_aureon_modules")
),
}
# Simple positive-improvement verdict:
# - new code added (classes or functions up) → positive
# - modules_total up (file added) → positive
# - no change at all → flagged "no-op"
# - only regressions → flagged "regression"
positives = [
delta["classes_delta"],
delta["functions_delta"],
delta["modules_total_delta"],
delta["loaded_aureon_modules_delta"],
]
is_noop = all(v == 0 for v in positives) and delta["decision_points_delta"] == 0
has_positive = any(v > 0 for v in positives) or any(v > 0 for v in kinds_delta.values())
has_regression = any(v < 0 for v in positives)
if is_noop:
delta["verdict"] = "no-op"
elif has_positive and not has_regression:
delta["verdict"] = "positive"
elif has_positive and has_regression:
delta["verdict"] = "mixed"
elif has_regression:
delta["verdict"] = "regression"
else:
delta["verdict"] = "neutral"
return delta
def _read_comparison_meta(self, pending_id: str) -> Tuple[str, str, str]:
"""Return (comparison_id, variant, sibling_pending_id) or ("", "", "")."""
fp = self._pending_file(pending_id)
if not fp.exists():
return "", "", ""
try:
d = json.loads(fp.read_text(encoding="utf-8"))
except Exception:
return "", "", ""
return (
str(d.get("comparison_id") or ""),
str(d.get("variant") or ""),
str(d.get("sibling_pending_id") or ""),
)
def _syntax_check(self, text: str, target_path: str) -> Tuple[bool, str]:
# We only full-parse .py files; everything else gets a pass.
if not target_path.endswith(".py"):
return True, ""
try:
ast.parse(text, filename=target_path)
except SyntaxError as e:
return False, f"SyntaxError at line {e.lineno}: {e.msg}"
except Exception as e:
return False, f"parse failed: {e}"
return True, ""
def _disk_compile(self, path: Path) -> Tuple[bool, str]:
if path.suffix != ".py":
return True, ""
try:
py_compile.compile(str(path), doraise=True)
except py_compile.PyCompileError as e:
return False, str(e)
except Exception as e:
return False, f"py_compile error: {e}"
return True, ""
def _pending_file(self, pending_id: str) -> Path:
return _PENDING_DIR / f"{pending_id}.json"
def _write_pending(self, pending: PendingEdit) -> None:
data = {
"pending_id": pending.pending_id,
"target_path": pending.target_path,
"rationale": pending.rationale,
"before_text": pending.before_text,
"after_text": pending.after_text,
"before_sha": pending.before_sha,
"after_sha": pending.after_sha,
"syntax_ok": pending.syntax_ok,
"syntax_error": pending.syntax_error,
"proposed_at": pending.proposed_at,
"match_count": pending.match_count,
}
self._pending_file(pending.pending_id).write_text(
json.dumps(data, indent=2), encoding="utf-8"
)
def _load_pending(self, pending_id: str) -> Optional[PendingEdit]:
fp = self._pending_file(pending_id)
if not fp.exists():
return None
try:
d = json.loads(fp.read_text(encoding="utf-8"))
except Exception:
return None
return PendingEdit(
pending_id=d["pending_id"],
target_path=d["target_path"],
rationale=d.get("rationale", ""),
before_text=d["before_text"],
after_text=d["after_text"],
before_sha=d["before_sha"],
after_sha=d["after_sha"],
syntax_ok=bool(d.get("syntax_ok", False)),
syntax_error=d.get("syntax_error", ""),
proposed_at=float(d.get("proposed_at", 0.0)),
match_count=int(d.get("match_count", 0)),
)
def _delete_pending(self, pending_id: str) -> None:
try:
self._pending_file(pending_id).unlink(missing_ok=True)
except Exception:
pass
def _backup(self, abs_path: Path, content: str) -> Path:
ts = time.strftime("%Y%m%d_%H%M%S")
rel = _relpath(abs_path)
bk_dir = _BACKUP_ROOT / rel
bk_dir.mkdir(parents=True, exist_ok=True)
bk_path = bk_dir / f"{ts}.bak"
try:
bk_path.write_text(content, encoding="utf-8")
except Exception as e:
logger.warning("backup write failed (%s): %s", bk_path, e)
return bk_path
def _append_log(self, path: Path, record: Dict[str, Any]) -> None:
try:
path.parent.mkdir(parents=True, exist_ok=True)
with path.open("a", encoding="utf-8") as f:
f.write(json.dumps(record, default=str) + "\n")
except Exception as e:
logger.warning("log append failed (%s): %s", path, e)
# ─────────────────────────────────────────────────────────────────────────────
# Singleton
# ─────────────────────────────────────────────────────────────────────────────
_instance: Optional[CodeIntegrator] = None
_instance_lock = threading.Lock()
def get_code_integrator() -> CodeIntegrator:
global _instance
with _instance_lock:
if _instance is None:
_instance = CodeIntegrator()
return _instance
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser(description="Aureon code integrator CLI.")
sub = parser.add_subparsers(dest="cmd", required=True)
p_list = sub.add_parser("list", help="List pending edits")
p_hist = sub.add_parser("history", help="Show applied/rejected history")
p_hist.add_argument("--kind", choices=("applied", "rejected"), default="applied")
p_hist.add_argument("--limit", type=int, default=10)
p_show = sub.add_parser("show", help="Show full pending edit with diff preview")
p_show.add_argument("pending_id")
p_confirm = sub.add_parser("confirm", help="Confirm a pending edit by id")
p_confirm.add_argument("pending_id")
p_reject = sub.add_parser("reject", help="Reject a pending edit by id")
p_reject.add_argument("pending_id")
p_reject.add_argument("--reason", default="")
p_propose = sub.add_parser("propose", help="Propose an edit from file contents on stdin")
p_propose.add_argument("--target", required=True)
p_propose.add_argument("--old", required=True, help="Path to a file containing old_text")
p_propose.add_argument("--new", required=True, help="Path to a file containing new_text")
p_propose.add_argument("--rationale", default="")
args = parser.parse_args()
ci = get_code_integrator()
if args.cmd == "list":
print(json.dumps(ci.list_pending(), indent=2, default=str))
elif args.cmd == "history":
print(json.dumps(ci.history(limit=args.limit, kind=args.kind), indent=2, default=str))
elif args.cmd == "show":
d = ci.inspect_pending(args.pending_id)
print(json.dumps(d, indent=2, default=str) if d else "not found")
elif args.cmd == "confirm":
print(json.dumps(ci.confirm_edit(args.pending_id), indent=2, default=str))
elif args.cmd == "reject":
print(json.dumps(ci.reject_edit(args.pending_id, reason=args.reason), indent=2, default=str))
elif args.cmd == "propose":
old_text = Path(args.old).read_text(encoding="utf-8")
new_text = Path(args.new).read_text(encoding="utf-8")
result = ci.propose_edit(
target_path=args.target,
old_text=old_text,
new_text=new_text,
rationale=args.rationale,
)
print(json.dumps(result, indent=2, default=str))
|