Spaces:
Sleeping
Sleeping
File size: 35,113 Bytes
a9df8b1 | 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 | from __future__ import annotations
import argparse
import hashlib
import json
import re
import sys
import time
import urllib.error
import urllib.request
from dataclasses import dataclass
from datetime import datetime
from pathlib import Path
from typing import Any
REPO_ROOT = Path(__file__).resolve().parents[1]
DEFAULT_RUN_ROOT = REPO_ROOT / "artifacts" / "input-skill-eval-runs"
ALLOWED_INPUT_TYPES = {"trajectory", "document", "api_doc", "script", "past_skills"}
@dataclass(frozen=True)
class FixtureCase:
source_id: str
input_type: str
domain: str
source_url: str
paper_or_project: str
expected_skill_shape: str
license_note: str
local_path: Path
content: str
content_sha256: str
manifest: dict[str, Any]
truncated: bool = False
def main(argv: list[str] | None = None) -> int:
parser = argparse.ArgumentParser(
description=(
"Run reusable SkillOS five-input evaluation from a manifest. "
"Raw fixture files are read-only; all outputs go under an isolated run directory."
)
)
parser.add_argument("--manifest", required=True, help="JSON manifest containing a fixtures array.")
parser.add_argument("--fixture-root", default="", help="Root for relative fixture content_file paths.")
parser.add_argument("--api-base", default="http://127.0.0.1:8001/api/v1")
parser.add_argument("--run-root", default=str(DEFAULT_RUN_ROOT))
parser.add_argument("--run-id", default=datetime.now().strftime("%Y%m%d-%H%M%S"))
parser.add_argument("--limit-per-type", type=int, default=0)
parser.add_argument("--max-chars", type=int, default=14000)
parser.add_argument("--request-timeout", type=int, default=180)
parser.add_argument("--create-candidates", action="store_true")
parser.add_argument("--max-candidates-per-fixture", type=int, default=1)
parser.add_argument("--snapshot", action="store_true", help="Create a version snapshot for created candidates.")
parser.add_argument("--fail-under", type=float, default=0.0)
args = parser.parse_args(argv)
manifest_path = Path(args.manifest).resolve()
fixture_root = Path(args.fixture_root).resolve() if args.fixture_root else manifest_path.parent
run_dir = Path(args.run_root).resolve() / f"input-skill-eval-{args.run_id}"
raw_dir = run_dir / "raw"
raw_dir.mkdir(parents=True, exist_ok=True)
cases = load_manifest(manifest_path, fixture_root=fixture_root, max_chars=args.max_chars)
cases = limit_cases(cases, args.limit_per_type)
client = ApiClient(args.api_base.rstrip("/"), raw_dir)
summary: dict[str, Any] = {
"started_at": datetime.now().isoformat(timespec="seconds"),
"api_base": client.base_url,
"manifest": str(manifest_path),
"fixture_root": str(fixture_root),
"run_dir": str(run_dir),
"run_id": args.run_id,
"isolation": {
"raw_fixtures_read_only": True,
"created_candidates": bool(args.create_candidates),
"outputs_are_run_scoped": True,
},
"options": {
"limit_per_type": args.limit_per_type,
"max_chars": args.max_chars,
"create_candidates": args.create_candidates,
"max_candidates_per_fixture": args.max_candidates_per_fixture,
"snapshot": args.snapshot,
},
"connectivity": {},
"fixture_count": len(cases),
"records": [],
"scores": {},
"notes": [],
}
summary["connectivity"]["skills_before"] = client.get("/skills?limit=1000", label="skills_before_eval")
before_count = count_skills(summary["connectivity"]["skills_before"])
for case in cases:
record = run_fixture_case(
case,
client,
run_id=args.run_id,
request_timeout=args.request_timeout,
create_candidates=args.create_candidates,
max_candidates_per_fixture=args.max_candidates_per_fixture,
snapshot=args.snapshot,
)
record["score"] = score_fixture_record(record)
summary["records"].append(record)
summary["connectivity"]["skills_after"] = client.get("/skills?limit=1000", label="skills_after_eval")
after_count = count_skills(summary["connectivity"]["skills_after"])
summary["skill_counts"] = {
"before": before_count,
"after": after_count,
"delta": after_count - before_count,
}
summary["scores"] = summarize_scores(summary["records"])
summary["finished_at"] = datetime.now().isoformat(timespec="seconds")
write_json(run_dir / "summary.json", summary)
(run_dir / "REPORT.md").write_text(render_report(summary), encoding="utf-8")
print(json.dumps({
"run_dir": str(run_dir),
"fixture_count": len(cases),
"overall": summary["scores"]["overall"],
"by_input_type": summary["scores"]["by_input_type"],
"skill_delta": summary["skill_counts"]["delta"],
}, indent=2, ensure_ascii=False))
if args.fail_under and summary["scores"]["overall"] < args.fail_under:
return 1
return 0
def load_manifest(manifest_path: Path, *, fixture_root: Path, max_chars: int = 0) -> list[FixtureCase]:
data = json.loads(manifest_path.read_text(encoding="utf-8"))
entries = data.get("fixtures") if isinstance(data, dict) else data
if not isinstance(entries, list):
raise ValueError("Manifest must be a JSON array or an object with a fixtures array.")
cases: list[FixtureCase] = []
for idx, entry in enumerate(entries):
if not isinstance(entry, dict):
raise ValueError(f"Fixture entry {idx} must be an object.")
input_type = str(entry.get("input_type") or entry.get("source_type") or "").strip()
if input_type not in ALLOWED_INPUT_TYPES:
raise ValueError(f"Fixture {idx} has unsupported input_type {input_type!r}.")
source_id = str(entry.get("source_id") or f"{input_type}-{idx + 1}").strip()
path_text = str(entry.get("content_file") or entry.get("local_path") or entry.get("path") or "").strip()
if not path_text:
raise ValueError(f"Fixture {source_id} must provide content_file/local_path/path.")
local_path = Path(path_text)
if not local_path.is_absolute():
local_path = fixture_root / local_path
content, truncated = read_fixture_text(local_path, max_chars=max_chars)
sha = hashlib.sha256(content.encode("utf-8")).hexdigest()
cases.append(FixtureCase(
source_id=source_id,
input_type=input_type,
domain=str(entry.get("domain") or "other"),
source_url=str(entry.get("source_url") or ""),
paper_or_project=str(entry.get("paper_or_project") or ""),
expected_skill_shape=str(entry.get("expected_skill_shape") or ""),
license_note=str(entry.get("license_note") or ""),
local_path=local_path,
content=content,
content_sha256=sha,
manifest=dict(entry),
truncated=truncated,
))
return cases
def limit_cases(cases: list[FixtureCase], limit_per_type: int) -> list[FixtureCase]:
if limit_per_type <= 0:
return cases
counts: dict[str, int] = {}
selected: list[FixtureCase] = []
for case in cases:
count = counts.get(case.input_type, 0)
if count >= limit_per_type:
continue
selected.append(case)
counts[case.input_type] = count + 1
return selected
def read_fixture_text(path: Path, *, max_chars: int) -> tuple[str, bool]:
content = path.read_text(encoding="utf-8", errors="replace")
if max_chars <= 0 or len(content) <= max_chars:
return content, False
marker = (
"\n\n[Fixture truncated for repeatable local evaluation. "
f"Original file: {path.name}; retained_chars={max_chars}.]\n"
)
keep = max(1000, max_chars - len(marker))
return content[:keep] + marker, True
def run_fixture_case(
case: FixtureCase,
client: "ApiClient",
*,
run_id: str,
request_timeout: int,
create_candidates: bool,
max_candidates_per_fixture: int,
snapshot: bool,
) -> dict[str, Any]:
label_base = safe_filename(f"{case.input_type}_{case.source_id}")
record: dict[str, Any] = {
"fixture": fixture_case_summary(case),
"parse": {},
"audit": {"status": "not_run"},
"create": {"status": "not_run"},
"graph": {"status": "not_run"},
"version": {"status": "not_run"},
"harness": {"status": "not_run"},
"skillsbench": {
"status": "not_run",
"note": "Reserved for SkillsBench/BenchFlow adapter. This runner keeps the field explicit instead of faking benchmark execution.",
},
}
parse_result = client.post(
"/ingest/parse",
{
"source_type": case.input_type,
"content": case.content,
"metadata": {
"eval_run_id": run_id,
"source_id": case.source_id,
"input_type": case.input_type,
"domain": case.domain,
"source_url": case.source_url,
"paper_or_project": case.paper_or_project,
"expected_skill_shape": case.expected_skill_shape,
"license_note": case.license_note,
"content_sha256": case.content_sha256,
"origin_file": str(case.local_path),
"ephemeral_preview_only": not create_candidates,
"max_candidates": max(1, max_candidates_per_fixture),
},
},
label=f"parse_{label_base}",
timeout_s=request_timeout,
)
record["parse"] = analyze_parse(case, parse_result)
units = parse_result.get("units") if isinstance(parse_result, dict) else []
if not create_candidates or not units:
return record
created: list[dict[str, Any]] = []
audits: list[dict[str, Any]] = []
graph_checks: list[dict[str, Any]] = []
version_checks: list[dict[str, Any]] = []
for idx, unit in enumerate(units[:max(1, max_candidates_per_fixture)]):
payload = candidate_review_payload(unit, case, run_id=run_id)
audit_result = client.post(
"/ingest/audit-candidate",
payload,
label=f"audit_{label_base}_{idx + 1}",
timeout_s=request_timeout,
)
audits.append(analyze_audit(payload, audit_result))
create_result = client.post(
"/ingest/create-candidate",
payload,
label=f"create_{label_base}_{idx + 1}",
timeout_s=request_timeout,
)
created_info = analyze_create(create_result)
created.append(created_info)
skill_id = created_info.get("created_skill_id")
if not skill_id:
continue
graph_checks.append(check_graph_for_skill(client, str(skill_id), label=f"graph_{label_base}_{idx + 1}"))
version_checks.append(check_version_for_skill(
client,
str(skill_id),
label=f"version_{label_base}_{idx + 1}",
snapshot=snapshot,
))
record["audit"] = summarize_audits(audits)
record["create"] = summarize_creates(created)
record["graph"] = summarize_graph_checks(graph_checks)
record["version"] = summarize_version_checks(version_checks)
return record
def analyze_parse(case: FixtureCase, result: dict[str, Any]) -> dict[str, Any]:
units = result.get("units") or []
unit_infos = [analyze_unit(unit) for unit in units if isinstance(unit, dict)]
return {
"http_status": result.get("_http_status"),
"success": bool(result.get("success")),
"unit_count": int(result.get("unit_count") or len(unit_infos)),
"error_count": len(result.get("errors") or []),
"errors": result.get("errors") or [],
"elapsed_ms": result.get("_elapsed_ms"),
"schema_completeness": average([unit["schema_completeness"] for unit in unit_infos]),
"ctx2skill_evidence_completeness": average([unit["ctx2skill_evidence_completeness"] for unit in unit_infos]),
"layer_correctness": expected_layer_score(case, unit_infos),
"graph_relation_preview_count": sum(unit["graph_relation_preview_count"] for unit in unit_infos),
"units": unit_infos,
}
def analyze_unit(unit: dict[str, Any]) -> dict[str, Any]:
meta = unit.get("metadata") or {}
interface = meta.get("candidate_interface") if isinstance(meta.get("candidate_interface"), dict) else {}
evidence = meta.get("ctx2skill_evidence") if isinstance(meta.get("ctx2skill_evidence"), dict) else {}
relations = meta.get("candidate_relations") if isinstance(meta.get("candidate_relations"), dict) else {}
graph_preview = meta.get("graph_relation_preview") or []
input_schema = interface.get("input_schema") if isinstance(interface.get("input_schema"), dict) else {}
output_schema = interface.get("output_schema") if isinstance(interface.get("output_schema"), dict) else {}
schema_bits = [
bool(input_schema),
bool(output_schema),
bool(interface.get("preconditions")),
bool(interface.get("postconditions")),
]
evidence_bits = [
bool(evidence),
bool(evidence.get("challenges")),
bool(evidence.get("judge_results")),
bool(evidence.get("selected_candidate") or evidence.get("selected_reason")),
]
return {
"unit_id": unit.get("unit_id"),
"proposed_skill_name": unit.get("proposed_skill_name"),
"proposed_type": unit.get("proposed_type"),
"confidence": unit.get("confidence"),
"schema_completeness": round(sum(1 for bit in schema_bits if bit) / len(schema_bits), 2),
"ctx2skill_evidence_completeness": round(sum(1 for bit in evidence_bits if bit) / len(evidence_bits), 2),
"challenge_count": len(evidence.get("challenges") or []),
"judge_result_count": len(evidence.get("judge_results") or []),
"candidate_score_count": len(evidence.get("candidate_scores") or []),
"layering_reason_present": bool(meta.get("layering_reason")),
"graph_relation_preview_count": len(graph_preview) if isinstance(graph_preview, list) else 0,
"dependency_count": len(relations.get("dependency_ids") or []),
"component_count": len(relations.get("component_ids") or []),
"parent_count": len(relations.get("parent_skill_ids") or []),
}
def expected_layer_score(case: FixtureCase, units: list[dict[str, Any]]) -> float:
if not units:
return 0.0
expected = case.expected_skill_shape.strip().lower()
if expected not in {"atomic", "functional", "strategic"}:
return 0.5
matches = sum(1 for unit in units if unit.get("proposed_type") == expected)
return round(matches / len(units), 2)
def candidate_review_payload(unit: dict[str, Any], case: FixtureCase, *, run_id: str) -> dict[str, Any]:
meta = unit.get("metadata") or {}
interface = meta.get("candidate_interface") if isinstance(meta.get("candidate_interface"), dict) else {}
implementation = meta.get("candidate_implementation") if isinstance(meta.get("candidate_implementation"), dict) else {}
relations = meta.get("candidate_relations") if isinstance(meta.get("candidate_relations"), dict) else {}
evaluation = meta.get("candidate_evaluation") if isinstance(meta.get("candidate_evaluation"), dict) else None
source_type = unit.get("source_type") or case.input_type
base_name = normalize_skill_name(unit.get("proposed_skill_name") or f"{case.source_id}_{source_type}_candidate")
source_suffix = normalize_skill_name(case.source_id)
if source_suffix and source_suffix not in base_name:
name = normalize_skill_name(f"{base_name}_{source_suffix}")[:120]
else:
name = base_name
description = unit.get("proposed_description") or unit.get("summary") or name
if evaluation:
evaluation = dict(evaluation)
evaluation.setdefault("test_case_refs", [f"{case.source_id}:{unit.get('unit_id', 'unit')}"])
evaluation.setdefault("benchmark_task_ids", list(case.manifest.get("target_benchmark_tasks") or []))
evaluation.setdefault("validation_summary", "Created by isolated input-skill evaluation runner.")
else:
evaluation = {
"verifier_specs": [{"type": "json_exists", "path": "output.result"}],
"test_case_refs": [f"{case.source_id}:{unit.get('unit_id', 'unit')}"],
"benchmark_task_ids": list(case.manifest.get("target_benchmark_tasks") or []),
"validation_summary": "Created by isolated input-skill evaluation runner.",
}
tags = meta.get("candidate_tags") if isinstance(meta.get("candidate_tags"), list) else []
return {
"source_type": source_type,
"unit_id": unit.get("unit_id") or f"{case.source_id}:unit",
"raw_content": unit.get("raw_content") or "",
"name": name,
"description": description,
"skill_type": unit.get("proposed_type") if unit.get("proposed_type") in {"atomic", "functional", "strategic"} else "atomic",
"tags": unique_strings([
*tags,
*(unit.get("index_keywords") or [])[:4],
"input-skill-eval",
f"eval-run:{run_id}",
f"source-id:{case.source_id}",
f"domain:{case.domain}",
])[:12],
"input_schema": interface.get("input_schema") or {"type": "object", "properties": {}},
"output_schema": interface.get("output_schema") or {"type": "object", "properties": {"result": {"type": "object"}}},
"preconditions": interface.get("preconditions") or [],
"postconditions": interface.get("postconditions") or ["Candidate returns a structured result."],
"prompt_template": implementation.get("prompt_template") or unit.get("summary") or description,
"provenance": {
"source_type": source_type,
"source_ids": unique_strings([case.source_id, unit.get("unit_id") or "unit"]),
"parent_skill_ids": relations.get("parent_skill_ids") or [],
"created_by_agent": "input_skill_eval_runner",
"creation_context": {
"eval_run_id": run_id,
"fixture_import_mode": "isolated_eval_candidate",
"source_id": case.source_id,
"input_type": case.input_type,
"domain": case.domain,
"source_url": case.source_url,
"paper_or_project": case.paper_or_project,
"expected_skill_shape": case.expected_skill_shape,
"license_note": case.license_note,
"local_path": str(case.local_path),
"content_sha256": case.content_sha256,
"ctx2skill_evidence": meta.get("ctx2skill_evidence"),
},
},
"evaluation": evaluation,
"dependency_ids": relations.get("dependency_ids") or [],
"component_ids": relations.get("component_ids") or [],
"sub_skill_ids": relations.get("sub_skill_ids") or relations.get("component_ids") or [],
"parent_skill_ids": relations.get("parent_skill_ids") or [],
"tool_calls": implementation.get("tool_calls") or [],
"author": "input_skill_eval_runner",
}
def analyze_audit(payload: dict[str, Any], result: dict[str, Any]) -> dict[str, Any]:
return {
"candidate_name": payload.get("name"),
"http_status": result.get("_http_status"),
"passed": bool(result.get("passed")),
"audit_score": result.get("audit_score"),
"issues": result.get("issues") or [],
"warnings": result.get("warnings") or [],
}
def analyze_create(result: dict[str, Any]) -> dict[str, Any]:
created = result.get("created_skill") if isinstance(result.get("created_skill"), dict) else {}
return {
"http_status": result.get("_http_status"),
"success": bool(result.get("success")),
"created_skill_id": created.get("skill_id"),
"created_skill_name": created.get("name"),
"state": created.get("state"),
"audit_passed": (result.get("audit") or {}).get("passed") if isinstance(result.get("audit"), dict) else None,
}
def check_graph_for_skill(client: "ApiClient", skill_id: str, *, label: str) -> dict[str, Any]:
result = client.get(f"/graph/view?view=skill_only&limit=500", label=label)
nodes = result.get("nodes") or []
edges = result.get("edges") or []
skill_present = any(str(node.get("skill_id") or node.get("id") or "").endswith(skill_id) for node in nodes)
relation_count = sum(
1 for edge in edges
if skill_id in str(edge.get("source") or edge.get("source_id") or "")
or skill_id in str(edge.get("target") or edge.get("target_id") or "")
)
return {
"skill_id": skill_id,
"http_status": result.get("_http_status"),
"skill_present": skill_present,
"relation_count": relation_count,
"node_count": len(nodes),
"edge_count": len(edges),
}
def check_version_for_skill(client: "ApiClient", skill_id: str, *, label: str, snapshot: bool) -> dict[str, Any]:
diff = client.get(f"/lifecycle/{skill_id}/diff", label=f"{label}_diff")
result = {
"skill_id": skill_id,
"http_status": diff.get("_http_status"),
"business_diff_available": "business_diff" in diff or "business_summary" in diff or "history" in diff,
"raw_diff_available": "raw_diff" in diff or "changes" in diff or "snapshot_diff" in diff or "history" in diff,
"snapshot_created": False,
}
if snapshot:
snap = client.post(
f"/lifecycle/{skill_id}/snapshot",
{
"description": "Input-skill eval snapshot.",
"author": "input_skill_eval_runner",
},
label=f"{label}_snapshot",
timeout_s=120,
)
result["snapshot_http_status"] = snap.get("_http_status")
result["snapshot_created"] = snap.get("_http_status") in {200, 201} and not snap.get("detail")
return result
def summarize_audits(items: list[dict[str, Any]]) -> dict[str, Any]:
if not items:
return {"status": "not_run"}
return {
"status": "completed",
"count": len(items),
"passed_count": sum(1 for item in items if item.get("passed")),
"average_audit_score": average([
float(item.get("audit_score") or 0.0)
for item in items
if isinstance(item.get("audit_score"), (int, float))
]),
"items": items,
}
def summarize_creates(items: list[dict[str, Any]]) -> dict[str, Any]:
if not items:
return {"status": "not_run"}
created_ids = [item.get("created_skill_id") for item in items if item.get("created_skill_id")]
return {
"status": "completed",
"count": len(items),
"success_count": sum(1 for item in items if item.get("success")),
"created_skill_ids": created_ids,
"http_statuses": [item.get("http_status") for item in items],
"items": items,
}
def summarize_graph_checks(items: list[dict[str, Any]]) -> dict[str, Any]:
if not items:
return {"status": "not_run"}
return {
"status": "completed",
"count": len(items),
"present_count": sum(1 for item in items if item.get("skill_present")),
"relation_count": sum(int(item.get("relation_count") or 0) for item in items),
"items": items,
}
def summarize_version_checks(items: list[dict[str, Any]]) -> dict[str, Any]:
if not items:
return {"status": "not_run"}
return {
"status": "completed",
"count": len(items),
"business_diff_available_count": sum(1 for item in items if item.get("business_diff_available")),
"snapshot_created_count": sum(1 for item in items if item.get("snapshot_created")),
"items": items,
}
def score_fixture_record(record: dict[str, Any]) -> float:
parse = record.get("parse") or {}
score = 0.0
if parse.get("http_status") == 200 and parse.get("success"):
score += 0.18
if parse.get("unit_count", 0) > 0:
score += 0.10
score += 0.08 * clamp01(float(parse.get("schema_completeness") or 0.0))
score += 0.08 * clamp01(float(parse.get("ctx2skill_evidence_completeness") or 0.0))
score += 0.06 * clamp01(float(parse.get("layer_correctness") or 0.0))
audit = record.get("audit") or {}
if audit.get("status") == "completed":
score += 0.10 * ratio(audit.get("passed_count"), audit.get("count"))
elif audit.get("http_status") == 200 and audit.get("passed"):
score += 0.10
create = record.get("create") or {}
if create.get("status") == "completed":
score += 0.12 * ratio(create.get("success_count"), create.get("count"))
elif create.get("http_status") in {200, 201} and create.get("success"):
score += 0.12
graph = record.get("graph") or {}
if graph.get("status") == "completed":
score += 0.10 * ratio(graph.get("present_count"), graph.get("count"))
if int(graph.get("relation_count") or 0) > 0:
score += 0.05
elif graph.get("http_status") == 200 and graph.get("skill_present"):
score += 0.10
if int(graph.get("relation_count") or 0) > 0:
score += 0.05
version = record.get("version") or {}
if version.get("status") == "completed":
score += 0.06 * ratio(version.get("business_diff_available_count"), version.get("count"))
if int(version.get("snapshot_created_count") or 0) > 0:
score += 0.03
elif version.get("http_status") == 200:
if version.get("business_diff_available"):
score += 0.06
if version.get("snapshot_created"):
score += 0.03
harness = record.get("harness") or {}
if harness.get("status") == "verified":
score += 0.12
elif harness.get("positive_pass") or harness.get("negative_rejected"):
score += 0.06
skillsbench = record.get("skillsbench") or {}
if skillsbench.get("status") in {"mapped", "completed"} and skillsbench.get("verifier_ran"):
score += 0.10
if skillsbench.get("generated_pass"):
score += 0.05
return round(min(score, 1.0), 3)
def summarize_scores(records: list[dict[str, Any]]) -> dict[str, Any]:
by_type_values: dict[str, list[float]] = {}
by_domain_values: dict[str, list[float]] = {}
for record in records:
score = float(record.get("score") or 0.0)
fixture = record.get("fixture") or {}
by_type_values.setdefault(str(fixture.get("input_type") or "unknown"), []).append(score)
by_domain_values.setdefault(str(fixture.get("domain") or "unknown"), []).append(score)
return {
"overall": average([float(record.get("score") or 0.0) for record in records]),
"by_input_type": {key: average(values) for key, values in by_type_values.items()},
"by_domain": {key: average(values) for key, values in by_domain_values.items()},
"sample_count": len(records),
}
def render_report(summary: dict[str, Any]) -> str:
lines = [
"# SkillOS Input-to-Skill Evaluation Report",
"",
f"- Started: `{summary['started_at']}`",
f"- Finished: `{summary.get('finished_at', '')}`",
f"- API base: `{summary['api_base']}`",
f"- Manifest: `{summary['manifest']}`",
f"- Fixture root: `{summary['fixture_root']}`",
f"- Run directory: `{summary['run_dir']}`",
f"- Created candidates: `{summary['options']['create_candidates']}`",
"",
"## Isolation",
"",
"- Raw fixture files are read only.",
"- API responses and reports are written only under this run directory.",
"- Candidate creation is disabled unless `--create-candidates` is passed.",
"- SkillsBench results are explicit fields and remain `not_run` until the benchmark adapter executes.",
"",
"## Scores",
"",
f"- Overall: `{summary['scores']['overall']}`",
f"- By input type: `{json.dumps(summary['scores']['by_input_type'], ensure_ascii=False)}`",
f"- By domain: `{json.dumps(summary['scores']['by_domain'], ensure_ascii=False)}`",
f"- Skill delta: `{summary['skill_counts']['delta']}`",
"",
"## Fixtures",
"",
]
for record in summary["records"]:
fixture = record["fixture"]
parse = record["parse"]
lines.extend([
f"### {fixture['source_id']} - {record['score']}",
f"- Input/domain: `{fixture['input_type']}` / `{fixture['domain']}`",
f"- Source: `{fixture.get('source_url', '')}`",
f"- Paper/project: `{fixture.get('paper_or_project', '')}`",
f"- Local file: `{fixture['local_path']}`",
f"- Parse: http=`{parse.get('http_status')}`, success=`{parse.get('success')}`, "
f"units=`{parse.get('unit_count')}`, schema=`{parse.get('schema_completeness')}`, "
f"ctx2skill=`{parse.get('ctx2skill_evidence_completeness')}`, layer=`{parse.get('layer_correctness')}`",
f"- Audit: `{record.get('audit', {}).get('status')}`",
f"- Create: `{record.get('create', {}).get('status')}`",
f"- Graph: `{record.get('graph', {}).get('status')}`",
f"- Version: `{record.get('version', {}).get('status')}`",
f"- Harness: `{record.get('harness', {}).get('status')}`",
f"- SkillsBench: `{record.get('skillsbench', {}).get('status')}`",
"",
])
lines.append("Raw API envelopes are saved under `raw/` next to this report.")
lines.append("")
return "\n".join(lines)
class ApiClient:
def __init__(self, base_url: str, raw_dir: Path) -> None:
self.base_url = base_url
self.raw_dir = raw_dir
def get(self, path: str, *, label: str, timeout_s: int = 30) -> dict[str, Any]:
return self._request("GET", self.base_url + path, None, label, timeout_s)
def post(self, path: str, payload: dict[str, Any], *, label: str, timeout_s: int = 120) -> dict[str, Any]:
return self._request("POST", self.base_url + path, payload, label, timeout_s)
def _request(
self,
method: str,
url: str,
payload: dict[str, Any] | None,
label: str,
timeout_s: int,
) -> dict[str, Any]:
data = None if payload is None else json.dumps(payload, ensure_ascii=False).encode("utf-8")
req = urllib.request.Request(url, data=data, method=method)
req.add_header("Accept", "application/json")
if payload is not None:
req.add_header("Content-Type", "application/json")
start = time.perf_counter()
try:
with urllib.request.urlopen(req, timeout=timeout_s) as resp:
body = resp.read().decode("utf-8", errors="replace")
status = resp.status
except urllib.error.HTTPError as exc:
body = exc.read().decode("utf-8", errors="replace")
status = exc.code
except Exception as exc:
body = json.dumps({"error": str(exc)}, ensure_ascii=False)
status = 0
elapsed_ms = round((time.perf_counter() - start) * 1000, 2)
parsed = parse_body(body)
envelope = {
"label": label,
"method": method,
"url": url,
"status_code": status,
"elapsed_ms": elapsed_ms,
"request_preview": preview_payload(payload),
"body": parsed,
}
write_json(self.raw_dir / f"{safe_filename(label)}.json", envelope)
if isinstance(parsed, dict):
result = dict(parsed)
result["_http_status"] = status
result["_elapsed_ms"] = elapsed_ms
return result
return {"_http_status": status, "_elapsed_ms": elapsed_ms, "body": parsed}
def fixture_case_summary(case: FixtureCase) -> dict[str, Any]:
return {
"source_id": case.source_id,
"input_type": case.input_type,
"domain": case.domain,
"source_url": case.source_url,
"paper_or_project": case.paper_or_project,
"expected_skill_shape": case.expected_skill_shape,
"license_note": case.license_note,
"local_path": str(case.local_path),
"content_sha256": case.content_sha256,
"truncated": case.truncated,
}
def count_skills(payload: Any) -> int:
if isinstance(payload, list):
return len(payload)
if isinstance(payload, dict):
body = payload.get("body")
if isinstance(body, list):
return len(body)
for key in ("value", "skills", "items"):
value = payload.get(key)
if isinstance(value, list):
return len(value)
if isinstance(payload.get("total"), int):
return int(payload["total"])
return 0
def write_json(path: Path, data: Any) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
path.write_text(json.dumps(data, indent=2, ensure_ascii=False), encoding="utf-8")
def parse_body(body: str) -> Any:
try:
return json.loads(body)
except json.JSONDecodeError:
return body
def preview_payload(payload: dict[str, Any] | None) -> Any:
if payload is None:
return None
preview = dict(payload)
content = preview.get("content")
if isinstance(content, str):
preview["content"] = f"{content[:500]}... [chars={len(content)}]"
return preview
def normalize_skill_name(value: str) -> str:
text = re.sub(r"[^a-zA-Z0-9_]+", "_", str(value or "").strip().lower())
text = re.sub(r"_+", "_", text).strip("_")
return text[:120] or "input_skill_eval_candidate"
def safe_filename(value: str) -> str:
normalized = normalize_skill_name(value).replace("/", "_")
if len(normalized) <= 96:
return normalized
digest = hashlib.sha1(str(value).encode("utf-8")).hexdigest()[:10]
return f"{normalized[:84].rstrip('_')}_{digest}"
def unique_strings(values: list[Any]) -> list[str]:
seen: set[str] = set()
result: list[str] = []
for value in values:
text = str(value or "").strip()
if not text or text in seen:
continue
seen.add(text)
result.append(text)
return result
def average(values: list[float]) -> float:
return round(sum(values) / len(values), 3) if values else 0.0
def ratio(numerator: Any, denominator: Any) -> float:
try:
denom = float(denominator or 0)
if denom <= 0:
return 0.0
return clamp01(float(numerator or 0) / denom)
except (TypeError, ValueError):
return 0.0
def clamp01(value: float) -> float:
return min(max(value, 0.0), 1.0)
if __name__ == "__main__":
sys.exit(main())
|