Files
Jyotisha/scripts/run_quality_gate.py
Jesse_Chen b5bcbaed73
Independent Staging Quality Gate / validate (push) Has been cancelled
Independent Staging Quality Gate / publish (push) Has been cancelled
fix(consult): record a tool call the tool never received
staging 手测 run 951a841e 第一次工具调用发出 tool.failed 后重试成功,但回执里
steps 只有 skill 与那次成功的 tool,stepBudget.used 为 2——失败的那次完全不存在。
客户端看见失败过一次,回执说没有,两边都查不到为什么。

工具的 inputSchema 是 strict 的,模型参数不合法时 Mastra 在调用 execute 之前就拒了,
于是工具体内一切都没跑:调用不计数、失败步不记录、连 chart-calculation 活动事件都没
发出(这也是本次定位的证据——失败那次没有任何 activity,重试那次有)。工具无法记录
一次它从未收到的调用。叠加两处:safeToolError 把非超时非取消的错误全塌成
calculation_failed,而即使失败落进工具体的 catch,consultationWorkflowFailureCode
对非 ConsultationWorkflowError 返回 undefined、append 处又写成可选省略,于是最需要
解释的那条记录恰好是唯一没有原因的记录。

改为在流层补记:流是唯一能观测到全部工具失败的位置,无论失败在 schema 这侧还是
execute 那侧,且它持有 startedAt 因而能给出时长。tool-error 分支比对「流已见的错误数」
与「state 里已有的失败 tool 步数」,只在前者更多时补一条,工具仍记录它能看见的失败,
两者不重复计。另新增 consultationToolFailureCode,令每个错误都解析出一个码。

failureCode 刻意仍不进公开回执:白名单与「the public receipt never carries the
internal failure classification」是刻意约束,workflow_rate_limited 这类后端内情不该
上线到客户端。原因走可观测日志的 toolCalls[].failureCode。客户端能看到「有一步失败」,
运维能在日志里看到为什么。

另记入 BUG-268 的线上实测值:单领域 referenceReads 两次均为 2,多领域为 0——不是
从不读方法,而是最需要方法的多领域路径一份都没打开。

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-18 15:30:51 +08:00

537 lines
20 KiB
Python

#!/usr/bin/env python3
"""Run the Jyotish skill quality gate used by local development and CI."""
from __future__ import annotations
import argparse
import contextlib
import json
import os
import py_compile
import subprocess
import sys
import tempfile
from pathlib import Path
ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(ROOT / "scripts"))
from local_env import load_local_env # noqa: E402
load_local_env(ROOT)
APP = ROOT / "frontend"
PYTHON = sys.executable
os.environ.setdefault("PYTHON", PYTHON)
COMPILE_DIRS = [
ROOT / "scripts",
ROOT / "jyotish_vedic",
]
EXTRA_COMPILE_TARGETS = [
ROOT / "mcp_server.py",
ROOT / "scripts" / "audit_fragments.py",
ROOT / "scripts" / "character_level_inventory_manifest.py",
ROOT / "scripts" / "dasha_reference_audit.py",
ROOT / "scripts" / "external_oracle_sanity_closure.py",
ROOT / "scripts" / "interpretation_source_inventory_gate.py",
ROOT / "scripts" / "oracle_boundary_audit.py",
ROOT / "scripts" / "oracle_collection_queue.py",
ROOT / "scripts" / "oracle_evidence_validator.py",
ROOT / "scripts" / "sync_final_evidence_packet_status.py",
ROOT / "tests" / "run_golden_cases.py",
ROOT / "tests" / "run_real_case_revalidation.py",
]
CORE_PYTEST_TARGETS = [
"tests/test_cli_smoke.py",
"tests/test_api_server_security.py",
"tests/test_jaimini.py",
"tests/test_shadbala_complete.py",
"tests/test_transit_trigger.py",
"tests/test_oracle_collection_queue.py",
"tests/test_oracle_evidence_validator.py",
"tests/test_external_oracle_sanity_closure.py",
# The staging gate runs the quick profile, so a guard absent from this list never runs in CI.
# These pin the answer-truth contract every product consultation is built on, and their failure
# mode is silent widening — nothing errors when they regress (BUG-267, BUG-270).
"tests/test_consultation_consumer_context.py",
]
RUNTIME_TRUTH_PYTEST_TARGETS = [
"tests/test_api_server_security.py::test_high_rigor_vedastro_official_summary_passes_through_contract_fields",
"tests/test_api_server_security.py::test_high_rigor_vedastro_official_summary_exposes_top_reader_contract_from_full_snapshot",
"tests/test_vedastro_external_technique_evidence.py::test_strict_workflow_uses_shared_consultation_executor",
"tests/test_vedastro_runtime_mode_diagnostics.py",
"tests/test_interpretation_source_inventory_gate.py::test_quality_gate_runs_interpretation_source_inventory_gate",
"tests/test_interpretation_source_runtime_coverage.py",
"tests/test_final_jhora_evidence_packet_acceptance.py",
]
RELEASE_CRITICAL_UNTRACKED_PATHS = [
"docs/research/desktop_packaging_spike_2026_06_23.md",
"docs/research/ephemeris_abstraction_feasibility_2026_06_23.md",
"docs/research/ephemeris_adapter_contract_2026_06_23.md",
"docs/research/ephemeris_candidate_adapter_spike_2026_06_23.md",
"docs/research/open_source_scan_2026_06_22.md",
"docs/research/product_gap_matrix_2026_06_22.md",
"docs/research/whole_machine_git_audit_2026_06_23.md",
"findings.md",
"progress.md",
"references/oracle/dasha_shadbala_oracle_cases.json",
"scripts/audit_fragments.py",
"scripts/deep_varga_avastha.py",
"scripts/dasha_reference_audit.py",
"scripts/ephemeris_adapter_contract.py",
"scripts/ephemeris_backend_probe.py",
"scripts/ephemeris_candidate_adapter_spike.py",
"scripts/oracle_boundary_audit.py",
"scripts/oracle_collection_queue.py",
"scripts/oracle_evidence_validator.py",
"task_plan.md",
"tests/test_api_server_security.py",
"tests/test_dasha_reference_audit.py",
"tests/test_deep_varga_avastha.py",
"tests/test_oracle_boundary_audit.py",
"tests/test_oracle_collection_queue.py",
"tests/test_oracle_evidence_validator.py",
]
QUALITY_GATE_PROFILES = {
"quick": {
"skip_slow": True,
"skip_yoga_logic": True,
"skip_frontend_runtime": False,
"check_release_hygiene": False,
"skip_real_cases": True,
"skip_dasha_audit": True,
"skip_oracle_audit": True,
"skip_local_accuracy_report": True,
"skip_vedastro_live": True,
},
"browser": {
"skip_slow": True,
"skip_yoga_logic": True,
"skip_frontend_runtime": False,
"check_release_hygiene": False,
"skip_real_cases": True,
"skip_dasha_audit": True,
"skip_oracle_audit": True,
"skip_local_accuracy_report": True,
"skip_vedastro_live": True,
},
"release": {
"skip_slow": False,
"skip_yoga_logic": False,
"skip_frontend_runtime": False,
"check_release_hygiene": True,
"skip_real_cases": False,
"skip_dasha_audit": False,
"skip_oracle_audit": False,
"skip_local_accuracy_report": False,
"skip_vedastro_live": True,
},
"accuracy": {
"skip_slow": True,
"skip_yoga_logic": False,
"skip_frontend_runtime": True,
"check_release_hygiene": False,
"skip_real_cases": False,
"skip_dasha_audit": False,
"skip_oracle_audit": False,
"skip_local_accuracy_report": False,
"skip_vedastro_live": True,
},
"vedastro-live": {
"skip_slow": True,
"skip_yoga_logic": True,
"skip_frontend_runtime": True,
"check_release_hygiene": False,
"skip_real_cases": True,
"skip_dasha_audit": True,
"skip_oracle_audit": True,
"skip_local_accuracy_report": True,
"skip_vedastro_live": False,
},
"runtime-truth": {
"skip_slow": True,
"skip_yoga_logic": True,
"skip_frontend_runtime": True,
"check_release_hygiene": False,
"skip_real_cases": True,
"skip_dasha_audit": True,
"skip_oracle_audit": True,
"skip_local_accuracy_report": True,
"skip_vedastro_live": True,
},
}
DASHA_REFERENCE_AUDIT_CMD = [
PYTHON,
"scripts/dasha_reference_audit.py",
"--year",
"1990",
"--month",
"4",
"--day",
"17",
"--hour",
"14",
"--minute",
"45",
"--second",
"20",
"--lat",
"36.466667",
"--lon",
"114.2",
"--tz",
"8",
"--target-start-date",
"2021-05-18",
"--target-source",
"synthetic_fixture",
]
ORACLE_BOUNDARY_AUDIT_CMD = [
PYTHON,
"scripts/oracle_boundary_audit.py",
"--oracle-file",
"references/oracle/dasha_shadbala_oracle_cases.json",
]
EXTERNAL_ORACLE_SANITY_CLOSURE_CMD = [
PYTHON,
"scripts/external_oracle_sanity_closure.py",
"--format",
"json",
]
ORACLE_COLLECTION_QUEUE_CMD = [
PYTHON,
"scripts/oracle_collection_queue.py",
"--oracle-file",
"references/oracle/dasha_shadbala_oracle_cases.json",
"--format",
"json",
]
ORACLE_COLLECTION_QUEUE_EXPECTED_FIELDS = ["evidence_packet", "capture_id", "target_fields"]
ORACLE_EVIDENCE_VALIDATOR_CMD = [
PYTHON,
"scripts/oracle_evidence_validator.py",
"--queue-file",
"{queue_file}",
]
def tail_text(text: str, *, limit: int = 2400) -> str:
text = text.strip()
if len(text) <= limit:
return text
return f"...\n{text[-limit:]}"
def extract_json_payload(text: str) -> dict:
text = text.strip()
if not text:
return {}
for start in [text.find("{"), text.rfind("{")]:
if start < 0:
continue
try:
payload = json.loads(text[start:])
except json.JSONDecodeError:
continue
if isinstance(payload, dict):
return payload
return {}
def format_failure_summary(
step: str,
cmd: list[str],
returncode: int,
*,
stdout: str = "",
stderr: str = "",
cwd: Path = ROOT,
) -> str:
combined_payload = extract_json_payload(stderr) or extract_json_payload(stdout)
reason = combined_payload.get("reason") if isinstance(combined_payload, dict) else None
snapshot = combined_payload.get("process_snapshot") if isinstance(combined_payload, dict) else None
lines = [
"",
"== Quality gate failed ==",
f"step: {step}",
f"exit code: {returncode}",
f"cwd: {cwd.relative_to(ROOT) if cwd != ROOT else '.'}",
f"command: {' '.join(cmd)}",
]
if reason:
lines.append(f"reason: {reason}")
if snapshot:
lines.append(f"process_snapshot: {json.dumps(snapshot, ensure_ascii=False)}")
if stdout.strip():
lines.append(f"stdout tail:\n{tail_text(stdout)}")
if stderr.strip():
lines.append(f"stderr tail:\n{tail_text(stderr)}")
lines.extend([
"普通用户启动路径:",
"1. 本地 API 服务:.venv/bin/python scripts/jyotish_api_server.py --host 127.0.0.1 --port 5200",
"2. 网页服务:npm run dev --prefix frontend",
"3. Open http://127.0.0.1:3000 and verify /api/health.",
"Next action: Run the focused command above, then rerun the affected Python or Next.js check.",
"",
])
return "\n".join(lines)
def run(cmd: list[str], *, optional: bool = False, step: str | None = None, cwd: Path = ROOT) -> bool:
label = step or " ".join(cmd[:2])
print(f"\n$ {' '.join(cmd)}")
completed = subprocess.run(cmd, cwd=cwd, text=True)
if completed.returncode == 0:
return True
if optional:
print(f"Optional step failed with exit code {completed.returncode}; continuing.")
return False
print(
format_failure_summary(label, cmd, completed.returncode, stdout="", stderr="", cwd=cwd),
file=sys.stderr,
)
raise SystemExit(completed.returncode)
def run_oracle_collection_queue_and_validator() -> None:
with tempfile.NamedTemporaryFile("w", suffix=".json", delete=False, encoding="utf-8") as handle:
queue_path = Path(handle.name)
try:
print(f"\n$ {' '.join(ORACLE_COLLECTION_QUEUE_CMD)}")
completed = subprocess.run(ORACLE_COLLECTION_QUEUE_CMD, cwd=ROOT, text=True, capture_output=True)
if completed.stdout:
print(completed.stdout, end="" if completed.stdout.endswith("\n") else "\n")
if completed.stderr:
print(completed.stderr, end="" if completed.stderr.endswith("\n") else "\n", file=sys.stderr)
if completed.returncode != 0:
print(
format_failure_summary(
"oracle_collection_queue",
ORACLE_COLLECTION_QUEUE_CMD,
completed.returncode,
stdout=completed.stdout,
stderr=completed.stderr,
),
file=sys.stderr,
)
raise SystemExit(completed.returncode)
queue_path.write_text(completed.stdout, encoding="utf-8")
validator_cmd = [part if part != "{queue_file}" else str(queue_path) for part in ORACLE_EVIDENCE_VALIDATOR_CMD]
run(validator_cmd, step="oracle_evidence_validator")
finally:
with contextlib.suppress(FileNotFoundError):
queue_path.unlink()
def git_untracked_files() -> set[str]:
completed = subprocess.run(
["git", "ls-files", "--others", "--exclude-standard"],
cwd=ROOT,
text=True,
capture_output=True,
check=False,
)
if completed.returncode != 0:
raise SystemExit(
format_failure_summary(
"release_hygiene_check",
["git", "ls-files", "--others", "--exclude-standard"],
completed.returncode,
stdout=completed.stdout,
stderr=completed.stderr,
)
)
return {line.strip() for line in completed.stdout.splitlines() if line.strip()}
def release_hygiene_check(require_external_parity: bool = False) -> None:
print("\n== Release hygiene check ==")
untracked = git_untracked_files()
critical = [path for path in RELEASE_CRITICAL_UNTRACKED_PATHS if path in untracked]
if critical:
payload = {
"reason": "release_critical_untracked_files",
"untracked_count": len(critical),
"untracked_files": critical,
"next_action": "Stage or commit these product files before running the release profile.",
}
print(json.dumps(payload, ensure_ascii=False, indent=2), file=sys.stderr)
raise SystemExit(1)
run([PYTHON, "scripts/public_release_privacy_scan.py", "--json"])
run([PYTHON, "scripts/report_renderer_isolation_poc.py", "--strict"])
parity_command = [PYTHON, "scripts/three_engine_parity_replay_validator.py", "references/oracle/three_engine_parity_replay_manifest.json"]
if require_external_parity:
parity_command.append("--require-pass")
run(parity_command)
print("release_hygiene_check ok: no release-critical product files are untracked")
def run_vedastro_live_smoke() -> None:
print("\n== VedAstro live adapter smoke ==")
endpoint = os.environ.get("VEDASTRO_API_ENDPOINT", "").strip()
network_enabled = os.environ.get("VEDASTRO_ENABLE_NETWORK", "").strip().lower() in {"1", "true", "yes"}
if not endpoint or not network_enabled:
print(json.dumps({
"status": "blocked",
"reason": "vedastro_live_endpoint_or_network_flag_missing",
"required_env": {
"endpoint": "VEDASTRO_API_ENDPOINT",
"network": "VEDASTRO_ENABLE_NETWORK",
},
"boundary": "Default CI stays deterministic; configure both env vars to run a real VedAstro live smoke.",
}, ensure_ascii=False, indent=2))
return
run([
PYTHON,
"scripts/vedastro_service_adapter.py",
"--range-scan",
"--domain",
"career",
"--case",
"beijing_first_use_demo",
"--start-date",
"2026-01-01",
"--end-date",
"2026-12-31",
])
def compile_targets() -> None:
print("\n== Compile core Python files ==")
targets: list[Path] = []
for directory in COMPILE_DIRS:
targets.extend(sorted(directory.glob("*.py")))
targets.extend(EXTRA_COMPILE_TARGETS)
seen: set[Path] = set()
for target in targets:
if target in seen or not target.exists():
continue
seen.add(target)
print(f"compile {target.relative_to(ROOT)}")
py_compile.compile(str(target), doraise=True)
def validate_json_files() -> None:
print("\n== Validate critical JSON files ==")
for relative in [
"references/technique_registry.json",
"references/yoga_rules.json",
"references/standard_test_charts.json",
"references/validation_logic_report.json",
"references/oracle/dasha_shadbala_oracle_cases.json",
"tests/golden/golden_cases.json",
]:
path = ROOT / relative
if not path.exists():
continue
with path.open("r", encoding="utf-8") as handle:
json.load(handle)
print(f"json ok {relative}")
def run_profile(args: argparse.Namespace) -> dict:
profile = dict(QUALITY_GATE_PROFILES[args.profile])
for key in [
"skip_slow",
"skip_yoga_logic",
"skip_frontend_runtime",
"skip_real_cases",
"skip_dasha_audit",
"skip_oracle_audit",
"skip_local_accuracy_report",
"skip_vedastro_live",
]:
if getattr(args, key):
profile[key] = True
return profile
def main() -> int:
parser = argparse.ArgumentParser(description="Run Jyotish skill quality gate")
parser.add_argument("--profile", choices=["quick", "browser", "release", "accuracy", "vedastro-live", "runtime-truth"], default="browser", help="Quality gate profile: quick, browser, release, accuracy, vedastro-live, or runtime-truth")
parser.add_argument("--skip-slow", action="store_true", help="Skip slow golden-case regressions")
parser.add_argument("--skip-yoga-logic", action="store_true", help="Skip Yoga logic comparison report refresh")
parser.add_argument("--skip-frontend-runtime", action="store_true", help="Skip Next.js tests, lint, and production build")
parser.add_argument("--skip-real-cases", action="store_true", help="Skip public real-person chart revalidation")
parser.add_argument("--skip-dasha-audit", action="store_true", help="Skip Dasha reference-drift audit")
parser.add_argument("--skip-oracle-audit", action="store_true", help="Skip combined Dasha/Shadbala external oracle boundary audit")
parser.add_argument("--skip-local-accuracy-report", action="store_true", help="Skip consolidated local accuracy report")
parser.add_argument("--skip-vedastro-live", action="store_true", help="Skip optional VedAstro live endpoint smoke")
parser.add_argument("--all-tests", action="store_true", help="Run every pytest file, including optional-dependency suites")
parser.add_argument("--require-external-parity", action="store_true", help="Fail the release gate unless the three-engine raw parity manifest passes.")
args = parser.parse_args()
profile = run_profile(args)
os.environ.setdefault("PYTHONPATH", str(ROOT / "scripts"))
print(f"\n== Quality gate profile: {args.profile} ==")
print(json.dumps(profile, ensure_ascii=False, indent=2))
if args.profile == "runtime-truth":
for target in [
ROOT / "scripts" / "jyotish_api_server.py",
ROOT / "scripts" / "diagnose_vedastro_mode.py",
ROOT / "scripts" / "diagnose_external_engine_adapters.py",
ROOT / "scripts" / "interpretation_source_runtime_coverage.py",
ROOT / "scripts" / "sync_final_evidence_packet_status.py",
ROOT / "scripts" / "external_validation_release_gate.py",
]:
py_compile.compile(str(target), doraise=True)
print(f"compiled {target.relative_to(ROOT)}")
run([PYTHON, "scripts/sync_final_evidence_packet_status.py"])
run([PYTHON, "scripts/interpretation_source_inventory_gate.py"])
run([PYTHON, "scripts/diagnose_vedastro_mode.py", "--json"])
run([PYTHON, "scripts/diagnose_external_engine_adapters.py", "--json"])
run([PYTHON, "scripts/external_validation_release_gate.py", "--require-match"])
else:
compile_targets()
validate_json_files()
run([PYTHON, "scripts/audit_capabilities.py", "--mode", "validate"])
run([PYTHON, "scripts/audit_fragments.py", "--strict"])
run([PYTHON, "scripts/interpretation_source_inventory_gate.py"])
run([PYTHON, "scripts/character_level_inventory_manifest.py", "--scope", "project", "--no-write", "--summary-only"])
if profile["check_release_hygiene"]:
release_hygiene_check(require_external_parity=args.require_external_parity)
run([PYTHON, "scripts/validate_bphs_invariants.py"])
if args.all_tests:
pytest_targets = ["tests"]
elif args.profile == "runtime-truth":
pytest_targets = RUNTIME_TRUTH_PYTEST_TARGETS
else:
pytest_targets = CORE_PYTEST_TARGETS
run([PYTHON, "-m", "pytest", *pytest_targets])
if not profile["skip_frontend_runtime"]:
run(["npm", "test"], optional=False, cwd=APP)
run(["npm", "run", "lint"], optional=False, cwd=APP)
run(["npm", "run", "build"], optional=False, cwd=APP)
if not profile["skip_slow"]:
run([PYTHON, "tests/run_golden_cases.py", "--python", PYTHON])
if not profile["skip_real_cases"]:
run([PYTHON, "tests/run_real_case_revalidation.py", "--python", PYTHON, "--summary"])
if not profile["skip_dasha_audit"]:
run(DASHA_REFERENCE_AUDIT_CMD)
if not profile["skip_oracle_audit"]:
run(ORACLE_BOUNDARY_AUDIT_CMD)
run(EXTERNAL_ORACLE_SANITY_CLOSURE_CMD)
run_oracle_collection_queue_and_validator()
if not profile["skip_yoga_logic"]:
run([PYTHON, "scripts/validate_logic_v2.py"], optional=True)
if not profile["skip_local_accuracy_report"]:
run([PYTHON, "scripts/local_accuracy_report.py", "--format", "json"])
if not profile["skip_vedastro_live"]:
run_vedastro_live_smoke()
print("\nQuality gate passed.")
return 0
if __name__ == "__main__":
raise SystemExit(main())