strengthen external parity diagnostics

benchmarks/jyotish/scripts/run_pyjhora_compare.py

@@ -3,6 +3,7 @@
 # or JYOTISH_SKILL_SCRIPT is provided. Raw output directories are generated locally
 # and are intentionally not committed.
 #!/usr/bin/env python3
+import argparse
 import csv
 import json
 import os
@@ -99,7 +100,7 @@ def tuple_to_date(t):
     return f'{int(y):04d}-{int(m):02d}-{int(d):02d}'
 
 
-def build_pyjhora_sample(sample):
+def build_pyjhora_sample(sample, *, node_mode='mean'):
     swe = patch_swisseph()
     from jhora import utils, const
     from jhora.panchanga import drik
@@ -109,6 +110,8 @@ def build_pyjhora_sample(sample):
     # Align benchmark口径: Lahiri + mean sidereal year. PyJHora default is TRUE_PUSHYA.
     const._DEFAULT_AYANAMSA_MODE = 'LAHIRI'
     drik.set_ayanamsa_mode('LAHIRI')
+    const.set_node_mode(node_mode == 'true')
+    drik.set_planet_list(set_rahu_ketu_as_true_nodes=(node_mode == 'true'))
     try:
         const.dhasa_year_duration_default = const.DHASA_YEAR_DURATION.MEAN_SIDEREAL_YEAR
     except Exception:
@@ -162,6 +165,7 @@ def build_pyjhora_sample(sample):
         dasha['error'] = f'{type(exc).__name__}: {exc}'
 
     return {
+        'settings': {'ayanamsa': 'lahiri', 'node_mode': node_mode},
         'sample_id': sample['id'],
         'engine': 'PyJHora_4_8_6_lahiri_patched',
         'parameters': {
@@ -320,14 +324,36 @@ def write_report(samples, rows):
     return '\n'.join(lines)
 
 
-def main():
+def main(argv=None):
+    parser = argparse.ArgumentParser(description='Compare public benchmark samples against PyJHora.')
+    parser.add_argument('--sample-id', action='append', default=[], help='Run only a named benchmark sample; repeatable.')
+    parser.add_argument('--build-local', action='store_true', help='Explicitly generate missing local canonical baselines.')
+    parser.add_argument('--node-mode', choices=['mean', 'true'], default='mean', help='Match the node convention before comparing.')
+    args = parser.parse_args(argv)
     PYJHORA_OUT.mkdir(parents=True, exist_ok=True)
     samples = json.loads(DATA.read_text())
+    if args.sample_id:
+        requested = set(args.sample_id)
+        samples = [sample for sample in samples if sample['id'] in requested]
+        missing = requested - {sample['id'] for sample in samples}
+        if missing:
+            parser.error(f'unknown sample id(s): {", ".join(sorted(missing))}')
     all_rows = []
     for sample in samples:
-        pyjhora = build_pyjhora_sample(sample)
+        local_path = LOCAL_CANON / f"{sample['id']}.canonical.json"
+        if not local_path.exists() and args.build_local:
+            from run_skill_baseline import run_sample
+            baseline = run_sample(sample)
+            if not baseline.get('ok'):
+                parser.error(f'failed to build local baseline for {sample["id"]}: {baseline.get("error", "unknown error")}')
+        if not local_path.exists():
+            parser.error(
+                f'missing local canonical baseline for {sample["id"]}; '
+                'run with --build-local or run_skill_baseline.py first'
+            )
+        pyjhora = build_pyjhora_sample(sample, node_mode=args.node_mode)
         (PYJHORA_OUT / f"{sample['id']}.pyjhora.json").write_text(json.dumps(pyjhora, ensure_ascii=False, indent=2))
-        local = json.loads((LOCAL_CANON / f"{sample['id']}.canonical.json").read_text())
+        local = json.loads(local_path.read_text())
         all_rows.extend(compare_one(sample['id'], local, pyjhora))
 
     matrix = OUT / 'pyjhora_comparison_matrix.csv'

benchmarks/jyotish/scripts/run_skill_baseline.py

@@ -10,7 +10,7 @@ import sys
 from pathlib import Path
 
 ROOT = Path(__file__).resolve().parents[1]
-SKILL_SCRIPT = Path(__file__).resolve().parents[2] / 'scripts' / 'jyotish_engine.py'
+SKILL_SCRIPT = Path(__file__).resolve().parents[3] / 'scripts' / 'jyotish_engine.py'
 PYTHON = Path(__import__('sys').executable)
 DATA = ROOT / 'data/benchmark_samples.json'
 OUT = ROOT / 'outputs'
@@ -31,6 +31,8 @@ def safe_get(obj, *keys, default=None):
 
 
 def run_sample(sample):
+    RAW.mkdir(parents=True, exist_ok=True)
+    CANON.mkdir(parents=True, exist_ok=True)
     birth = sample['birth']
     cmd = [
         str(PYTHON), str(SKILL_SCRIPT), 'full-reading',

docs/research/pre_work_error_ledger.md

@@ -79,6 +79,7 @@ For large architecture or release work, also read:
 | ERR-046 | Report-renderer SSRF/file PoC could not run because the Playwright Chromium binary was absent and installation exceeded the desktop outer timeout. | blocked environment | Keep route/JS-denial tests; rerun isolated HTTP/file PoC only after a verified Chromium installation, then update this ledger with the measured request count. |
 | ERR-047 | Initial `slow` marker partition for `test_api_server_security.py` still exceeded the 120-second desktop budget; heavy paths extend beyond VedAstro/high-rigor prefix groups. | active profiling blocker | Profile test node IDs in bounded subprocess batches, mark only measured heavy tests, and keep fast-security acceptance separate from long CI integration coverage. |
 | ERR-048 | Candidate-time scanner assumed all documented D4/D24/D30 divisions were exposed by `jyotish_engine.py varga`; actual `--d4` failed at runtime. | mitigated 2026-07-12 | Candidate scans must record unsupported Varga flags as `unavailable_vargas`; only successfully computed D1/D9/D10 fields may drive local sensitivity output until a unified Varga contract exists. |
+| ERR-049 | PyJHora benchmark runner executed on `--help`, used a wrong repository-root path in `run_skill_baseline.py`, and failed when reused without pre-created output directories. | mitigated 2026-07-12 | Keep `tests/test_pyjhora_compare_cli.py`; require explicit `--build-local`, safe argparse help, correct repo root, and directory creation inside `run_sample()`. |
 
 ## Fragment Sweep Command Set
 

scripts/candidate_time_sensitivity_scan.py

@@ -39,6 +39,21 @@ def _varga_ascendant(payload: dict[str, Any], varga: str) -> str | None:
     return None
 
 
+def _all_varga_ascendants(payload: dict[str, Any]) -> dict[str, str | None]:
+    values = {varga.upper(): None for varga in _VARGAS}
+    try:
+        raw = _engine_json("varga", {**payload, "varga": "all"})
+    except subprocess.CalledProcessError:
+        return values
+    for name, chart in (raw.get("divisional_charts") or {}).items():
+        if not isinstance(chart, dict):
+            continue
+        for varga in _VARGAS:
+            if name.startswith(varga.upper() + "_"):
+                values[varga.upper()] = chart.get("ascendant")
+    return values
+
+
 def scan_candidate_times(payload: dict[str, Any], *, uncertainty_minutes: int = 30, step_minutes: int = 1) -> dict[str, Any]:
     required = ("year", "month", "day", "hour", "minute", "lat", "lon", "tz")
     missing = [key for key in required if payload.get(key) is None]
@@ -53,7 +68,7 @@ def scan_candidate_times(payload: dict[str, Any], *, uncertainty_minutes: int =
         point = {**payload, "year": moment.year, "month": moment.month, "day": moment.day, "hour": moment.hour, "minute": moment.minute}
         chart = _engine_json("chart", point)
         asc = chart.get("ascendant", {})
-        divisional = {varga.upper(): _varga_ascendant(point, varga) for varga in _VARGAS}
+        divisional = _all_varga_ascendants(point)
         rows.append({
             "time": moment.strftime("%Y-%m-%d %H:%M"),
             "offset_minutes": offset,
@@ -62,6 +77,8 @@ def scan_candidate_times(payload: dict[str, Any], *, uncertainty_minutes: int =
             "divisional_ascendants": divisional,
         })
     signatures = [tuple([row["d1_ascendant"], *row["divisional_ascendants"].values()]) for row in rows]
+    unavailable_vargas = [varga.upper() for varga in _VARGAS if all(row["divisional_ascendants"][varga.upper()] is None for row in rows)]
+    supported_vargas = [varga.lower() for varga in _VARGAS if varga.upper() not in unavailable_vargas]
     modal = Counter(signatures).most_common(1)[0][0]
     for row, signature in zip(rows, signatures):
         row["sensitivity_count"] = sum(left != right for left, right in zip(signature, modal))
@@ -69,7 +86,6 @@ def scan_candidate_times(payload: dict[str, Any], *, uncertainty_minutes: int =
             name for name, current, typical in zip(("D1", "D4", "D9", "D10", "D24", "D30"), signature, modal)
             if current != typical
         ]
-    unavailable_vargas = [varga.upper() for varga in _VARGAS if all(row["divisional_ascendants"][varga.upper()] is None for row in rows)]
     transitions = []
     for previous, current in zip(rows, rows[1:]):
         changed = [name for name in ("d1_ascendant", "divisional_ascendants") if previous[name] != current[name]]
@@ -85,6 +101,7 @@ def scan_candidate_times(payload: dict[str, Any], *, uncertainty_minutes: int =
         "step_minutes": step_minutes,
         "rows": rows,
         "transitions": transitions,
+        "supported_vargas": [varga.upper() for varga in supported_vargas],
         "unavailable_vargas": unavailable_vargas,
         "pending_layers": ["UL", "A7", "A10", "KP_cusp"],
         "boundary": "Actual local D1/Varga differences only. Unsupported Varga CLI flags are explicitly unavailable. Event answers still require an explicit event-to-candidate adjudication model before minute-level rectification.",

scripts/jyotish_api_server.py

@@ -1308,10 +1308,12 @@ class JyotishAPIHandler(BaseHTTPRequestHandler):
             elif path == '/api/rectification/questionnaire':
                 self._json(build_rectification_questionnaire(body))
             elif path == '/api/rectification/sensitivity_scan':
+                uncertainty = int(body.get('time_uncertainty_minutes') or 30)
+                step_minutes = int(body.get('step_minutes') or (5 if uncertainty > 15 else 1))
                 self._json(scan_candidate_times(
                     body,
-                    uncertainty_minutes=int(body.get('time_uncertainty_minutes') or 30),
-                    step_minutes=int(body.get('step_minutes') or 1),
+                    uncertainty_minutes=uncertainty,
+                    step_minutes=step_minutes,
                 ))
             elif path == '/api/rectification/answers':
                 questionnaire = body.get('questionnaire')

scripts/pyjhora_parity_summary.py

@@ -0,0 +1,61 @@
+#!/usr/bin/env python3
+"""Summarize reviewable PyJHora comparison matrices without overstating coverage."""
+
+from __future__ import annotations
+
+import argparse
+import csv
+import json
+from collections import Counter, defaultdict
+from pathlib import Path
+from typing import Any
+
+
+REQUIRED_FULL_PARITY = ("D1", "D9", "D10", "D2", "D4", "Vimshottari", "Shadbala", "Ashtakavarga")
+MATRIX_SECTION_MAP = {"ascendant": "D1", "planet": "D1", "dasha": "Vimshottari", "D9": "D9", "D10": "D10"}
+
+
+def summarize_matrix(path: str | Path, *, settings: dict[str, Any]) -> dict[str, Any]:
+    path = Path(path)
+    rows = list(csv.DictReader(path.open(encoding="utf-8")))
+    status_counts = Counter(str(row.get("status") or "unknown") for row in rows)
+    sections: dict[str, dict[str, int]] = defaultdict(lambda: {"total": 0, "match": 0, "mismatch": 0})
+    covered = set()
+    for row in rows:
+        section = MATRIX_SECTION_MAP.get(str(row.get("section") or ""))
+        if not section:
+            continue
+        covered.add(section)
+        sections[section]["total"] += 1
+        if row.get("status") == "match":
+            sections[section]["match"] += 1
+        elif row.get("status") == "mismatch":
+            sections[section]["mismatch"] += 1
+    missing = [field for field in REQUIRED_FULL_PARITY if field not in covered]
+    return {
+        "scope": "pyjhora_same_chart_parity_summary",
+        "matrix_path": str(path),
+        "tested": bool(rows),
+        "settings": settings,
+        "row_counts": dict(status_counts),
+        "coverage": dict(sorted(sections.items())),
+        "covered_outputs": sorted(covered),
+        "missing_required_outputs": missing,
+        "status": "partial_verified" if rows and not status_counts.get("mismatch") else "partial_mismatch",
+        "full_parity_verified": not missing and not status_counts.get("mismatch"),
+        "boundary": "Only covered outputs are compared. This summary cannot promote full parity while required outputs are absent.",
+    }
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("matrix")
+    parser.add_argument("--ayanamsa", default="lahiri")
+    parser.add_argument("--node-mode", default="mean", choices=["mean", "true"])
+    args = parser.parse_args()
+    print(json.dumps(summarize_matrix(args.matrix, settings={"ayanamsa": args.ayanamsa, "node_mode": args.node_mode}), ensure_ascii=False, indent=2))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

tests/test_candidate_time_sensitivity_scan.py

@@ -7,7 +7,12 @@ def test_scanner_reports_real_divisional_transitions(monkeypatch):
         if command == "chart":
             return {"ascendant": {"sign": "Leo", "degree_in_sign": 10 + minute / 100}}
         ascendant = "Aries" if minute % 2 else "Taurus"
-        return {"divisional_charts": {"D": {"ascendant": ascendant}}}
+        return {
+            "divisional_charts": {
+                "D9_Navamsa": {"ascendant": ascendant},
+                "D10_Dasamsa": {"ascendant": ascendant},
+            }
+        }
 
     monkeypatch.setattr(scanner, "_engine_json", fake_engine)
     report = scanner.scan_candidate_times(

tests/test_pyjhora_compare_cli.py

@@ -0,0 +1,17 @@
+import subprocess
+import sys
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[1]
+RUNNER = ROOT / "benchmarks" / "jyotish" / "scripts" / "run_pyjhora_compare.py"
+
+
+def test_pyjhora_compare_help_is_non_executing():
+    result = subprocess.run(
+        [sys.executable, str(RUNNER), "--help"], cwd=ROOT, capture_output=True, text=True, timeout=15
+    )
+
+    assert result.returncode == 0
+    assert "--build-local" in result.stdout
+    assert "FileNotFoundError" not in result.stderr

tests/test_pyjhora_parity_summary.py

@@ -0,0 +1,17 @@
+from pathlib import Path
+
+from scripts.pyjhora_parity_summary import summarize_matrix
+
+
+def test_summary_marks_partial_verified_when_only_d1_d9_d10_dasha_are_covered(tmp_path: Path):
+    matrix = tmp_path / "matrix.csv"
+    matrix.write_text(
+        "section,status\nascendant,match\nD9,match\nD10,match\ndasha,match\n",
+        encoding="utf-8",
+    )
+
+    result = summarize_matrix(matrix, settings={"ayanamsa": "lahiri", "node_mode": "mean"})
+
+    assert result["status"] == "partial_verified"
+    assert result["full_parity_verified"] is False
+    assert result["missing_required_outputs"] == ["D2", "D4", "Shadbala", "Ashtakavarga"]

web/rectification.html

@@ -9,7 +9,7 @@ let questionnaire;
 const asObject=f=>Object.fromEntries(new FormData(f).entries());
 const fail=e=>document.querySelector('#error').textContent=`请求失败：${e.message}。请检查输入后重试。`;
 async function post(url,body){const r=await fetch(url,{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify(body)});const d=await r.json();if(!r.ok)throw new Error(d.error||r.status);return d}
-document.querySelector('#birth').onsubmit=async e=>{e.preventDefault();document.querySelector('#error').textContent='正在计算候选盘…';try{const p=asObject(e.target);for(const k of Object.keys(p))p[k]=Number(p[k]);const [q,scan]=await Promise.all([post('/api/rectification/questionnaire',p),post('/api/rectification/sensitivity_scan',p)]);questionnaire=q;document.querySelector('#error').textContent='';document.querySelector('#scan').innerHTML=`<h2>实际候选盘差异</h2><pre>${JSON.stringify({candidate_count:scan.candidate_count,transitions:scan.transitions,unavailable_vargas:scan.unavailable_vargas,pending_layers:scan.pending_layers,boundary:scan.boundary},null,2)}</pre>`;render(questionnaire.questions||[])}catch(err){fail(err)}};
+document.querySelector('#birth').onsubmit=async e=>{e.preventDefault();document.querySelector('#error').textContent='正在计算候选盘…';try{const p=asObject(e.target);for(const k of Object.keys(p))p[k]=Number(p[k]);const [q,scan]=await Promise.all([post('/api/rectification/questionnaire',p),post('/api/rectification/sensitivity_scan',p)]);questionnaire=q;document.querySelector('#error').textContent='';document.querySelector('#scan').innerHTML=`<h2>实际候选盘差异</h2><pre>${JSON.stringify({candidate_count:scan.candidate_count,step_minutes:scan.step_minutes,transitions:scan.transitions,supported_vargas:scan.supported_vargas,unavailable_vargas:scan.unavailable_vargas,pending_layers:scan.pending_layers,boundary:scan.boundary},null,2)}</pre>`;render(questionnaire.questions||[])}catch(err){fail(err)}};
 function render(qs){const root=document.querySelector('#questions');root.innerHTML=qs.map(q=>`<section class="question"><strong>${q.prompt}</strong>${q.options.map(o=>`<label><input type="radio" name="${q.id}" value="${o.key}"> ${o.key}. ${o.label}</label>`).join('')}</section>`).join('')+'<button id="score">提交本轮答案</button>';document.querySelector('#score').onclick=score}
 async function score(){try{const answers={};document.querySelectorAll('#questions input:checked').forEach(e=>answers[e.name]=e.value);const d=await post('/api/rectification/answers',{questionnaire,answers});document.querySelector('#result').innerHTML=`<h2>候选簇排序</h2><pre>${JSON.stringify({candidate_cluster_rankings:d.candidate_cluster_rankings,next_round:d.next_round,boundary:d.boundary},null,2)}</pre>`}catch(err){fail(err)}}
 </script></html>