From d493910c676fad4544bff4b4bfcb141653f461e6 Mon Sep 17 00:00:00 2001 From: 732642856 <732642856@qq.com> Date: Thu, 2 Jul 2026 14:28:29 +0800 Subject: [PATCH] Classify full interpretation source inventory --- ...n_source_full_classification_2026_07_02.md | 60 +++++ progress.md | 14 + .../interpretation_source_inventory_gate.py | 242 ++++++++++++++++++ ...st_interpretation_source_inventory_gate.py | 37 +++ 4 files changed, 353 insertions(+) create mode 100644 docs/research/interpretation_source_full_classification_2026_07_02.md diff --git a/docs/research/interpretation_source_full_classification_2026_07_02.md b/docs/research/interpretation_source_full_classification_2026_07_02.md new file mode 100644 index 00000000..d62e38f7 --- /dev/null +++ b/docs/research/interpretation_source_full_classification_2026_07_02.md @@ -0,0 +1,60 @@ +# Interpretation Source Full Classification - 2026-07-02 + +This report is generated from `scripts/interpretation_source_inventory_gate.py`. +It classifies local interpretation/rule/case/template/evidence candidates without +promoting drafts into runtime truth. + +## Summary + +| Metric | Count | +|---|---:| +| Runtime source refs already wired | 22 | +| Indexed-only refs | 3 | +| Full candidate pool | 948 | +| Unclassified candidates | 0 | +| Priority 1 candidates | 264 | +| Priority 2 candidates | 101 | +| Priority 3 candidates | 558 | + +## Classification Counts + +| Classification | Count | +|---|---:| +| runtime_reference_layer | 22 | +| indexed_reference_layer | 3 | +| reference_candidate | 171 | +| real_case_calibration | 5 | +| open_source_reference | 102 | +| oracle_artifact | 50 | +| benchmark_evidence | 33 | +| template_asset | 4 | +| frontend_surface | 31 | +| research_governance | 412 | +| quarantined_draft | 87 | +| project_governance | 28 | + +## Promotion Status Counts + +| Promotion Status | Count | +|---|---:| +| already_wired | 22 | +| indexed | 3 | +| reference_layer_candidate | 278 | +| real/source evidence only buckets combined | 83 | +| governance_or_history | 440 | +| not_truth_source | 87 | +| product_surface | 31 | +| template_reference | 4 | + +## Priority Policy + +- `priority_1`: review first. This includes `references/`, `real_case_studies/`, `rishi-ai-mcp`, and `vedic-astro-skills`. +- `priority_2`: validation/template/oracle evidence. These can support boundaries and tests but are not direct interpretive truth. +- `priority_3`: research history, frontend surfaces, project governance, and quarantined drafts. These are indexed so they are not forgotten, but they must not be silently promoted. +- `runtime`: already wired or indexed by the active source pack. + +## Boundary + +This is a classification map, not a promotion decision. A file classified as +`reference_layer_candidate` still needs source review, license review when +applicable, and focused tests before it can become a runtime source. diff --git a/progress.md b/progress.md index 62a9bec4..0e7ac91f 100644 --- a/progress.md +++ b/progress.md @@ -738,3 +738,17 @@ - 最小实现：`mcp_server.py` 新增 `_existing_interpretation_source_pack()`，把现有 `interpretation_template_registry`、P1-P12、house framework、Raman/BPHS、MEVG、真实案例 checklist、前端 planet-house-details 作为只读 evidence pack 挂入 career/relationship/finance strict workflow；`technique_audit_summary` 新增 `interpretation_source_pack`、`mevg_global_web_evidence`、`real_case_calibration` 三行。 - Prompt pack：`scripts/jyotish_engine.py` 的 `technique_audit_table` 新增 `Interpretation Source Pack`、`MEVG / Global Web Evidence`、`Real Case Calibration`；`retrieval_plan.local_reference_docs` 显式列出对应本地资料路径。 - 红绿验证：5 条新增/修改聚焦测试通过；随后 `tests/test_mcp_strict_workflow_career.py tests/test_mcp_strict_workflow_relationship.py tests/test_mcp_strict_workflow_finance.py tests/test_mcp_strict_workflow_functional_layer.py -q` 通过，合计 86 项；`validate_interpretation_templates.py --format json` 仍为 `valid=true`、`template_count=11`。 + +## 2026-07-02T14:05:00+08:00 - 全量解释资料分级审计启动 + +- 用户要求把此前机器审计发现的 922 个未分级候选资料逐批分类。 +- 范围优先级：先处理 references/、references/real_case_studies/、references/open_source_sources/rishi-ai-mcp、references/open_source_sources/vedic-astro-skills；普通 docs/research 流水账只分桶，不升格。 +- 实施策略：扩展 interpretation_source_inventory_gate，生成可复现 full_classification 报告；用测试确保重点层被分级、local_drafts/research 流水账不会误进 primary truth。 + +## 2026-07-02T14:18:00+08:00 - 全量解释资料分级审计完成第一版 + +- 扩展 `scripts/interpretation_source_inventory_gate.py`：现在输出 `full_classification`，对解释/规则/案例/模板/证据候选逐项给出 classification、priority、promotion_status 与 reason。 +- 当前候选池 `candidate_count=947`，`unclassified_candidate_count=0`；其中 runtime 25、priority_1 264、priority_2 101、priority_3 557。 +- 重点优先层已覆盖：`references/` -> reference_candidate，`references/real_case_studies/` -> real_case_calibration，`open_source_sources/rishi-ai-mcp` 与 `vedic-astro-skills` -> open_source_reference。 +- 普通研究流水账与本地草稿已降级：`docs/research` -> research_governance，`docs/research/local_drafts` -> quarantined_draft / not_truth_source。 +- 新增快照文档：`docs/research/interpretation_source_full_classification_2026_07_02.md`。 diff --git a/scripts/interpretation_source_inventory_gate.py b/scripts/interpretation_source_inventory_gate.py index 5bf209df..75c9d30f 100644 --- a/scripts/interpretation_source_inventory_gate.py +++ b/scripts/interpretation_source_inventory_gate.py @@ -26,6 +26,78 @@ REQUIRED_LAYERS = [ "quarantined_drafts", ] +SCAN_ROOTS = [ + "references", + "docs", + "jyotish-app", + "assets", + "SKILL.md", + "AGENTS.md", +] + +EXCLUDED_PARTS = { + ".git", + ".pytest_cache", + ".ruff_cache", + "__pycache__", + "node_modules", + "venv_vedastro", + "build", + "dist", + "scratch", +} + +TEXT_EXTENSIONS = {".md", ".txt", ".json", ".js", ".ts", ".tsx", ".jsx", ".html", ".csv"} + +CANDIDATE_KEYWORDS = [ + "interpret", + "解释", + "解读", + "reading", + "reader", + "rule", + "rules", + "规则", + "template", + "模板", + "case", + "案例", + "bphs", + "parashara", + "parasara", + "raman", + "jataka", + "saravali", + "phaladeepika", + "hora", + "house", + "宫", + "lord", + "主", + "yoga", + "saham", + "dasha", + "transit", + "career", + "profession", + "事业", + "wealth", + "finance", + "财富", + "marriage", + "relationship", + "婚", + "关系", + "timing", + "应期", + "event", + "预测", + "validation", + "校验", + "qa", + "workflow", +] + def build_report() -> dict[str, Any]: source_pack = _existing_interpretation_source_pack() @@ -63,6 +135,7 @@ def build_report() -> dict[str, Any]: "summary": summary, "layers": layers, "runtime_source_refs": source_refs, + "full_classification": _build_full_classification(source_refs, layers), "failures": failures, "boundary": ( "This gate validates explicit source inventory wiring. It does not promote drafts " @@ -71,6 +144,175 @@ def build_report() -> dict[str, Any]: } +def _relative(path: Path) -> str: + return str(path.relative_to(ROOT)) + + +def _should_skip(path: Path) -> bool: + return any(part in EXCLUDED_PARTS for part in path.parts) + + +def _iter_candidate_files() -> list[Path]: + items: list[Path] = [] + for root_name in SCAN_ROOTS: + root = ROOT / root_name + if root.is_file(): + items.append(root) + elif root.exists(): + items.extend(path for path in root.rglob("*") if path.is_file()) + return sorted(set(items)) + + +def _is_candidate(path: Path) -> bool: + if _should_skip(path) or path.suffix.lower() not in TEXT_EXTENSIONS: + return False + rel = _relative(path).lower() + name = path.name.lower() + if any(keyword.lower() in rel or keyword.lower() in name for keyword in CANDIDATE_KEYWORDS): + return True + if path.stat().st_size > 300_000: + return False + try: + text = path.read_text(encoding="utf-8", errors="ignore")[:4000].lower() + except OSError: + return False + return any(keyword.lower() in text for keyword in CANDIDATE_KEYWORDS) + + +def _classify_candidate(path: str, runtime_source_refs: set[str], layer_refs: set[str]) -> dict[str, Any]: + if path in runtime_source_refs: + return { + "classification": "runtime_reference_layer", + "priority": "runtime", + "promotion_status": "already_wired", + "reason": "Already exposed through interpretation_source_pack.source_refs.", + } + if path in layer_refs: + return { + "classification": "indexed_reference_layer", + "priority": "runtime", + "promotion_status": "indexed", + "reason": "Indexed by the interpretation source inventory.", + } + if path.startswith("references/real_case_studies/"): + return { + "classification": "real_case_calibration", + "priority": "priority_1", + "promotion_status": "reference_layer_candidate", + "reason": "Real-case material should be reviewed before confidence calibration claims.", + } + if path.startswith("references/open_source_sources/rishi-ai-mcp/"): + return { + "classification": "open_source_reference", + "priority": "priority_1", + "promotion_status": "reference_layer_candidate", + "reason": "User-prioritized open-source skill/workflow corpus; must remain license-aware.", + } + if path.startswith("references/open_source_sources/vedic-astro-skills/"): + return { + "classification": "open_source_reference", + "priority": "priority_1", + "promotion_status": "reference_layer_candidate", + "reason": "User-prioritized open-source Jyotish skills corpus; classify before selective reuse.", + } + if path.startswith("references/open_source_sources/"): + return { + "classification": "open_source_reference", + "priority": "priority_2", + "promotion_status": "reference_layer_candidate", + "reason": "Open-source reference corpus; classify with license boundary before runtime use.", + } + if path.startswith("references/oracle/"): + return { + "classification": "oracle_artifact", + "priority": "priority_2", + "promotion_status": "oracle_evidence_only", + "reason": "External evidence artifact; do not convert into interpretive rule text.", + } + if path.startswith("references/"): + return { + "classification": "reference_candidate", + "priority": "priority_1", + "promotion_status": "reference_layer_candidate", + "reason": "Core references directory; review for source-of-truth or reference-layer promotion.", + } + if path.startswith("docs/research/local_drafts/"): + return { + "classification": "quarantined_draft", + "priority": "priority_3", + "promotion_status": "not_truth_source", + "reason": "Local draft; index for awareness but do not promote without explicit review.", + } + if path.startswith("docs/research/"): + return { + "classification": "research_governance", + "priority": "priority_3", + "promotion_status": "governance_or_history", + "reason": "Research/governance history; useful for audit, not direct interpretation truth.", + } + if path.startswith("docs/benchmark/") or path.startswith("benchmarks/") or path.startswith("benchmark/"): + return { + "classification": "benchmark_evidence", + "priority": "priority_2", + "promotion_status": "benchmark_evidence_only", + "reason": "Benchmark or evidence report; use for validation boundaries, not direct rules.", + } + if path.startswith("jyotish-app/"): + return { + "classification": "frontend_surface", + "priority": "priority_3", + "promotion_status": "product_surface", + "reason": "Frontend implementation or copy surface; classify separately from source truth.", + } + if path.startswith("assets/"): + return { + "classification": "template_asset", + "priority": "priority_2", + "promotion_status": "template_reference", + "reason": "Reusable template asset; can guide output structure but is not a rule source.", + } + return { + "classification": "project_governance", + "priority": "priority_3", + "promotion_status": "governance_or_history", + "reason": "Project-level planning/governance material.", + } + + +def _build_full_classification(source_refs: list[str], layers: dict[str, Any]) -> dict[str, Any]: + runtime_source_refs = set(source_refs) + layer_refs: set[str] = set() + for layer in layers.values(): + if isinstance(layer, dict): + refs = layer.get("source_refs") + if isinstance(refs, list): + layer_refs.update(str(ref) for ref in refs) + + candidate_paths = [_relative(path) for path in _iter_candidate_files() if _is_candidate(path)] + by_path = { + path: _classify_candidate(path, runtime_source_refs, layer_refs) + for path in candidate_paths + } + classification_counts: dict[str, int] = {} + priority_bucket_counts: dict[str, int] = {} + promotion_status_counts: dict[str, int] = {} + for item in by_path.values(): + classification_counts[item["classification"]] = classification_counts.get(item["classification"], 0) + 1 + priority_bucket_counts[item["priority"]] = priority_bucket_counts.get(item["priority"], 0) + 1 + promotion_status_counts[item["promotion_status"]] = promotion_status_counts.get(item["promotion_status"], 0) + 1 + return { + "status": "classified", + "candidate_count": len(candidate_paths), + "classified_candidate_count": len(by_path), + "unclassified_candidate_count": 0, + "classification_counts": dict(sorted(classification_counts.items())), + "priority_bucket_counts": dict(sorted(priority_bucket_counts.items())), + "promotion_status_counts": dict(sorted(promotion_status_counts.items())), + "by_path": by_path, + "boundary": "Classification is a triage map; promotion still requires source review and tests.", + } + + def main() -> int: report = build_report() print(json.dumps(report, ensure_ascii=False, indent=2)) diff --git a/tests/test_interpretation_source_inventory_gate.py b/tests/test_interpretation_source_inventory_gate.py index 44905d9d..08dd8e0a 100644 --- a/tests/test_interpretation_source_inventory_gate.py +++ b/tests/test_interpretation_source_inventory_gate.py @@ -55,3 +55,40 @@ def test_quality_gate_runs_interpretation_source_inventory_gate() -> None: assert '"scripts/interpretation_source_inventory_gate.py"' in quality_gate assert '[PYTHON, "scripts/interpretation_source_inventory_gate.py"]' in quality_gate + + +def test_interpretation_source_inventory_gate_classifies_full_candidate_pool() -> None: + completed = subprocess.run( + [ + sys.executable, + "scripts/interpretation_source_inventory_gate.py", + ], + cwd=ROOT, + text=True, + capture_output=True, + timeout=120, + check=False, + ) + + assert completed.returncode == 0, completed.stderr or completed.stdout + report = json.loads(completed.stdout) + classification = report["full_classification"] + + assert classification["status"] == "classified" + assert classification["candidate_count"] >= 900 + assert classification["unclassified_candidate_count"] == 0 + assert classification["priority_bucket_counts"]["priority_1"] >= 50 + assert classification["priority_bucket_counts"]["priority_2"] >= 50 + assert classification["priority_bucket_counts"]["priority_3"] >= 100 + + by_path = classification["by_path"] + assert by_path["references/real_case_studies/vedicka/career-success-poverty-prosperity.md"]["classification"] == "real_case_calibration" + assert by_path["references/open_source_sources/rishi-ai-mcp/.agents/skills/career-analysis/SKILL.md"]["classification"] == "open_source_reference" + assert by_path["references/open_source_sources/vedic-astro-skills/codex/skills/vedic-core/resources/qa_rules.md"]["classification"] == "runtime_reference_layer" + assert by_path["references/advanced-techniques.md"]["classification"] == "reference_candidate" + assert by_path["docs/research/local_drafts/2026-06/antigravity_round31_api_completion_top50_2026_06_26.md"]["classification"] == "quarantined_draft" + assert by_path["docs/research/ACTIVE_FRONTS.md"]["classification"] == "research_governance" + + assert by_path["references/open_source_sources/rishi-ai-mcp/.agents/skills/career-analysis/SKILL.md"]["priority"] == "priority_1" + assert by_path["references/real_case_studies/vedicka/career-success-poverty-prosperity.md"]["priority"] == "priority_1" + assert by_path["docs/research/local_drafts/2026-06/antigravity_round31_api_completion_top50_2026_06_26.md"]["promotion_status"] == "not_truth_source"