Answer questions over this repo's regulation-ready Markdown data layer. Use when an AI agent needs to answer questions about rules, notices, attachments, formulas, thresholds, implementation notices, or reporting obligations by reading `Knowledge_Base_MarkDown/ready_data/` first, narrowing scope by title or number, then grounding the final answer in structured sections or raw Markdown instead of freeform chatbot guessing.
Produce verifiable answers from this repo's ready_data artifacts with the smallest correct scope and the least unsupported inference.
catalog, locate, summary, formula, comparison, version, compliance, or analysis.doc_catalog.jsonl or manifest.json directly instead of retrieval.doc_catalog.jsonl and title_aliases.jsonl first, then keep the initial scope to that doc_id.doc_summaries.jsonl first and use summary_structured, focus_points, and headings to build the outline.sections_structured.jsonlformula_cards.jsonl to locate the formula and use the matching structured section to explain variable meaning, applicability, and nearby rule text.ready_data artifacts are insufficient or obviously noisy.Use the unified CLI from the repo root when answering as an agent:
python .\cross2.py plan --question "规则第2号主要内容是什么"
python .\cross2.py evidence --question "控制风险最低资本按照哪一项规则计量？"
python .\cross2.py search sections --query "最低资本由哪些部分组成？" --doc-id "rules/保险公司偿付能力监管规则第2号：最低资本.md"
python .\cross2.py search formulas --query "规则第2号最低资本公式" --doc-id "rules/保险公司偿付能力监管规则第2号：最低资本.md"
python .\cross2.py explain formula --formula-id "rules/保险公司偿付能力监管规则第2号：最低资本.md#formula-1"
python .\cross2.py trace relations --doc-id "rules/保险公司偿付能力监管规则第2号：最低资本.md"
python .\cross2.py answer --question "最低资本由哪些部分组成？" --mode verified
doc_catalog.jsonl: document identity, numbering, aliases, headings, and direct catalog answerstitle_aliases.jsonl: shorthand, rule number, attachment number, and notice-title lookupdoc_summaries.jsonl: overview and scope narrowingsections_structured.jsonl: article-level, requirement-level, threshold-level, table-level, and formula-adjacent evidenceformula_cards.jsonl: formula lookup only; do not rely on it alone for semantic explanationrelations_graph.json and related_doc_ids: expansion hints, not authoritative regulatory conclusionsrelated_doc_ids and relation edges as navigation hints; confirm the substantive claim in section text or raw Markdown before using it.MC, LA, Stress_t, and Spread_t.summary_short is noisy because of OCR notes, image placeholders, or publish-date boilerplate, use summary_structured, headings, and section text instead.