muse-agent-example/.claude/skills/access-database/scripts/seed_schemas.py

#!/usr/bin/env python3
"""A3：23 型结构本体 YAML → meta 表行（W1 种子演练，幂等可重跑）。

映射约定（详见 db/表映射.md 与本次收口记录）：
- muse_meta_schema         一型一行；schema_key=target_type；状态 启用→active / 待启用→inactive
- muse_meta_schema_version 每型 v1，active_flag=TRUE；field_contract_snapshot=完整 YAML（判据/说明/设计发现全在）
- muse_meta_field          规范化投影行：基础字段 sort_order 1–9，特有字段 11 起（段位即基础/特有约定）
- muse_meta_visibility_policy 版本级一行：ai_context=是否存在 AI 可见字段；
                           字段级细则进 policy_snapshot.fieldAiContext（主仓无字段级列——设计发现，待回填）
- 幂等：schema/version/policy upsert；field 行重跑=删旧插新（meta 种子行豁免软删约定，见表映射.md）
"""
import json
import pathlib
import sys

import psycopg
import yaml
from psycopg.types.json import Jsonb

DSN = "postgresql://root:f6710e2d0294eb1c10e26a805a64bc54@100.64.0.8:5433/muse-example"
SCHEMA_DIR = pathlib.Path(__file__).resolve().parents[4] / "meta" / "schemas"
TENANT, ACTOR = 1, "1"  # 实验写入约定：系统主账号

# 基础字段（所有型共有，README §使用规则；aiContext 为实验设计判断，可在 B4 优化环调整）：
# 名称/别名/摘要/标签全用途可见；来源(出处)生成不可见防抄袭腔；状态是系统轨道值不进上下文（授权过滤在查询层）
BASE_FIELDS = [
    ("名称", "text", True, None, True),
    ("别名", "array", False, None, True),
    ("一句话摘要", "text", True, None, True),
    ("标签", "array", False, None, True),
    ("来源", "text", False, None, ["planning", "detection", "extraction"]),
    ("状态", "enum", True, ["草稿", "已确认"], False),
]


def seed_one(conn, doc: dict, src_name: str):
    """单型入库：schema → version(v1) → fields → visibility_policy；返回审查摘要。"""
    key = doc["target_type"]
    status = "active" if doc.get("状态", "启用") == "启用" else "inactive"
    # 1) schema 行 upsert（uk: tenant_id+domain+scope+target_type+schema_key）
    row = conn.execute(
        """INSERT INTO muse_meta_schema (schema_key, domain, scope, target_type, display_name, status,
                                         creator, updater, tenant_id)
           VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s)
           ON CONFLICT (tenant_id, domain, scope, target_type, schema_key)
           DO UPDATE SET display_name=EXCLUDED.display_name, status=EXCLUDED.status, updater=EXCLUDED.updater
           RETURNING id""",
        (key, doc["domain"], doc["scope"], key, doc.get("中文名", key), status,
         ACTOR, ACTOR, TENANT)).fetchone()
    schema_id = row[0]

    # 2) version v1 upsert（uk: tenant_id+schema_key+version_no）；快照=完整 YAML
    row = conn.execute(
        """INSERT INTO muse_meta_schema_version (schema_id, schema_key, version_no, status, active_flag,
                field_contract_snapshot, change_note, published_by, activated_by, creator, updater, tenant_id)
           VALUES (%s,%s,1,'published',TRUE,%s,%s,%s,%s,%s,%s,%s)
           ON CONFLICT (tenant_id, schema_key, version_no)
           DO UPDATE SET field_contract_snapshot=EXCLUDED.field_contract_snapshot,
                         change_note=EXCLUDED.change_note, updater=EXCLUDED.updater
           RETURNING id""",
        (schema_id, key, Jsonb(doc), f"A3 种子入库(源:{src_name})", ACTOR, ACTOR, ACTOR, ACTOR, TENANT)).fetchone()
    version_id = row[0]
    conn.execute("UPDATE muse_meta_schema SET active_version_id=%s WHERE id=%s", (version_id, schema_id))

    # 3) field 行：删旧插新（幂等重建）
    conn.execute("DELETE FROM muse_meta_field WHERE tenant_id=%s AND schema_version_id=%s", (TENANT, version_id))
    ai_map = {}
    n_special = 0
    for i, (fk, ftype, req, enum, ai) in enumerate(BASE_FIELDS, start=1):
        conn.execute(
            """INSERT INTO muse_meta_field (schema_version_id, field_key, display_name, field_type,
                   is_required, enum_values, sort_order, creator, updater, tenant_id)
               VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s)""",
            (version_id, fk, fk, ftype, req, Jsonb(enum) if enum else None, i, ACTOR, ACTOR, TENANT))
        ai_map[fk] = ai
    for i, f in enumerate(doc.get("特有字段") or [], start=11):
        conn.execute(
            """INSERT INTO muse_meta_field (schema_version_id, field_key, display_name, field_type,
                   is_required, sort_order, creator, updater, tenant_id)
               VALUES (%s,%s,%s,'text',FALSE,%s,%s,%s,%s)""",
            (version_id, f["key"], f["key"], i, ACTOR, ACTOR, TENANT))
        ai_map[f["key"]] = f.get("aiContext", True)
        n_special += 1

    # 4) visibility_policy：版本级一行 upsert（无 uk，先查后写）；字段级细则进 policy_snapshot
    any_ai = any(v is not False for v in ai_map.values())
    snap = Jsonb({"fieldAiContext": ai_map,
                  "用途枚举": ["planning", "generation", "detection", "extraction"],
                  "语义": "true=任何用途可入AI上下文;false=一律不入;[用途]=仅列出的用途可入"})
    exist = conn.execute(
        "SELECT id FROM muse_meta_visibility_policy WHERE tenant_id=%s AND schema_version_id=%s",
        (TENANT, version_id)).fetchone()
    if exist:
        conn.execute("UPDATE muse_meta_visibility_policy SET ai_context=%s, policy_snapshot=%s, updater=%s WHERE id=%s",
                     (any_ai, snap, ACTOR, exist[0]))
    else:
        conn.execute(
            """INSERT INTO muse_meta_visibility_policy (schema_version_id, ui_visible, ai_context,
                   user_editable, user_searchable, exportable, policy_snapshot, creator, updater, tenant_id)
               VALUES (%s,TRUE,%s,TRUE,FALSE,FALSE,%s,%s,%s,%s)""",
            (version_id, any_ai, snap, ACTOR, ACTOR, TENANT))

    full_open = sum(1 for v in ai_map.values() if v is True)
    scoped = sum(1 for v in ai_map.values() if isinstance(v, list))
    closed = sum(1 for v in ai_map.values() if v is False)
    return (key, doc.get("中文名", ""), doc["domain"], doc["scope"], status,
            len(BASE_FIELDS) + n_special, full_open, scoped, closed)


def main():
    files = sorted(p for p in SCHEMA_DIR.glob("*.yaml"))
    if not files:
        sys.exit(f"未找到 schema YAML: {SCHEMA_DIR}")
    rows = []
    with psycopg.connect(DSN) as conn:
        for p in files:
            doc = yaml.safe_load(p.read_text())
            rows.append(seed_one(conn, doc, p.name))
        conn.commit()
    # 审查面：逐型卡片行
    print(f"{'型':<20}{'中文名':<10}{'domain':<11}{'scope':<10}{'状态':<9}{'字段':<5}{'AI全开':<7}{'限用途':<7}{'关闭'}")
    for r in rows:
        print(f"{r[0]:<20}{r[1]:<10}{r[2]:<11}{r[3]:<10}{r[4]:<9}{r[5]:<5}{r[6]:<7}{r[7]:<7}{r[8]}")
    print(f"\n共 {len(rows)} 型入库（基础字段 {len(BASE_FIELDS)} + 各型特有；快照含判据/说明/设计发现全文）")


if __name__ == "__main__":
    main()