76d38f2dc7
- 模型锁定完整 ID 等值：role_policy 废弃子串匹配，前置校验+事后 MODEL_POLICY_VIOLATION 熔断+回执 modelMatch，治理链角色放行 BUDGET_CHAIN - 工具白名单只读机械强制：任务包 allowlist ⊆ 只读注册表，TOOL_NOT_READONLY - 本地向量检索 truthful 化：aiContext 裁剪、指针行跳过、资格失败关闭 （bindingStatus/productionRetrievalEligible 不再伪造 active） - 角色提示词 name 回归英文系统 ID；planner 补 fine_outline 绑定； writer 数据契约对齐 writer-candidate-body-v1 - 测试账本隔离：sqlite_path 三层透传（bridge/two_phase），9 处派发测试 改用临时库；清除 muse.db 测试残留 5 runs+26 events+reviews 5/6（有备份） - test-inventory 补 3 条登记并机械重算 summary；评测场景补 output_contract/fail_closed 与 stability 两类；死代码清理 （project_paths.py、offline_only 死参数、harness/harness 残骸） - 新增 metaphysical-diff-review 技能（红线 4.2 载体）并登记，共 59 技能
480 lines
16 KiB
Python
480 lines
16 KiB
Python
"""本地 SQLite（data/muse.db）：WAL、版本化迁移、run/event/review/card。"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import hashlib
|
||
import json
|
||
import math
|
||
import os
|
||
import sqlite3
|
||
import struct
|
||
from datetime import datetime, timezone
|
||
from pathlib import Path
|
||
from typing import Any, Iterable, Mapping, Sequence
|
||
|
||
PROJECT_ROOT = next(
|
||
parent
|
||
for parent in (Path(__file__).resolve().parent, *Path(__file__).resolve().parents)
|
||
if (parent / "AGENTS.md").is_file() and (parent / ".git").exists()
|
||
)
|
||
MIGRATIONS_DIR = PROJECT_ROOT / "data" / "migrations"
|
||
_DEFAULT_DB = PROJECT_ROOT / "data" / "muse.db"
|
||
|
||
|
||
def default_db_path() -> Path:
|
||
override = os.environ.get("MUSE_DB")
|
||
return Path(override).expanduser() if override else _DEFAULT_DB
|
||
|
||
|
||
def _now() -> str:
|
||
return datetime.now(timezone.utc).replace(microsecond=0).isoformat()
|
||
|
||
|
||
def pack_embedding(vector: Sequence[float]) -> bytes:
|
||
return struct.pack(f"<{len(vector)}f", *[float(item) for item in vector])
|
||
|
||
|
||
def unpack_embedding(blob: bytes) -> list[float]:
|
||
if not blob:
|
||
return []
|
||
count = len(blob) // 4
|
||
return list(struct.unpack(f"<{count}f", blob))
|
||
|
||
|
||
def cosine(left: Sequence[float], right: Sequence[float]) -> float:
|
||
if not left or not right or len(left) != len(right):
|
||
return 0.0
|
||
dot = sum(a * b for a, b in zip(left, right))
|
||
norm_l = math.sqrt(sum(a * a for a in left))
|
||
norm_r = math.sqrt(sum(b * b for b in right))
|
||
if norm_l == 0.0 or norm_r == 0.0:
|
||
return 0.0
|
||
return dot / (norm_l * norm_r)
|
||
|
||
|
||
def skill_set_hash(root: Path | None = None) -> str:
|
||
base = root or PROJECT_ROOT
|
||
digest = hashlib.sha256()
|
||
skills = base / ".agent" / "skills"
|
||
if skills.is_dir():
|
||
for path in sorted(skills.rglob("SKILL.md")):
|
||
digest.update(path.as_posix().encode("utf-8"))
|
||
digest.update(path.read_bytes())
|
||
return "sha256:" + digest.hexdigest()
|
||
|
||
|
||
def baseline_hash(path: Path | None = None) -> str | None:
|
||
db_path = path or default_db_path()
|
||
if not db_path.is_file():
|
||
return None
|
||
digest = hashlib.sha256()
|
||
with db_path.open("rb") as handle:
|
||
for chunk in iter(lambda: handle.read(1024 * 1024), b""):
|
||
digest.update(chunk)
|
||
return "sha256:" + digest.hexdigest()
|
||
|
||
|
||
def connect(path: str | Path | None = None) -> sqlite3.Connection:
|
||
db_path = Path(path) if path is not None else default_db_path()
|
||
db_path.parent.mkdir(parents=True, exist_ok=True)
|
||
conn = sqlite3.connect(str(db_path))
|
||
conn.row_factory = sqlite3.Row
|
||
conn.execute("PRAGMA foreign_keys=ON")
|
||
mode = conn.execute("PRAGMA journal_mode=WAL").fetchone()[0]
|
||
if str(mode).lower() != "wal":
|
||
conn.close()
|
||
raise RuntimeError(f"无法开启 WAL：journal_mode={mode!r}")
|
||
_apply_migrations(conn)
|
||
return conn
|
||
|
||
|
||
def _apply_migrations(conn: sqlite3.Connection) -> None:
|
||
conn.execute(
|
||
"""CREATE TABLE IF NOT EXISTS schema_migrations (
|
||
version TEXT PRIMARY KEY,
|
||
applied_at TEXT NOT NULL
|
||
)"""
|
||
)
|
||
applied = {
|
||
row["version"]
|
||
for row in conn.execute("SELECT version FROM schema_migrations")
|
||
}
|
||
for script in sorted(MIGRATIONS_DIR.glob("*.sql")):
|
||
version = script.stem
|
||
if version in applied:
|
||
continue
|
||
conn.executescript(script.read_text(encoding="utf-8"))
|
||
conn.execute(
|
||
"INSERT INTO schema_migrations(version, applied_at) VALUES (?, ?)",
|
||
(version, _now()),
|
||
)
|
||
conn.commit()
|
||
|
||
|
||
def upsert_card(
|
||
*,
|
||
card_id: str,
|
||
kind: str,
|
||
title: str | None,
|
||
payload: Mapping[str, Any],
|
||
embedding: Sequence[float] | None = None,
|
||
work_id: int | None = None,
|
||
source_path: str | None = None,
|
||
content_hash: str | None = None,
|
||
path: str | Path | None = None,
|
||
) -> None:
|
||
blob = pack_embedding(embedding) if embedding else None
|
||
with connect(path) as conn:
|
||
conn.execute(
|
||
"""INSERT INTO cards(
|
||
id, kind, title, payload_json, embedding, work_id,
|
||
source_path, content_hash, created_at
|
||
) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
|
||
ON CONFLICT(id) DO UPDATE SET
|
||
kind=excluded.kind,
|
||
title=excluded.title,
|
||
payload_json=excluded.payload_json,
|
||
embedding=excluded.embedding,
|
||
work_id=excluded.work_id,
|
||
source_path=excluded.source_path,
|
||
content_hash=excluded.content_hash
|
||
""",
|
||
(
|
||
card_id,
|
||
kind,
|
||
title,
|
||
json.dumps(payload, ensure_ascii=False),
|
||
blob,
|
||
work_id,
|
||
source_path,
|
||
content_hash,
|
||
_now(),
|
||
),
|
||
)
|
||
conn.commit()
|
||
|
||
|
||
def _ai_context_visible(ai_rule: Any, purpose: str) -> bool:
|
||
"""aiContext 判定，与 search-knowledge PG 面同语义：
|
||
true 全用途可见；false 不可见；[用途] 仅列出的可见；无规则默认可见。"""
|
||
|
||
if ai_rule is None:
|
||
return True
|
||
if isinstance(ai_rule, bool):
|
||
return ai_rule
|
||
return purpose in ai_rule
|
||
|
||
|
||
def search_card_vectors(
|
||
intent_vector: Sequence[float],
|
||
*,
|
||
kind: str | None = None,
|
||
work_id: int | None = None,
|
||
top: int = 5,
|
||
path: str | Path | None = None,
|
||
purpose: str = "generation",
|
||
ai_rules: Mapping[str, Mapping[str, Any]] | None = None,
|
||
) -> list[dict[str, Any]]:
|
||
"""本地 SQLite 向量召回（PG 检索面的降级镜像）。
|
||
|
||
与 PG 面同口径应用 aiContext 字段裁剪；本地卡没有绑定/状态面，
|
||
bindingStatus 与 productionRetrievalEligible 一律失败关闭（None/False），
|
||
不得伪造 active 资格——生产消费方（如 ProductionCardIndexRepository）
|
||
会因此拒绝本地镜像结果，而不是静默放行。
|
||
"""
|
||
|
||
if purpose not in {"generation", "planning", "detection", "extraction"}:
|
||
raise ValueError("purpose 非法")
|
||
rules_by_type: Mapping[str, Mapping[str, Any]] = ai_rules or {}
|
||
sql = "SELECT id, kind, title, payload_json, embedding, work_id, source_path, content_hash FROM cards WHERE embedding IS NOT NULL"
|
||
args: list[Any] = []
|
||
if kind:
|
||
sql += " AND kind = ?"
|
||
args.append(kind)
|
||
if work_id is not None:
|
||
sql += " AND work_id = ?"
|
||
args.append(work_id)
|
||
with connect(path) as conn:
|
||
rows = list(conn.execute(sql, args))
|
||
scored: list[tuple[float, sqlite3.Row]] = []
|
||
for row in rows:
|
||
score = cosine(intent_vector, unpack_embedding(row["embedding"]))
|
||
scored.append((score, row))
|
||
scored.sort(key=lambda item: item[0], reverse=True)
|
||
results: list[dict[str, Any]] = []
|
||
for score, row in scored[:top]:
|
||
payload = json.loads(row["payload_json"] or "{}")
|
||
fields = payload.get("字段") if isinstance(payload.get("字段"), dict) else {}
|
||
# 无内容指针行（如 PG 迁移的 embedding 指针）不产出检索结果：
|
||
# 它们没有名称/字段/摘要，召回它们只会产出空壳卡片。
|
||
if not (payload.get("名称") or payload.get("一句话摘要") or fields):
|
||
continue
|
||
card_type = payload.get("型") or row["kind"]
|
||
type_rules = rules_by_type.get(card_type, {})
|
||
visible_fields = {
|
||
key: item
|
||
for key, item in fields.items()
|
||
if _ai_context_visible(type_rules.get(key), purpose)
|
||
}
|
||
results.append(
|
||
{
|
||
"cardId": row["id"],
|
||
"type": card_type,
|
||
"name": payload.get("名称") or row["title"],
|
||
"score": float(score),
|
||
"summary": payload.get("一句话摘要"),
|
||
"visibleFields": visible_fields,
|
||
"omittedFields": sorted(set(fields) - set(visible_fields)),
|
||
"sourceId": f"sqlite-card:{row['id']}",
|
||
"sourceVersion": f"hash:{row['content_hash'] or 'none'}",
|
||
"sourceOffset": 0,
|
||
"sourceRefs": payload.get("sourceRefs") if isinstance(payload.get("sourceRefs"), list) else [],
|
||
"milestones": [],
|
||
"sourceKind": "local_card",
|
||
"sourceStatus": None,
|
||
"bindingStatus": None,
|
||
"retrievalScope": "work" if work_id is not None else "admin",
|
||
"productionRetrievalEligible": False,
|
||
}
|
||
)
|
||
return results
|
||
|
||
|
||
class FanoutSink:
|
||
"""把同一条事件转给多个 sink；runtime 不认识本类型。"""
|
||
|
||
def __init__(self, *sinks: Any) -> None:
|
||
self._sinks = sinks
|
||
|
||
def emit(self, event_type: str, **kwargs: Any) -> Any:
|
||
result = None
|
||
for sink in self._sinks:
|
||
result = sink.emit(event_type, **kwargs)
|
||
return result
|
||
|
||
|
||
class SqliteRecorder:
|
||
"""缓冲会话事件，结束时一次写入 run + events。"""
|
||
|
||
def __init__(
|
||
self,
|
||
run_id: str,
|
||
*,
|
||
kind: str,
|
||
input_obj: Mapping[str, Any],
|
||
skill_set_hash_value: str,
|
||
role: str | None = None,
|
||
path: str | Path | None = None,
|
||
) -> None:
|
||
self.run_id = run_id
|
||
self.kind = kind
|
||
self.input_obj = dict(input_obj)
|
||
self.skill_set_hash_value = skill_set_hash_value
|
||
self.role = role
|
||
self.path = path
|
||
self.events: list[dict[str, Any]] = []
|
||
self._committed = False
|
||
|
||
def emit(self, event_type: str, **kwargs: Any) -> int:
|
||
seq = len(self.events) + 1
|
||
self.events.append(
|
||
{
|
||
"seq": seq,
|
||
"kind": event_type,
|
||
"role": self.role,
|
||
"tool_name": kwargs.get("tool_name"),
|
||
"payload": {key: value for key, value in kwargs.items() if key != "tool_name"},
|
||
"created_at": _now(),
|
||
}
|
||
)
|
||
return seq
|
||
|
||
def commit(self, output_text: str, meta: Mapping[str, Any] | None = None) -> None:
|
||
if self._committed:
|
||
return
|
||
self._committed = True
|
||
db_path = Path(self.path) if self.path is not None else default_db_path()
|
||
prior = baseline_hash(db_path)
|
||
payload = json.dumps(self.input_obj, ensure_ascii=False, sort_keys=True, default=str)
|
||
meta_json = json.dumps(dict(meta or {}), ensure_ascii=False, sort_keys=True, default=str)
|
||
with connect(db_path) as conn:
|
||
try:
|
||
conn.execute(
|
||
"""INSERT INTO runs(
|
||
id, created_at, kind, input_json, output_text,
|
||
meta_json, skill_set_hash, baseline_hash
|
||
) VALUES (?, ?, ?, ?, ?, ?, ?, ?)""",
|
||
(
|
||
self.run_id,
|
||
_now(),
|
||
self.kind,
|
||
payload,
|
||
output_text or "",
|
||
meta_json,
|
||
self.skill_set_hash_value,
|
||
prior,
|
||
),
|
||
)
|
||
except sqlite3.IntegrityError:
|
||
return
|
||
for event in self.events:
|
||
conn.execute(
|
||
"""INSERT INTO events(
|
||
run_id, seq, kind, role, tool_name, payload_json, created_at
|
||
) VALUES (?, ?, ?, ?, ?, ?, ?)""",
|
||
(
|
||
self.run_id,
|
||
event["seq"],
|
||
event["kind"],
|
||
event["role"],
|
||
event["tool_name"],
|
||
json.dumps(event["payload"], ensure_ascii=False, default=str),
|
||
event["created_at"],
|
||
),
|
||
)
|
||
conn.commit()
|
||
|
||
|
||
def add_review(
|
||
*,
|
||
run_id: str,
|
||
target: str,
|
||
action: str,
|
||
reviewer: str,
|
||
reason: str | None = None,
|
||
before_text: str | None = None,
|
||
after_text: str | None = None,
|
||
kind: str = "prose",
|
||
path: str | Path | None = None,
|
||
) -> int:
|
||
if not reviewer or not str(reviewer).strip():
|
||
raise ValueError("reviewer 必填")
|
||
if action == "revise" and (before_text is None or after_text is None):
|
||
raise ValueError("revise 必须提供 before_text 与 after_text")
|
||
with connect(path) as conn:
|
||
cursor = conn.execute(
|
||
"""INSERT INTO reviews(run_id, target, action, reviewer, reason, created_at)
|
||
VALUES (?, ?, ?, ?, ?, ?)""",
|
||
(run_id, target, action, reviewer, reason, _now()),
|
||
)
|
||
review_id = int(cursor.lastrowid)
|
||
if action == "revise":
|
||
conn.execute(
|
||
"""INSERT INTO revisions(review_id, kind, before_text, after_text, created_at)
|
||
VALUES (?, ?, ?, ?, ?)""",
|
||
(review_id, kind, before_text or "", after_text or "", _now()),
|
||
)
|
||
conn.commit()
|
||
return review_id
|
||
|
||
|
||
def register_lesson(
|
||
*,
|
||
lesson_id: str,
|
||
source_run_ids: Sequence[str],
|
||
source_review_ids: Sequence[str],
|
||
kind: str,
|
||
title: str,
|
||
content: str,
|
||
target_ref: str,
|
||
lessons_dir: str | Path | None = None,
|
||
path: str | Path | None = None,
|
||
) -> Path:
|
||
if not source_run_ids or not source_review_ids:
|
||
raise ValueError("lesson 必须绑定 source_run_ids 与 source_review_ids")
|
||
root = PROJECT_ROOT / "lessons" / "pending"
|
||
dest_dir = Path(lessons_dir) if lessons_dir is not None else root
|
||
dest_dir.mkdir(parents=True, exist_ok=True)
|
||
content_path = dest_dir / f"{lesson_id}.md"
|
||
content_path.write_text(content, encoding="utf-8")
|
||
with connect(path) as conn:
|
||
conn.execute(
|
||
"""INSERT INTO lessons(
|
||
id, source_run_ids, source_review_ids, kind, title,
|
||
content_path, status, target_ref
|
||
) VALUES (?, ?, ?, ?, ?, ?, 'proposed', ?)""",
|
||
(
|
||
lesson_id,
|
||
json.dumps(list(source_run_ids)),
|
||
json.dumps(list(source_review_ids)),
|
||
kind,
|
||
title,
|
||
str(content_path),
|
||
target_ref,
|
||
),
|
||
)
|
||
conn.commit()
|
||
return content_path
|
||
|
||
|
||
def approve_and_merge_lesson(
|
||
lesson_id: str,
|
||
*,
|
||
decided_by: str,
|
||
rationale: str,
|
||
references_dir: str | Path,
|
||
path: str | Path | None = None,
|
||
) -> Path:
|
||
with connect(path) as conn:
|
||
row = conn.execute("SELECT * FROM lessons WHERE id=?", (lesson_id,)).fetchone()
|
||
if row is None:
|
||
raise ValueError(f"lesson 不存在: {lesson_id}")
|
||
source = Path(row["content_path"])
|
||
text = source.read_text(encoding="utf-8")
|
||
dest_dir = Path(references_dir)
|
||
dest_dir.mkdir(parents=True, exist_ok=True)
|
||
dest = dest_dir / f"lesson-{lesson_id}.md"
|
||
dest.write_text(text, encoding="utf-8")
|
||
conn.execute(
|
||
"""UPDATE lessons
|
||
SET status='promoted', decided_by=?, decided_at=?, rationale=?
|
||
WHERE id=?""",
|
||
(decided_by, _now(), rationale, lesson_id),
|
||
)
|
||
conn.commit()
|
||
add_review(
|
||
run_id=json.loads(row["source_run_ids"])[0],
|
||
target="lesson",
|
||
action="adopt",
|
||
reviewer=decided_by,
|
||
reason=rationale,
|
||
path=path,
|
||
)
|
||
return dest
|
||
|
||
|
||
def get_run(run_id: str, path: str | Path | None = None) -> dict[str, Any] | None:
|
||
with connect(path) as conn:
|
||
row = conn.execute("SELECT * FROM runs WHERE id = ?", (run_id,)).fetchone()
|
||
if row is None:
|
||
return None
|
||
return dict(row)
|
||
|
||
|
||
def list_events(run_id: str, path: str | Path | None = None) -> list[dict[str, Any]]:
|
||
with connect(path) as conn:
|
||
rows = conn.execute(
|
||
"SELECT * FROM events WHERE run_id = ? ORDER BY seq ASC", (run_id,)
|
||
).fetchall()
|
||
return [dict(row) for row in rows]
|
||
|
||
|
||
__all__ = [
|
||
"FanoutSink",
|
||
"SqliteRecorder",
|
||
"add_review",
|
||
"baseline_hash",
|
||
"connect",
|
||
"cosine",
|
||
"default_db_path",
|
||
"approve_and_merge_lesson",
|
||
"get_run",
|
||
"list_events",
|
||
"register_lesson",
|
||
"pack_embedding",
|
||
"search_card_vectors",
|
||
"skill_set_hash",
|
||
"unpack_embedding",
|
||
"upsert_card",
|
||
]
|