初始化提交

backend/app/modules/big_cycle/extract.py

@@ -0,0 +1,114 @@
+"""解析模型返回为演化候选草稿（纯函数，无 IO）。
+
+模型可能返回纯 JSON、markdown 代码块或夹杂解释文本；本模块容错提取首个 JSON
+数组，并归一化为演化候选草稿。``evidence`` 序号由 application 映射为记忆 ID。
+
+缺省字段取**安全悲观默认**：``level=L4``（必须人工确认）、``auto_apply_allowed=False``，
+避免模型漏字段时误自动应用高风险演化。
+"""
+
+import json
+import re
+from dataclasses import dataclass, field
+from typing import Any
+
+
+@dataclass
+class EvolutionCandidateDraft:
+    type: str
+    title: str
+    proposal: str
+    level: str = "L4"
+    scope: str = "character_scope"
+    source: str = "big_cycle"
+    delta: dict[str, Any] = field(default_factory=dict)
+    confidence: float = 0.5
+    benefit_score: float = 0.5
+    risk_score: float = 0.5
+    drift_score: float = 0.0
+    auto_apply_allowed: bool = False
+    evidence: list[int] = field(default_factory=list)
+
+
+def parse_evolution_candidates(raw: str) -> list[EvolutionCandidateDraft]:
+    drafts: list[EvolutionCandidateDraft] = []
+    for item in _extract_json_array(raw):
+        if not isinstance(item, dict):
+            continue
+        type_ = str(item.get("type", "")).strip()
+        title = str(item.get("title", "")).strip()
+        proposal = str(item.get("proposal", "")).strip()
+        if not type_ or not title or not proposal:
+            continue
+        drafts.append(
+            EvolutionCandidateDraft(
+                type=type_,
+                title=title,
+                proposal=proposal,
+                level=_level(item.get("level")),
+                scope=_scope(item.get("scope")),
+                source=str(item.get("source", "big_cycle")).strip() or "big_cycle",
+                delta=_dict(item.get("delta")),
+                confidence=_clamp_float(item.get("confidence"), 0.5),
+                benefit_score=_clamp_float(item.get("benefit_score"), 0.5),
+                risk_score=_clamp_float(item.get("risk_score"), 0.5),
+                drift_score=_clamp_float(item.get("drift_score"), 0.0),
+                auto_apply_allowed=bool(item.get("auto_apply_allowed", False)),
+                evidence=_int_list(item.get("evidence")),
+            )
+        )
+    return drafts
+
+
+def _extract_json_array(raw: str) -> list[Any]:
+    text = raw.strip()
+    matched = re.search(r"```(?:json)?\s*(\[.*?\])\s*```", text, re.DOTALL)
+    if matched:
+        text = matched.group(1)
+    else:
+        start, end = text.find("["), text.rfind("]")
+        if start != -1 and end > start:
+            text = text[start : end + 1]
+    try:
+        data = json.loads(text)
+    except json.JSONDecodeError:
+        return []
+    return data if isinstance(data, list) else []
+
+
+_VALID_LEVELS = {"L0", "L1", "L2", "L3", "L4"}
+_VALID_SCOPES = {"character_scope", "user_scope", "task_scope", "global_scope"}
+
+
+def _level(value: Any) -> str:
+    text = str(value or "").strip().upper()
+    return text if text in _VALID_LEVELS else "L4"
+
+
+def _scope(value: Any) -> str:
+    text = str(value or "").strip()
+    return text if text in _VALID_SCOPES else "character_scope"
+
+
+def _dict(value: Any) -> dict[str, Any]:
+    return value if isinstance(value, dict) else {}
+
+
+def _clamp_float(value: Any, default: float) -> float:
+    try:
+        num = float(value)
+    except (TypeError, ValueError):
+        return default
+    return max(0.0, min(1.0, num))
+
+
+def _int_list(value: Any) -> list[int]:
+    if not isinstance(value, list):
+        return []
+    result: list[int] = []
+    for item in value:
+        try:
+            result.append(int(item))
+        except (TypeError, ValueError):
+            continue
+    return result