初始化提交

backend/app/modules/imports/extract.py

@@ -0,0 +1,143 @@
+"""解析模型返回为候选草稿（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，
+并归一化为候选草稿。evidence 序号由 application 映射为 ``import_items.id``。
+"""
+
+import json
+import re
+from dataclasses import dataclass, field
+from typing import Any
+
+
+@dataclass
+class ProfileCandidateDraft:
+    field_name: str
+    value: str
+    confidence: float
+    reason: str
+    evidence: list[int] = field(default_factory=list)
+
+
+@dataclass
+class MemoryCandidateDraft:
+    type: str
+    title: str
+    content: str
+    importance: float
+    sensitivity: str
+    reason: str
+    evidence: list[int] = field(default_factory=list)
+
+
+@dataclass
+class PersonalityProposalDraft:
+    field_name: str
+    change: str
+    reason: str
+    evidence: list[int] = field(default_factory=list)
+
+
+def parse_profile_candidates(raw: str) -> list[ProfileCandidateDraft]:
+    drafts: list[ProfileCandidateDraft] = []
+    for item in _extract_json_array(raw):
+        if not isinstance(item, dict):
+            continue
+        field_name = str(item.get("field", "")).strip()
+        value = str(item.get("value", "")).strip()
+        if not field_name or not value:
+            continue
+        drafts.append(
+            ProfileCandidateDraft(
+                field_name=field_name,
+                value=value,
+                confidence=_clamp_float(item.get("confidence"), 0.5),
+                reason=str(item.get("reason", "")).strip(),
+                evidence=_int_list(item.get("evidence")),
+            )
+        )
+    return drafts
+
+
+def parse_memory_candidates(raw: str) -> list[MemoryCandidateDraft]:
+    drafts: list[MemoryCandidateDraft] = []
+    for item in _extract_json_array(raw):
+        if not isinstance(item, dict):
+            continue
+        title = str(item.get("title", "")).strip()
+        content = str(item.get("content", "")).strip()
+        if not title or not content:
+            continue
+        drafts.append(
+            MemoryCandidateDraft(
+                type=str(item.get("type", "other")).strip() or "other",
+                title=title,
+                content=content,
+                importance=_clamp_float(item.get("importance"), 0.5),
+                sensitivity=_sensitivity(item.get("sensitivity")),
+                reason=str(item.get("reason", "")).strip(),
+                evidence=_int_list(item.get("evidence")),
+            )
+        )
+    return drafts
+
+
+def parse_personality_proposals(raw: str) -> list[PersonalityProposalDraft]:
+    drafts: list[PersonalityProposalDraft] = []
+    for item in _extract_json_array(raw):
+        if not isinstance(item, dict):
+            continue
+        field_name = str(item.get("field", "")).strip()
+        change = str(item.get("change", "")).strip()
+        if not field_name or not change:
+            continue
+        drafts.append(
+            PersonalityProposalDraft(
+                field_name=field_name,
+                change=change,
+                reason=str(item.get("reason", "")).strip(),
+                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 []
+
+
+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
+
+
+def _sensitivity(value: Any) -> str:
+    text = str(value or "").strip().lower()
+    return text if text in {"low", "medium", "high"} else "low"