209 lines
7.8 KiB
Python
209 lines
7.8 KiB
Python
# core/requirement_analyzer.py - 需求分解 & 函数签名生成
|
||
import json
|
||
from typing import List, Optional, Callable
|
||
|
||
import config
|
||
from core.llm_client import LLMClient
|
||
from database.models import FunctionalRequirement
|
||
|
||
|
||
class RequirementAnalyzer:
|
||
"""负责需求分解、模块分类、函数签名生成"""
|
||
|
||
def __init__(self, llm: LLMClient):
|
||
self.llm = llm
|
||
|
||
# ══════════════════════════════════════════════════
|
||
# 需求分解
|
||
# ══════════════════════════════════════════════════
|
||
|
||
def decompose(
|
||
self,
|
||
raw_requirement: str,
|
||
project_id: int,
|
||
raw_req_id: int,
|
||
knowledge: str = "",
|
||
) -> List[FunctionalRequirement]:
|
||
"""
|
||
将原始需求文本分解为功能需求列表（含模块分类）。
|
||
|
||
Args:
|
||
raw_requirement: 原始需求文本
|
||
project_id: 所属项目 ID
|
||
raw_req_id: 原始需求记录 ID
|
||
knowledge: 知识库文本（可选）
|
||
|
||
Returns:
|
||
FunctionalRequirement 对象列表（未持久化，id=None）
|
||
"""
|
||
knowledge_section = (
|
||
f"【参考知识库】\n{knowledge}\n" if knowledge else ""
|
||
)
|
||
prompt = config.DECOMPOSE_PROMPT_TEMPLATE.format(
|
||
raw_requirement = raw_requirement,
|
||
knowledge_section = knowledge_section,
|
||
)
|
||
|
||
try:
|
||
items = self.llm.chat_json(prompt)
|
||
if not isinstance(items, list):
|
||
raise ValueError("LLM 返回结果不是数组")
|
||
except Exception as e:
|
||
raise RuntimeError(f"需求分解失败: {e}")
|
||
|
||
reqs = []
|
||
for i, item in enumerate(items, 1):
|
||
req = FunctionalRequirement(
|
||
project_id = project_id,
|
||
raw_req_id = raw_req_id,
|
||
index_no = i,
|
||
title = item.get("title", f"功能{i}"),
|
||
description = item.get("description", ""),
|
||
function_name = item.get("function_name", f"function_{i}"),
|
||
priority = item.get("priority", "medium"),
|
||
module = item.get("module", config.DEFAULT_MODULE),
|
||
status = "pending",
|
||
is_custom = False,
|
||
)
|
||
reqs.append(req)
|
||
return reqs
|
||
|
||
# ══════════════════════════════════════════════════
|
||
# 模块分类（独立步骤，可对已有需求列表重新分类）
|
||
# ══════════════════════════════════════════════════
|
||
|
||
def classify_modules(
|
||
self,
|
||
func_reqs: List[FunctionalRequirement],
|
||
knowledge: str = "",
|
||
) -> List[dict]:
|
||
"""
|
||
对功能需求列表进行模块分类，返回 {function_name: module} 映射列表。
|
||
|
||
Args:
|
||
func_reqs: 功能需求列表
|
||
knowledge: 知识库文本（可选）
|
||
|
||
Returns:
|
||
[{"function_name": "...", "module": "..."}, ...]
|
||
"""
|
||
req_list = [
|
||
{
|
||
"index_no": r.index_no,
|
||
"title": r.title,
|
||
"description": r.description,
|
||
"function_name": r.function_name,
|
||
}
|
||
for r in func_reqs
|
||
]
|
||
knowledge_section = f"【参考知识库】\n{knowledge}\n" if knowledge else ""
|
||
prompt = config.MODULE_CLASSIFY_PROMPT_TEMPLATE.format(
|
||
requirements_json = json.dumps(req_list, ensure_ascii=False, indent=2),
|
||
knowledge_section = knowledge_section,
|
||
)
|
||
try:
|
||
result = self.llm.chat_json(prompt)
|
||
if not isinstance(result, list):
|
||
raise ValueError("LLM 返回结果不是数组")
|
||
return result
|
||
except Exception as e:
|
||
raise RuntimeError(f"模块分类失败: {e}")
|
||
|
||
# ══════════════════════════════════════════════════
|
||
# 函数签名生成
|
||
# ══════════════════════════════════════════════════
|
||
|
||
def build_function_signature(
|
||
self,
|
||
func_req: FunctionalRequirement,
|
||
requirement_id: str = "",
|
||
knowledge: str = "",
|
||
) -> dict:
|
||
"""
|
||
为单个功能需求生成函数签名 dict。
|
||
|
||
Args:
|
||
func_req: 功能需求对象
|
||
requirement_id: 需求编号字符串（如 "REQ.01"）
|
||
knowledge: 知识库文本
|
||
|
||
Returns:
|
||
函数签名 dict
|
||
|
||
Raises:
|
||
RuntimeError: LLM 调用或解析失败
|
||
"""
|
||
knowledge_section = f"【参考知识库】\n{knowledge}\n" if knowledge else ""
|
||
prompt = config.FUNC_SIGNATURE_PROMPT_TEMPLATE.format(
|
||
requirement_id = requirement_id or f"REQ.{func_req.index_no:02d}",
|
||
title = func_req.title,
|
||
description = func_req.description,
|
||
function_name = func_req.function_name,
|
||
module = func_req.module or config.DEFAULT_MODULE,
|
||
knowledge_section = knowledge_section,
|
||
)
|
||
try:
|
||
sig = self.llm.chat_json(prompt)
|
||
if not isinstance(sig, dict):
|
||
raise ValueError("LLM 返回结果不是 dict")
|
||
# 确保 module 字段存在
|
||
if "module" not in sig:
|
||
sig["module"] = func_req.module or config.DEFAULT_MODULE
|
||
return sig
|
||
except Exception as e:
|
||
raise RuntimeError(f"签名生成失败 [{func_req.function_name}]: {e}")
|
||
|
||
def build_function_signatures_batch(
|
||
self,
|
||
func_reqs: List[FunctionalRequirement],
|
||
knowledge: str = "",
|
||
on_progress: Optional[Callable] = None,
|
||
) -> List[dict]:
|
||
"""
|
||
批量生成函数签名，失败时使用降级结构。
|
||
|
||
Args:
|
||
func_reqs: 功能需求列表
|
||
knowledge: 知识库文本
|
||
on_progress: 进度回调 fn(index, total, req, signature, error)
|
||
|
||
Returns:
|
||
与 func_reqs 等长的签名 dict 列表（索引一一对应）
|
||
"""
|
||
signatures = []
|
||
total = len(func_reqs)
|
||
|
||
for i, req in enumerate(func_reqs, 1):
|
||
req_id = f"REQ.{req.index_no:02d}"
|
||
try:
|
||
sig = self.build_function_signature(req, req_id, knowledge)
|
||
error = None
|
||
except Exception as e:
|
||
sig = self._fallback_signature(req, req_id)
|
||
error = e
|
||
|
||
signatures.append(sig)
|
||
if on_progress:
|
||
on_progress(i, total, req, sig, error)
|
||
|
||
return signatures
|
||
|
||
@staticmethod
|
||
def _fallback_signature(
|
||
req: FunctionalRequirement,
|
||
requirement_id: str,
|
||
) -> dict:
|
||
"""生成降级签名结构（LLM 失败时使用）"""
|
||
return {
|
||
"name": req.function_name,
|
||
"requirement_id": requirement_id,
|
||
"description": req.description,
|
||
"type": "function",
|
||
"module": req.module or config.DEFAULT_MODULE,
|
||
"parameters": {},
|
||
"return": {
|
||
"type": "any",
|
||
"on_success": {"value": "...", "description": "成功时返回值"},
|
||
"on_failure": {"value": "None", "description": "失败时返回 None"},
|
||
},
|
||
} |