init the project
This commit is contained in:
evyzacq
+127
@@ -0,0 +1,127 @@
|
||||
---
|
||||
name: 文档分析代理
|
||||
description: 一个智能代理，用于分析文档（.docx, .pdf），提取和结构化内容，检测文本与图表之间的冲突，并生成结构化的JSON中间表示。
|
||||
---
|
||||
|
||||
# 文档分析代理
|
||||
|
||||
## 环境变量配置
|
||||
|
||||
在执行任何分析之前，必须先检查用户是否配置了DASHSCOPE_API_KEY，如果没有提示用户设置为环境变量DASHSCOPE_API_KEY。
|
||||
所有脚本通过该环境变量读取 API Key。严禁在对话或命令行中明文写入或显示 API Key。
|
||||
|
||||
### 配置方式
|
||||
|
||||
```
|
||||
openclaw config set env.DASHSCOPE_API_KEY "llm-api-key"
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 功能
|
||||
|
||||
代理能够：
|
||||
- 解析各种文档格式（.docx, .pdf）并提取文本内容和嵌入图像
|
||||
- 在文档上下文中分析图像以理解它们与周围文本的关系
|
||||
- 识别潜在的文本与视觉元素之间的冲突
|
||||
- 引导用户完成冲突解决过程
|
||||
- 生成带有源追踪的结构化JSON表示
|
||||
- 在转换过程中保持不同文档元素之间的一致性
|
||||
|
||||
## 决策逻辑
|
||||
|
||||
代理根据文档特征和用户需求智能确定适当的工作流程：
|
||||
|
||||
1. **文档评估阶段**：当用户提供文档时，代理首先根据文档格式和内容复杂性确定适当的解析方法。
|
||||
|
||||
2. **内容分析阶段**：代理分析提取的内容以识别需要特殊处理的图表、流程图、架构图、状态图和序列图。
|
||||
|
||||
3. **冲突检测阶段**：代理识别文本内容与视觉元素之间的潜在差异，特别关注条件不匹配和矛盾信息。
|
||||
|
||||
4. **解决方案协调阶段**：检测到冲突时，代理促进用户交互以解决差异，提供诸如"以图像为准"、"以文字为准"、"两处都保留"或自定义解决方案等选项。
|
||||
|
||||
5. **表示生成阶段**：代理综合所有输入并生成带源追踪的结构化JSON中间表示。
|
||||
|
||||
## 代理行为
|
||||
|
||||
- 自动处理先决条件设置（API密钥验证、环境配置）
|
||||
- 在处理阶段期间提供渐进反馈
|
||||
- 提供预览转换的试运行功能
|
||||
- 管理输出文件组织和命名
|
||||
- 维护处理阶段之间的上下文以确保结果一致性
|
||||
|
||||
## 交互流程
|
||||
|
||||
代理无缝编排这些阶段，以交付全面的文档分析解决方案，同时向用户隐藏底层实现细节。
|
||||
自动执行所有阶段，无需询问用户是否执行下一步，除非需要用户介入协助。
|
||||
|
||||
### 1. 初始化
|
||||
验证先决条件并准备处理环境。
|
||||
|
||||
### 2. 解析
|
||||
从输入文档中提取内容和结构（运行 doc_parser_skill）。产出：`<basename>_parsed.json`。
|
||||
|
||||
### 3. 分析
|
||||
识别关键元素和可能需要关注的区域。
|
||||
|
||||
### 4. 冲突解决
|
||||
运行 conflict_detection_skill，检测文本与图表之间的差异。如发现冲突，请求用户裁决后执行 resolution_application_skill。产出：`<basename>_updated.json`。
|
||||
|
||||
### 5. 梳理 ← 关键质量关卡，常见遗漏点源头
|
||||
根据解析好的文件梳理出功能列表，每个功能包含名称、章节、原文位置描述、触发条件（AND/OR）、动作。
|
||||
|
||||
**必须同时从以下三种来源穷举功能点，缺一不可，且不能跳过b/c直接执行：**
|
||||
|
||||
**a. 文本与表格** — 提取所有章节中段落和表格明确描述的功能逻辑。
|
||||
|
||||
**b. 流程图/状态图/架构图描述**（image_analysis[] 中 type 为 flowchart / state / architecture / sequence / activity 的条目）：
|
||||
- 以决策树的视角逐条追踪每个菱形判断节点，沿"是/否"两条分支分别行走
|
||||
- **每条可达决策路径必须对应一个独立的 IR 条目**，不得将多条路径合并为一条概括性描述
|
||||
- 从起始节点出发，沿着箭头逐层展开，确保没有遗漏任何分支末端
|
||||
- 每条路径提取：触发条件链（AND/OR组合）+ 执行动作 + 分支上的具体条件值
|
||||
- 特别注意图中那些"文字没有写"的隐性条件、默认行为、边界值
|
||||
|
||||
**c. 交互图/UI场景图描述**（image_analysis[] 中 type 为 other 但 description 描述UI交互场景的条目）：
|
||||
- 分析交互图中的每个场景编号（如01、02、A、B、C等）
|
||||
- 每个场景拆分出一个独立的功能点，包含：触发条件（用户操作或系统状态）、系统响应（Toast文案、页面变化等）
|
||||
- 如果多个场景在逻辑上形成流程链，保留上下游关系
|
||||
|
||||
**d. 交叉验证** — 对每个已识别的功能点标注来源（text / image_rIdXX），确保 text 来源覆盖了文档所有文字，image 来源覆盖了所有 image_analysis 中有行为逻辑的条目。
|
||||
|
||||
### 6. 合成
|
||||
根据梳理的功能列表，生成最终结构化表示（IR），输出到输出目录的 `<文档名>_ir.json` 文件中。
|
||||
使用 ir_generation_skill 脚本。
|
||||
|
||||
### 7. 检查
|
||||
对比解析好的文件（parsed.json/updated.json）和合成的 IR 文件，**对照以下清单逐项验证**，找到遗漏点：
|
||||
|
||||
**7a. 文本覆盖检查** — 遍历 parsed.json 中所有 sections[].blocks（包括 para 和 table），对每个段/格判断：
|
||||
- 是否描述了某个功能行为？
|
||||
- 若是，该行为是否已在 IR 中？
|
||||
- 遗漏的标记出来。
|
||||
|
||||
**7b. 流程图路径完整性检查** — 逐张流程图 description，对照决策树结构：
|
||||
- □ 每条"是"分支路径 → IR 中有对应条目
|
||||
- □ 每条"否"分支路径 → IR 中有对应条目
|
||||
- □ 起始节点的每个出口都被追踪到
|
||||
- □ 边角情况（分支合并、默认走法）也已覆盖
|
||||
|
||||
**7c. 交互图场景检查** — 逐张交互图 description：
|
||||
- □ 每个场景编号（01/02/A/B/C等）→ IR 中有对应条目
|
||||
- □ 每个 UX 文案 → IR 中有对应条目或作为已有条目的动作细节
|
||||
- □ 特殊行为 → IR 中有对应条件
|
||||
|
||||
**7d. 功能列表交叉检查** — 如果文档有功能列表表格：
|
||||
- □ 表格中的每一行 → IR 中有对应条目
|
||||
- □ 章节中的详细描述 → IR 中有对应条目，且粒度不粗于功能列表
|
||||
|
||||
### 8. 补充
|
||||
根据检查结果，将遗漏功能加入已有 IR 文件。
|
||||
使用 ir_generation_skill 添加每个遗漏点。
|
||||
**每补充完一批，回到步骤7重新检查**，直到所有检查项均为 ✅（即功能点一致）。
|
||||
|
||||
### 9. 输出
|
||||
提供带追踪信息的完整分析结果，包括：
|
||||
- 功能点清单总览
|
||||
- 冲突解决记录
|
||||
- 文档质量备注（如图片错误、占位符、不一致等）
|
||||
+2
@@ -0,0 +1,2 @@
|
||||
openai
|
||||
python-docx
|
||||
@@ -0,0 +1,37 @@
|
||||
---
|
||||
name: 冲突检测技能
|
||||
description: 分析解析后的文档，检测图表类图像与其相应文本描述之间的矛盾和条件不匹配。
|
||||
---
|
||||
|
||||
# 冲突检测技能
|
||||
|
||||
## 概述
|
||||
|
||||
此技能识别解析文档中文本内容与视觉内容之间的潜在冲突。它特别针对图表类图像（流程图、架构图、状态图、序列图和活动图）并交叉检查其描述与同文档部分的文本内容。
|
||||
|
||||
## 功能
|
||||
|
||||
该技能：
|
||||
- 从解析的文档结构中识别图表类图像
|
||||
- 将图像描述与同一文档部分中的相应文本内容进行交叉引用
|
||||
- 检测视觉表示和文本表示之间的矛盾和条件不匹配
|
||||
- 生成包含其位置的已识别冲突的结构化列表
|
||||
- 专门针对流程图、架构图、状态图、序列图和活动图
|
||||
|
||||
## 输入要求
|
||||
|
||||
- 解析文档JSON文件的路径（由文档解析技能生成）
|
||||
- 可选输出目录规范
|
||||
- 可选试运行标志，在不调用API的情况下预览大语言模型提示
|
||||
|
||||
## 输出
|
||||
|
||||
该技能生成一个结构化JSON文件，文件名为输入文档的基本名称后跟'_conflicts.json'，包含：
|
||||
- 带有关于差异详情的冲突对象列表
|
||||
- 标识每个冲突发生位置的节标识符
|
||||
- 冲突图像和文本内容的片段
|
||||
- 每个冲突的类型分类（例如，矛盾、条件不匹配）
|
||||
|
||||
## 集成点
|
||||
|
||||
此技能消耗文档解析技能的输出并为解决方案应用技能提供输入。冲突解决过程通常需要人工输入才能进入下一阶段。
|
||||
+105
@@ -0,0 +1,105 @@
|
||||
import logging
|
||||
import os
|
||||
import time
|
||||
from typing import Optional
|
||||
|
||||
from openai import OpenAI
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
class LLMClient:
|
||||
"""Low-level OpenAI-compatible LLM client with retry and token tracking.
|
||||
|
||||
Usage::
|
||||
|
||||
llm = LLMClient()
|
||||
content = llm.chat("qwen3.5-flash", [{"role": "user", "content": "Hello"}])
|
||||
print(llm.usage)
|
||||
"""
|
||||
|
||||
IMAGE_MODEL = "qwen3-vl-plus"
|
||||
TEXT_MODEL = "qwen3.5-flash-2026-02-23"
|
||||
TIMEOUT = 120
|
||||
MAX_RETRIES = 3
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
*,
|
||||
base_url: str = "https://dashscope.aliyuncs.com/compatible-mode/v1",
|
||||
timeout: int | None = None,
|
||||
):
|
||||
key = os.environ.get("DASHSCOPE_API_KEY", "")
|
||||
if not key:
|
||||
raise ValueError("DASHSCOPE_API_KEY environment variable is not set.")
|
||||
self._client = OpenAI(api_key=key, base_url=base_url)
|
||||
self._timeout = timeout or self.TIMEOUT
|
||||
self._prompt_tokens = 0
|
||||
self._completion_tokens = 0
|
||||
|
||||
@property
|
||||
def usage(self) -> dict:
|
||||
"""Return accumulated token counts as ``{prompt, completion, total}``."""
|
||||
return {
|
||||
"prompt_tokens": self._prompt_tokens,
|
||||
"completion_tokens": self._completion_tokens,
|
||||
"total_tokens": self._prompt_tokens + self._completion_tokens,
|
||||
}
|
||||
|
||||
@staticmethod
|
||||
def estimate_tokens(text: str) -> int:
|
||||
"""Quick token estimate. CJK ≈1.7/token, others ≈3.0/token."""
|
||||
cjk = sum(1 for c in text if '一' <= c <= '鿿' or ' ' <= c <= '〿')
|
||||
other = len(text) - cjk
|
||||
return max(1, int(cjk / 1.7 + other / 3.0))
|
||||
|
||||
@staticmethod
|
||||
def estimate_image_tokens() -> int:
|
||||
"""Fixed estimate for one vision-model image (~500 tokens)."""
|
||||
return 500
|
||||
|
||||
def chat(
|
||||
self, model: str, messages: list[dict], *, timeout: int | None = None,
|
||||
response_format: dict | None = None,
|
||||
) -> str:
|
||||
"""Send a chat completion request and return the response content.
|
||||
|
||||
Automatically retries on failure and accumulates token usage.
|
||||
"""
|
||||
label = f"chat({model})"
|
||||
|
||||
def _call():
|
||||
t0 = time.time()
|
||||
kwargs = dict(model=model, messages=messages, timeout=timeout or self._timeout)
|
||||
if response_format is not None:
|
||||
kwargs["response_format"] = response_format
|
||||
kwargs["temperature"] = 0
|
||||
resp = self._client.chat.completions.create(**kwargs)
|
||||
content = resp.choices[0].message.content
|
||||
usg = resp.usage
|
||||
if usg:
|
||||
self._prompt_tokens += usg.prompt_tokens
|
||||
self._completion_tokens += usg.completion_tokens
|
||||
elapsed = time.time() - t0
|
||||
logger.info("%s: %d chars in %.1fs", label, len(content) if content else 0, elapsed)
|
||||
if not content:
|
||||
raise RuntimeError("Empty response from LLM")
|
||||
return content
|
||||
|
||||
return self._retry(_call, label)
|
||||
|
||||
def _retry(self, fn, label: str) -> str:
|
||||
"""Call *fn()* with exponential-backoff retry."""
|
||||
last_error: Optional[Exception] = None
|
||||
for attempt in range(self.MAX_RETRIES):
|
||||
try:
|
||||
return fn()
|
||||
except Exception as e:
|
||||
last_error = e
|
||||
logger.warning(
|
||||
"%s error (attempt %d/%d): %s",
|
||||
label, attempt + 1, self.MAX_RETRIES, e,
|
||||
)
|
||||
if attempt < self.MAX_RETRIES - 1:
|
||||
time.sleep(2 ** attempt)
|
||||
raise RuntimeError(f"{label}: all retries exhausted") from last_error
|
||||
Executable
+280
@@ -0,0 +1,280 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Detect logical conflicts between image analysis and text in ``_parsed.json``.
|
||||
|
||||
Usage::
|
||||
|
||||
python scripts/detect_conflicts.py D:/projects/jike/output/车机娱乐系统禁止功能文档_精简_parsed.json [--output-dir DIR]
|
||||
|
||||
For each diagram-type image (flowchart, architecture, state, sequence, activity),
|
||||
the script locates its section via *image_sources*, grabs the corresponding text
|
||||
blocks, and calls an LLM to find contradictions/condition-mismatches between the
|
||||
image description and the text.
|
||||
|
||||
Output: ``<basename>_conflicts.json``
|
||||
"""
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import logging
|
||||
import os
|
||||
import re
|
||||
import sys
|
||||
import time
|
||||
|
||||
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
|
||||
|
||||
from LLM import LLMClient
|
||||
|
||||
logging.basicConfig(
|
||||
level=logging.INFO,
|
||||
format="%(asctime)s [%(levelname)s] %(message)s",
|
||||
datefmt="%Y-%m-%d %H:%M:%S",
|
||||
)
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
RATE_LIMIT_DELAY = 0.5
|
||||
DIAGRAM_TYPES = {"flowchart", "architecture", "state", "sequence", "activity"}
|
||||
MIN_TEXT_CHARS = 20
|
||||
|
||||
PROMPT_DETECT_CONFLICT = """你是一个文档一致性检查专家。以下内容来自同一份需求文档的同一个章节，包含两部分：
|
||||
|
||||
## 部分1：图片（流程图/架构图/状态图）的描述
|
||||
```
|
||||
{image_description}
|
||||
```
|
||||
|
||||
## 部分2：同章节的文字描述
|
||||
```
|
||||
{text_description}
|
||||
```
|
||||
|
||||
## 你的任务
|
||||
检查这两部分之间是否存在**逻辑矛盾或条件不一致**。
|
||||
|
||||
你需要关注的冲突类型：
|
||||
|
||||
1. **condition_mismatch**（条件不一致）：两者描述了同一规则，但触发条件、阈值、时序不同。
|
||||
例如：图片说"车速≥15km/h且持续5秒"，文字说"车速≥10km/h且持续3秒"。
|
||||
例如：图片说"非P档限制"，文字说"车速>0限制"。
|
||||
|
||||
2. **contradiction**（直接矛盾）：两者对同一事物的描述完全相反。
|
||||
例如：图片说"功能X被禁止"，文字说"功能X可用"。
|
||||
例如：图片说"开关默认关闭"，文字说"开关默认开启"。
|
||||
|
||||
3. **scope_mismatch**（范围不一致）：两者描述的场景/地域/设备范围不同。
|
||||
例如：图片说"国内方案"，文字说"海外方案"。
|
||||
例如：图片说"CSD中控屏"，文字描述包含"PSD副驾屏"。
|
||||
|
||||
## 输出格式
|
||||
|
||||
如果**没有冲突**，只输出：
|
||||
```
|
||||
[[NO_CONFLICT]]
|
||||
```
|
||||
|
||||
如果**有冲突**，输出以下JSON数组（不要任何其他文字）：
|
||||
|
||||
```json
|
||||
[
|
||||
{{
|
||||
"conflict_type": "condition_mismatch",
|
||||
"severity": "high",
|
||||
"section": "{section_name}",
|
||||
"image_snippet": "图片中描述的关键内容（摘录）",
|
||||
"text_snippet": "文字中描述的关键内容（摘录）",
|
||||
"description": "用中文说明冲突的具体差异"
|
||||
}}
|
||||
]
|
||||
```
|
||||
|
||||
注意：
|
||||
- 每个冲突一个条目，不要合并
|
||||
- severity: high（功能正确性受影响）| medium（边界条件模糊）| low（表达方式差异）
|
||||
- 输出必须是**严格合法的JSON数组**，不要有尾随逗号
|
||||
- 如果没有严格冲突，输出 [[NO_CONFLICT]]
|
||||
"""
|
||||
|
||||
|
||||
def _build_text_for_section(sections: list[dict], section_name: str) -> str:
|
||||
"""Build a single text block for the given section name."""
|
||||
texts: list[str] = []
|
||||
for sec in sections:
|
||||
if sec.get("source", "") == section_name:
|
||||
for blk in sec.get("blocks", []):
|
||||
if blk["type"] == "para":
|
||||
texts.append(blk["text"])
|
||||
elif blk["type"] == "table":
|
||||
table_lines = [f"表格 {blk['table']}:"]
|
||||
for ri, row in enumerate(blk.get("rows", [])):
|
||||
cols = row.get("columns", [])
|
||||
parts = [f"{c['name']}: {c['text']}" for c in cols]
|
||||
table_lines.append(f" 行{ri + 1}: {' | '.join(parts)}")
|
||||
texts.append("\n".join(table_lines))
|
||||
return "\n\n".join(texts)
|
||||
|
||||
|
||||
def _parse_conflict_json(content: str) -> list[dict]:
|
||||
"""Extract JSON array from LLM response, handling markdown fences."""
|
||||
stripped = content.strip()
|
||||
|
||||
if "[[NO_CONFLICT]]" in stripped:
|
||||
return []
|
||||
|
||||
# Remove markdown code fences
|
||||
if "```json" in stripped:
|
||||
stripped = stripped.split("```json", 1)[1]
|
||||
if "```" in stripped:
|
||||
stripped = stripped.split("```", 1)[0]
|
||||
elif "```" in stripped:
|
||||
stripped = stripped.split("```", 1)[1]
|
||||
if "```" in stripped:
|
||||
stripped = stripped.split("```", 1)[0]
|
||||
|
||||
stripped = stripped.strip()
|
||||
if not stripped:
|
||||
return []
|
||||
|
||||
# Try to find a JSON array
|
||||
match = re.search(r"\[\s*\{.*\}\s*\]", stripped, re.DOTALL)
|
||||
if match:
|
||||
stripped = match.group()
|
||||
|
||||
try:
|
||||
conflicts = json.loads(stripped)
|
||||
if isinstance(conflicts, list):
|
||||
return conflicts
|
||||
return []
|
||||
except json.JSONDecodeError as e:
|
||||
logger.warning("Failed to parse conflict JSON: %s", e)
|
||||
logger.debug("Raw content: %s", stripped)
|
||||
return []
|
||||
|
||||
|
||||
def detect_conflicts(
|
||||
parsed_path: str,
|
||||
output_dir: str | None = None,
|
||||
*,
|
||||
dry_run: bool = False,
|
||||
) -> list[dict]:
|
||||
"""Load ``_parsed.json`` and detect image-vs-text conflicts.
|
||||
|
||||
Returns a flat list of conflict dicts and writes to ``<basename>_conflicts.json``.
|
||||
"""
|
||||
with open(parsed_path, "r", encoding="utf-8") as f:
|
||||
data = json.load(f)
|
||||
|
||||
basename = os.path.splitext(os.path.basename(parsed_path))[0]
|
||||
if basename.endswith("_parsed"):
|
||||
basename = basename[:-7]
|
||||
|
||||
if output_dir is None:
|
||||
output_dir = os.path.dirname(os.path.abspath(parsed_path))
|
||||
os.makedirs(output_dir, exist_ok=True)
|
||||
|
||||
sections = data.get("sections", [])
|
||||
image_sources = data.get("image_sources", {})
|
||||
image_analysis = data.get("image_analysis", [])
|
||||
|
||||
llm = LLMClient()
|
||||
all_conflicts: list[dict] = []
|
||||
|
||||
# ---- For each diagram image, compare with its section text -------------
|
||||
for img in image_analysis:
|
||||
img_type = img.get("type", "other")
|
||||
rid = img.get("rid", "")
|
||||
description = img.get("description", "").strip()
|
||||
|
||||
if img_type not in DIAGRAM_TYPES or not description:
|
||||
logger.info("Skip conflict check: rid=%s type=%s", rid, img_type)
|
||||
continue
|
||||
|
||||
# Find source section
|
||||
src = image_sources.get(rid, {})
|
||||
section_name = src.get("section", "")
|
||||
|
||||
if not section_name:
|
||||
logger.warning("No section found for rid=%s, skipping", rid)
|
||||
continue
|
||||
|
||||
# Build text from the same section
|
||||
text_content = _build_text_for_section(sections, section_name)
|
||||
text_len = len(text_content.strip())
|
||||
if text_len < MIN_TEXT_CHARS:
|
||||
logger.info("Section text too short (%d chars) for rid=%s, skip", text_len, rid)
|
||||
continue
|
||||
|
||||
logger.info("Checking conflicts: rid=%s section=%s (desc=%d chars, text=%d chars)",
|
||||
rid, section_name, len(description), text_len)
|
||||
|
||||
if dry_run:
|
||||
logger.info(" [DRY RUN] would call LLM to detect conflicts")
|
||||
continue
|
||||
|
||||
prompt = PROMPT_DETECT_CONFLICT.format(
|
||||
image_description=description,
|
||||
text_description=text_content,
|
||||
section_name=section_name,
|
||||
)
|
||||
|
||||
try:
|
||||
raw = llm.chat(
|
||||
model=LLMClient.TEXT_MODEL,
|
||||
messages=[{"role": "user", "content": prompt}],
|
||||
)
|
||||
logger.info("Conflict check response: %d chars", len(raw))
|
||||
except RuntimeError as e:
|
||||
logger.error("Conflict check failed: %s", e)
|
||||
continue
|
||||
|
||||
conflicts = _parse_conflict_json(raw)
|
||||
|
||||
# Enrich with location info
|
||||
for c in conflicts:
|
||||
c["rid"] = rid
|
||||
c["image_path"] = img.get("path", "")
|
||||
if "section" not in c:
|
||||
c["section"] = section_name
|
||||
if src.get("table"):
|
||||
c.setdefault("source_location", {})["table"] = src["table"]
|
||||
if src.get("row"):
|
||||
c.setdefault("source_location", {})["image_row"] = src["row"]
|
||||
|
||||
all_conflicts.extend(conflicts)
|
||||
logger.info(" Found %d conflicts for rid=%s", len(conflicts), rid)
|
||||
|
||||
if any(x.get("type") in DIAGRAM_TYPES
|
||||
for x in image_analysis
|
||||
if x.get("rid", "") != rid):
|
||||
time.sleep(RATE_LIMIT_DELAY)
|
||||
|
||||
# ---- Save ---------------------------------------------------------------
|
||||
conflicts_path = os.path.join(output_dir, f"{basename}_conflicts.json")
|
||||
with open(conflicts_path, "w", encoding="utf-8") as f:
|
||||
json.dump(all_conflicts, f, ensure_ascii=False, indent=2)
|
||||
logger.info("Saved: %s (%d conflicts)", conflicts_path, len(all_conflicts))
|
||||
|
||||
# ---- Summary ------------------------------------------------------------
|
||||
usg = llm.usage
|
||||
logger.info("Tokens: %d prompt + %d completion = %d total",
|
||||
usg["prompt_tokens"], usg["completion_tokens"], usg["total_tokens"])
|
||||
|
||||
return all_conflicts
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# CLI
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
if __name__ == "__main__":
|
||||
parser = argparse.ArgumentParser(
|
||||
description="Detect image-vs-text conflicts in parsed document.",
|
||||
)
|
||||
parser.add_argument("input", metavar="parsed.json", help="Path to _parsed.json from doc_parser")
|
||||
parser.add_argument("--output-dir", metavar="DIR", default=None,
|
||||
help="Output directory (default: same as input)")
|
||||
parser.add_argument("--dry-run", action="store_true",
|
||||
help="Print LLM prompts without calling the API.")
|
||||
|
||||
args = parser.parse_args()
|
||||
detect_conflicts(args.input, output_dir=args.output_dir, dry_run=args.dry_run)
|
||||
@@ -0,0 +1,36 @@
|
||||
---
|
||||
name: 文档解析技能
|
||||
description: 解析文档（.docx, .pdf）以提取图像和文本结构，并使用视觉大语言模型分析每个图像的类型和描述。
|
||||
---
|
||||
|
||||
# 文档解析技能
|
||||
|
||||
## 概述
|
||||
|
||||
此技能从文档（.docx, .pdf）中提取内容并准备进行进一步分析。它提取文本内容和嵌入图像，并对图像执行初始分析以了解其类型和内容。
|
||||
|
||||
## 功能
|
||||
|
||||
该技能：
|
||||
- 从文档中提取文本结构（段落、表格、标题）
|
||||
- 识别并提取嵌入的图像
|
||||
- 使用视觉大语言模型分析每个图像并确定其类型和内容描述
|
||||
- 生成结构化输出，将图像映射到其在文档中的位置
|
||||
- 创建文档的初始解析表示，供后续处理阶段使用
|
||||
|
||||
## 输入要求
|
||||
|
||||
- 文档文件路径（必需，支持.docx和.pdf格式）
|
||||
- 可选输出目录（默认为'output/'）
|
||||
- 可选试运行标志，在不调用API的情况下预览大语言模型提示
|
||||
|
||||
## 输出
|
||||
|
||||
该技能生成一个结构化JSON文件，文件名为输入文档的基本名称后跟'_parsed.json'，包含：
|
||||
- `sections`：按标题分组的文档文本结构
|
||||
- `image_sources`：从图像标识符到其在文档中位置的映射
|
||||
- `image_analysis`：由视觉大语言模型确定的每个图像的类型和内容描述
|
||||
|
||||
## 集成点
|
||||
|
||||
此技能作为文档分析管道中的初始处理步骤。其输出被冲突检测技能消费以识别文本和视觉内容之间的差异。
|
||||
@@ -0,0 +1,105 @@
|
||||
import logging
|
||||
import os
|
||||
import time
|
||||
from typing import Optional
|
||||
|
||||
from openai import OpenAI
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
class LLMClient:
|
||||
"""Low-level OpenAI-compatible LLM client with retry and token tracking.
|
||||
|
||||
Usage::
|
||||
|
||||
llm = LLMClient()
|
||||
content = llm.chat("qwen3.5-flash", [{"role": "user", "content": "Hello"}])
|
||||
print(llm.usage)
|
||||
"""
|
||||
|
||||
IMAGE_MODEL = "qwen3-vl-plus"
|
||||
TEXT_MODEL = "qwen3.5-flash-2026-02-23"
|
||||
TIMEOUT = 120
|
||||
MAX_RETRIES = 3
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
*,
|
||||
base_url: str = "https://dashscope.aliyuncs.com/compatible-mode/v1",
|
||||
timeout: int | None = None,
|
||||
):
|
||||
key = os.environ.get("DASHSCOPE_API_KEY", "")
|
||||
if not key:
|
||||
raise ValueError("DASHSCOPE_API_KEY environment variable is not set.")
|
||||
self._client = OpenAI(api_key=key, base_url=base_url)
|
||||
self._timeout = timeout or self.TIMEOUT
|
||||
self._prompt_tokens = 0
|
||||
self._completion_tokens = 0
|
||||
|
||||
@property
|
||||
def usage(self) -> dict:
|
||||
"""Return accumulated token counts as ``{prompt, completion, total}``."""
|
||||
return {
|
||||
"prompt_tokens": self._prompt_tokens,
|
||||
"completion_tokens": self._completion_tokens,
|
||||
"total_tokens": self._prompt_tokens + self._completion_tokens,
|
||||
}
|
||||
|
||||
@staticmethod
|
||||
def estimate_tokens(text: str) -> int:
|
||||
"""Quick token estimate. CJK ≈1.7/token, others ≈3.0/token."""
|
||||
cjk = sum(1 for c in text if '一' <= c <= '鿿' or ' ' <= c <= '〿')
|
||||
other = len(text) - cjk
|
||||
return max(1, int(cjk / 1.7 + other / 3.0))
|
||||
|
||||
@staticmethod
|
||||
def estimate_image_tokens() -> int:
|
||||
"""Fixed estimate for one vision-model image (~500 tokens)."""
|
||||
return 500
|
||||
|
||||
def chat(
|
||||
self, model: str, messages: list[dict], *, timeout: int | None = None,
|
||||
response_format: dict | None = None,
|
||||
) -> str:
|
||||
"""Send a chat completion request and return the response content.
|
||||
|
||||
Automatically retries on failure and accumulates token usage.
|
||||
"""
|
||||
label = f"chat({model})"
|
||||
|
||||
def _call():
|
||||
t0 = time.time()
|
||||
kwargs = dict(model=model, messages=messages, timeout=timeout or self._timeout)
|
||||
if response_format is not None:
|
||||
kwargs["response_format"] = response_format
|
||||
kwargs["temperature"] = 0
|
||||
resp = self._client.chat.completions.create(**kwargs)
|
||||
content = resp.choices[0].message.content
|
||||
usg = resp.usage
|
||||
if usg:
|
||||
self._prompt_tokens += usg.prompt_tokens
|
||||
self._completion_tokens += usg.completion_tokens
|
||||
elapsed = time.time() - t0
|
||||
logger.info("%s: %d chars in %.1fs", label, len(content) if content else 0, elapsed)
|
||||
if not content:
|
||||
raise RuntimeError("Empty response from LLM")
|
||||
return content
|
||||
|
||||
return self._retry(_call, label)
|
||||
|
||||
def _retry(self, fn, label: str) -> str:
|
||||
"""Call *fn()* with exponential-backoff retry."""
|
||||
last_error: Optional[Exception] = None
|
||||
for attempt in range(self.MAX_RETRIES):
|
||||
try:
|
||||
return fn()
|
||||
except Exception as e:
|
||||
last_error = e
|
||||
logger.warning(
|
||||
"%s error (attempt %d/%d): %s",
|
||||
label, attempt + 1, self.MAX_RETRIES, e,
|
||||
)
|
||||
if attempt < self.MAX_RETRIES - 1:
|
||||
time.sleep(2 ** attempt)
|
||||
raise RuntimeError(f"{label}: all retries exhausted") from last_error
|
||||
+106
@@ -0,0 +1,106 @@
|
||||
import argparse
|
||||
import json
|
||||
import logging
|
||||
import os
|
||||
import sys
|
||||
import time
|
||||
|
||||
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
|
||||
|
||||
from image_parser import ImageParser
|
||||
from LLM import LLMClient
|
||||
from word_parser import WordParser
|
||||
|
||||
logging.basicConfig(
|
||||
level=logging.INFO,
|
||||
format="%(asctime)s [%(levelname)s] %(message)s",
|
||||
datefmt="%Y-%m-%d %H:%M:%S",
|
||||
)
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
RATE_LIMIT_DELAY = 0.5
|
||||
|
||||
|
||||
def parse_document(
|
||||
docx_path: str,
|
||||
output_dir: str = "output",
|
||||
*,
|
||||
dry_run: bool = False,
|
||||
) -> dict:
|
||||
"""Parse a .docx file: extract text structure and parse embedded images.
|
||||
|
||||
Produces ``<basename>_parsed.json`` in *output_dir*.
|
||||
"""
|
||||
word = WordParser(docx_path)
|
||||
|
||||
basename = os.path.splitext(os.path.basename(docx_path))[0]
|
||||
os.makedirs(output_dir, exist_ok=True)
|
||||
images_dir = os.path.join(output_dir, "images")
|
||||
|
||||
# ---- extract sections and images -----------------------------------------
|
||||
sections, image_sources = word.extract_sections()
|
||||
logger.info("Document has %d sections, %d image sources", len(sections), len(image_sources))
|
||||
|
||||
# ---- parse images ----------------------------------------------------------
|
||||
images = word.extract_images(images_dir)
|
||||
logger.info("Found %d images in document", len(images))
|
||||
|
||||
image_analysis: list[dict] = []
|
||||
|
||||
if images:
|
||||
llm = ImageParser()
|
||||
for i, img in enumerate(images):
|
||||
logger.info("[image %d/%d] rid=%s", i + 1, len(images), img["rid"])
|
||||
if dry_run:
|
||||
est = LLMClient.estimate_image_tokens()
|
||||
logger.info(" [DRY RUN] would call vision LLM (~%d tokens)", est)
|
||||
result = {"type": "other", "description": "[DRY RUN]"}
|
||||
else:
|
||||
result = llm.parse_image(img["path"])
|
||||
if result is None:
|
||||
result = {"type": "other", "description": ""}
|
||||
result["rid"] = img["rid"]
|
||||
result["path"] = img["path"]
|
||||
image_analysis.append(result)
|
||||
if i < len(images) - 1:
|
||||
time.sleep(RATE_LIMIT_DELAY)
|
||||
|
||||
usg = llm.usage
|
||||
logger.info("Tokens: %d prompt + %d completion = %d total",
|
||||
usg["prompt_tokens"], usg["completion_tokens"], usg["total_tokens"])
|
||||
else:
|
||||
logger.info("No images found in document")
|
||||
|
||||
# ---- build output --------------------------------------------------------
|
||||
output = {
|
||||
"source": os.path.abspath(docx_path),
|
||||
"sections": sections,
|
||||
"image_sources": image_sources,
|
||||
"image_analysis": image_analysis,
|
||||
}
|
||||
|
||||
parsed_path = os.path.join(output_dir, f"{basename}_parsed.json")
|
||||
with open(parsed_path, "w", encoding="utf-8") as f:
|
||||
json.dump(output, f, ensure_ascii=False, indent=2)
|
||||
logger.info("Saved: %s", parsed_path)
|
||||
|
||||
return output
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# CLI
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
if __name__ == "__main__":
|
||||
parser = argparse.ArgumentParser(
|
||||
description="Parse a .docx file: extract text structure and parse images.",
|
||||
)
|
||||
parser.add_argument("input", metavar="input.docx", help="Path to the Word document")
|
||||
parser.add_argument("output_dir", nargs="?", default="output", metavar="output_dir",
|
||||
help="Directory for output files (default: output/)")
|
||||
parser.add_argument("--dry-run", action="store_true",
|
||||
help="Print LLM prompts without calling the API.")
|
||||
|
||||
args = parser.parse_args()
|
||||
parse_document(args.input, args.output_dir, dry_run=args.dry_run)
|
||||
+123
@@ -0,0 +1,123 @@
|
||||
import base64
|
||||
import logging
|
||||
import os
|
||||
from typing import Optional
|
||||
|
||||
from LLM import LLMClient
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# Prompts
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
PROMPT_IMAGE = """请分析这张图片，判断类型并输出文字描述。
|
||||
|
||||
## 判断图片类型
|
||||
|
||||
如果是 **流程图 / 架构图 / 状态图 / 时序图 / 活动图**，详细描述：
|
||||
- 图中所有节点/步骤/状态/组件的名称
|
||||
- 所有连线/箭头/转换关系及其方向
|
||||
- 所有分支条件、判断逻辑和判断结果
|
||||
- 所有文字标注、注释、标签
|
||||
- 图的整体结构和逻辑流程
|
||||
- 如果图片包含多个子图，拆解描述
|
||||
|
||||
如果是 **其他类型**（UI原型图 / 界面截图 / 设计稿 / 手机屏幕截图 / 网页截图等），简要描述图片内容。
|
||||
|
||||
## 输出格式
|
||||
|
||||
**1. 类型标签（单独一行）：**
|
||||
type: <flowchart|architecture|state|sequence|activity|other>
|
||||
|
||||
**2. 文字描述：**
|
||||
该图片的详细文字描述。
|
||||
|
||||
不要输出 ---YAML--- 分隔符或 YAML 内容，不要添加任何额外的解释或问候语。"""
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# ImageParser
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
class ImageParser:
|
||||
"""Vision LLM wrapper for parsing images (type + description).
|
||||
|
||||
Usage::
|
||||
|
||||
parser = ImageParser()
|
||||
result = parser.parse_image("images/img1.png")
|
||||
"""
|
||||
|
||||
_VALID_TYPES = {"flowchart", "architecture", "state", "sequence", "activity", "text"}
|
||||
|
||||
def __init__(self, llm: LLMClient | None = None):
|
||||
self._llm = llm or LLMClient()
|
||||
|
||||
@property
|
||||
def usage(self) -> dict:
|
||||
return self._llm.usage
|
||||
|
||||
def parse_image(self, image_path: str) -> Optional[dict]:
|
||||
"""Parse an image and return its type and description (no YAML IR).
|
||||
|
||||
Returns ``{type, description}``, or *None* for UI mockups.
|
||||
"""
|
||||
logger.info("Parsing image: %s", image_path)
|
||||
|
||||
with open(image_path, "rb") as f:
|
||||
img_b64 = base64.b64encode(f.read()).decode()
|
||||
mime = self._mime_type(image_path)
|
||||
|
||||
try:
|
||||
content = self._llm.chat(
|
||||
model=LLMClient.IMAGE_MODEL,
|
||||
messages=[{
|
||||
"role": "user",
|
||||
"content": [
|
||||
{"type": "image_url", "image_url": {"url": f"data:{mime};base64,{img_b64}"}},
|
||||
{"type": "text", "text": PROMPT_IMAGE},
|
||||
],
|
||||
}],
|
||||
)
|
||||
except RuntimeError as e:
|
||||
logger.error(str(e))
|
||||
return {"type": "other", "description": "", "error": str(e)}
|
||||
|
||||
parsed = self._parse_type_and_description(content)
|
||||
if parsed is None:
|
||||
return None
|
||||
return {"type": parsed[0], "description": parsed[1]}
|
||||
|
||||
# ---- internals ----------------------------------------------------------
|
||||
|
||||
def _parse_type_and_description(self, content: str) -> Optional[tuple[str, str]]:
|
||||
"""Extract ``(type, description)`` from LLM response.
|
||||
|
||||
Returns *None* for ``[[UI]]`` (skip).
|
||||
"""
|
||||
content = content.strip()
|
||||
if content == "[[UI]]" or content.startswith("[[UI]]"):
|
||||
return None
|
||||
|
||||
parsed_type = "other"
|
||||
desc_lines: list[str] = []
|
||||
for line in content.splitlines():
|
||||
stripped = line.strip()
|
||||
if (stripped.startswith("type:") or stripped.startswith("类型:")) and parsed_type == "other":
|
||||
type_val = stripped.split(":", 1)[1].strip().lower()
|
||||
if type_val in self._VALID_TYPES:
|
||||
parsed_type = type_val
|
||||
else:
|
||||
desc_lines.append(line)
|
||||
|
||||
return parsed_type, "\n".join(desc_lines).strip()
|
||||
|
||||
@staticmethod
|
||||
def _mime_type(image_path: str) -> str:
|
||||
ext = os.path.splitext(image_path)[1].lstrip(".").lower()
|
||||
return {
|
||||
"png": "image/png", "jpg": "image/jpeg", "jpeg": "image/jpeg",
|
||||
"gif": "image/gif", "bmp": "image/bmp",
|
||||
"webp": "image/webp", "svg": "image/svg+xml", "tiff": "image/tiff",
|
||||
}.get(ext, "image/png")
|
||||
+239
@@ -0,0 +1,239 @@
|
||||
import logging
|
||||
import os
|
||||
|
||||
from docx import Document
|
||||
from docx.table import Table
|
||||
from docx.text.paragraph import Paragraph
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
IMAGE_EXT = {
|
||||
"image/png": ".png",
|
||||
"image/jpeg": ".jpg",
|
||||
"image/gif": ".gif",
|
||||
"image/bmp": ".bmp",
|
||||
"image/tiff": ".tiff",
|
||||
"image/webp": ".webp",
|
||||
"image/x-emf": ".emf",
|
||||
"image/x-wmf": ".wmf",
|
||||
"image/svg+xml": ".svg",
|
||||
}
|
||||
|
||||
|
||||
class WordParser:
|
||||
"""Parse a .docx file — extract images, split body into sections.
|
||||
|
||||
Usage::
|
||||
|
||||
parser = WordParser("doc.docx")
|
||||
parser.extract_images("images/")
|
||||
sections, image_sources = parser.extract_sections()
|
||||
"""
|
||||
|
||||
HEADER_CELL_MAX_LEN = 20 # max chars per cell to treat first row as header
|
||||
|
||||
WML_NS = "http://schemas.openxmlformats.org/wordprocessingml/2006/main"
|
||||
DRAW_NS = "http://schemas.openxmlformats.org/drawingml/2006/main"
|
||||
REL_NS = "http://schemas.openxmlformats.org/officeDocument/2006/relationships"
|
||||
|
||||
def __init__(self, docx_path: str):
|
||||
if not os.path.isfile(docx_path):
|
||||
raise FileNotFoundError(f"Document not found: {docx_path}")
|
||||
self._doc = Document(docx_path)
|
||||
|
||||
# ---- public API ---------------------------------------------------------
|
||||
|
||||
def extract_images(self, images_dir: str) -> list[dict]:
|
||||
"""Save all images to *images_dir*. Returns ``[{rid, path}, ...]``."""
|
||||
os.makedirs(images_dir, exist_ok=True)
|
||||
images: list[dict] = []
|
||||
for rel in self._doc.part.rels.values():
|
||||
if "image" not in rel.reltype:
|
||||
continue
|
||||
ext = IMAGE_EXT.get(rel.target_part.content_type, ".png")
|
||||
name = f"image_{rel.rId}{ext}"
|
||||
path = os.path.join(images_dir, name)
|
||||
with open(path, "wb") as f:
|
||||
f.write(rel.target_part.blob)
|
||||
images.append({"rid": rel.rId, "path": path})
|
||||
return images
|
||||
|
||||
def extract_sections(self) -> tuple[list[dict], dict[str, dict]]:
|
||||
"""Walk document body and split into sections by heading.
|
||||
|
||||
Returns:
|
||||
*sections* — ``[{source, blocks, images}, ...]``
|
||||
Each block is ``{type, index, text}`` (paragraph) or
|
||||
``{type, table, headers, rows}`` (table).
|
||||
*image_sources* — ``rid → {section, table?, row?, column?, name?}``
|
||||
"""
|
||||
sections: list[dict] = []
|
||||
current_source = ""
|
||||
blocks: list[dict] = []
|
||||
section_images: list[str] = []
|
||||
image_sources: dict[str, dict] = {}
|
||||
para_idx = 0
|
||||
tbl_idx = 0
|
||||
|
||||
for child in self._doc.element.body:
|
||||
tag = child.tag.split("}")[-1] if "}" in child.tag else child.tag
|
||||
|
||||
if tag == "p":
|
||||
para = Paragraph(child, self._doc)
|
||||
|
||||
if self._heading_level(para) is not None:
|
||||
heading_text = para.text.strip()
|
||||
if heading_text: # ignore empty heading-like paragraphs
|
||||
if blocks or section_images:
|
||||
sections.append({
|
||||
"source": current_source,
|
||||
"blocks": blocks,
|
||||
"images": list(section_images),
|
||||
})
|
||||
blocks = []
|
||||
section_images = []
|
||||
para_idx = 0
|
||||
tbl_idx = 0
|
||||
current_source = heading_text
|
||||
continue
|
||||
|
||||
text = para.text.strip()
|
||||
|
||||
# Scan for images — append [[IMAGE:rid]] markers
|
||||
for run in para.runs:
|
||||
for rid in self._images_in(run._element):
|
||||
text += f" [[IMAGE:{rid}]]"
|
||||
section_images.append(rid)
|
||||
image_sources[rid] = {"section": current_source}
|
||||
|
||||
if text.strip():
|
||||
blocks.append({"type": "para", "index": para_idx + 1, "text": text.strip()})
|
||||
para_idx += 1
|
||||
|
||||
elif tag == "tbl":
|
||||
tbl_idx += 1
|
||||
table = Table(child, self._doc)
|
||||
|
||||
# Collect all rows as [[cell_text, ...], ...]
|
||||
all_rows: list[list[str]] = []
|
||||
all_images: list[list[list[str]]] = [] # row → col → [rids]
|
||||
for row in table.rows:
|
||||
row_texts: list[str] = []
|
||||
row_cell_images: list[list[str]] = []
|
||||
for cell in row.cells:
|
||||
cell_text = cell.text.strip()
|
||||
cell_imgs: list[str] = []
|
||||
for cp in cell.paragraphs:
|
||||
for run in cp.runs:
|
||||
for rid in self._images_in(run._element):
|
||||
cell_imgs.append(rid)
|
||||
# Replace images with markers in text
|
||||
for rid in cell_imgs:
|
||||
cell_text += f" [[IMAGE:{rid}]]"
|
||||
section_images.append(rid)
|
||||
row_texts.append(cell_text.strip())
|
||||
row_cell_images.append(cell_imgs)
|
||||
if any(row_texts) or any(row_cell_images):
|
||||
all_rows.append(row_texts)
|
||||
all_images.append(row_cell_images)
|
||||
|
||||
if len(all_rows) >= 2:
|
||||
# Heuristic: first row is a header if every cell is short
|
||||
first_row = all_rows[0]
|
||||
has_header = all(len(c) < self.HEADER_CELL_MAX_LEN for c in first_row)
|
||||
if has_header:
|
||||
headers = first_row
|
||||
data_rows_slice = zip(all_rows[1:], all_images[1:])
|
||||
else:
|
||||
headers = [f"列{ci + 1}" for ci in range(len(first_row))]
|
||||
data_rows_slice = zip(all_rows, all_images)
|
||||
|
||||
data_rows: list[dict] = []
|
||||
|
||||
for ri, (row_data, row_imgs) in enumerate(data_rows_slice):
|
||||
columns: list[dict] = []
|
||||
max_cols = max(len(headers), len(row_data))
|
||||
for ci in range(max_cols):
|
||||
hdr = headers[ci] if ci < len(headers) else ""
|
||||
txt = row_data[ci] if ci < len(row_data) else ""
|
||||
columns.append({
|
||||
"name": hdr,
|
||||
"row": ri + 1,
|
||||
"col": ci + 1,
|
||||
"text": txt,
|
||||
})
|
||||
|
||||
# Register image sources with structured location
|
||||
imgs = row_imgs[ci] if ci < len(row_imgs) else []
|
||||
for rid in imgs:
|
||||
image_sources[rid] = {
|
||||
"section": current_source,
|
||||
"table": tbl_idx,
|
||||
"row": ri + 1,
|
||||
"column": ci + 1,
|
||||
"name": hdr,
|
||||
}
|
||||
|
||||
data_rows.append({"columns": columns})
|
||||
|
||||
blocks.append({
|
||||
"type": "table",
|
||||
"table": tbl_idx,
|
||||
"headers": headers,
|
||||
"rows": data_rows,
|
||||
})
|
||||
elif all_rows:
|
||||
# Degenerate table (only header or single row) — treat as plain rows
|
||||
for ri, row_data in enumerate(all_rows):
|
||||
row_text = " | ".join(row_data)
|
||||
if row_text.strip():
|
||||
blocks.append({
|
||||
"type": "para",
|
||||
"index": para_idx + 1,
|
||||
"text": row_text,
|
||||
})
|
||||
para_idx += 1
|
||||
|
||||
if blocks or section_images:
|
||||
sections.append({
|
||||
"source": current_source,
|
||||
"blocks": blocks,
|
||||
"images": list(section_images),
|
||||
})
|
||||
|
||||
return sections, image_sources
|
||||
|
||||
# ---- internals ----------------------------------------------------------
|
||||
|
||||
def _heading_level(self, para: Paragraph) -> int | None:
|
||||
"""Heading level 1-9, or *None* if not a heading."""
|
||||
if para.style and para.style.name:
|
||||
name = para.style.name
|
||||
for prefix in ("Heading", "标题"):
|
||||
if name.startswith(prefix):
|
||||
try:
|
||||
return int(name.split()[-1])
|
||||
except (ValueError, IndexError):
|
||||
pass
|
||||
pPr = para._element.find(f"{{{self.WML_NS}}}pPr")
|
||||
if pPr is not None:
|
||||
ol = pPr.find(f"{{{self.WML_NS}}}outlineLvl")
|
||||
if ol is not None:
|
||||
val = ol.get(f"{{{self.WML_NS}}}val")
|
||||
if val is not None:
|
||||
try:
|
||||
return int(val) + 1
|
||||
except ValueError:
|
||||
pass
|
||||
return None
|
||||
|
||||
def _images_in(self, element) -> list[str]:
|
||||
"""Return rId values for drawings embedded in *element*."""
|
||||
rids: list[str] = []
|
||||
for drawing in element.findall(f".//{{{self.WML_NS}}}drawing"):
|
||||
blip = drawing.find(f".//{{{self.DRAW_NS}}}blip")
|
||||
if blip is not None:
|
||||
rid = blip.get(f"{{{self.REL_NS}}}embed")
|
||||
if rid:
|
||||
rids.append(rid)
|
||||
return rids
|
||||
@@ -0,0 +1,34 @@
|
||||
---
|
||||
name: ir_generation_skill
|
||||
description: 通过自然语言描述，向JSON文件中添加以IR表示的功能点。
|
||||
---
|
||||
|
||||
# add_ir
|
||||
|
||||
## 概述
|
||||
|
||||
通过自然语言描述由大语言模型自动转换为标准IR格式并写入到JSON文件或追加到JSON文件中。
|
||||
|
||||
## 用法
|
||||
|
||||
- **命令行传入**：直接传入功能的文字描述
|
||||
- **文件传入**：通过 `-f` 参数从文本文件读取描述
|
||||
- **追加模式**：新生成的功能点追加到已有 `_ir.json` 末尾，不影响已有条目
|
||||
|
||||
## 输入要求
|
||||
|
||||
- 功能的自然语言描述（命令行参数或 `-f` 文件）
|
||||
- 目标IR JSON文件路径（`--ir`，默认 `output/ir.json`）
|
||||
|
||||
## 输出
|
||||
|
||||
在目标IR JSON文件中追加新的功能点条目，每个条目包含：
|
||||
- `function`：功能名称
|
||||
- `source.section`：所属章节
|
||||
- `source.location`：位置描述
|
||||
- `trigger`：触发条件（可选）
|
||||
- `actions`：执行动作（可选）
|
||||
|
||||
## 使用提示
|
||||
|
||||
描述中应尽量包含功能的触发条件、执行动作和所属章节信息，以提高转换质量。
|
||||
@@ -0,0 +1,105 @@
|
||||
import logging
|
||||
import os
|
||||
import time
|
||||
from typing import Optional
|
||||
|
||||
from openai import OpenAI
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
class LLMClient:
|
||||
"""Low-level OpenAI-compatible LLM client with retry and token tracking.
|
||||
|
||||
Usage::
|
||||
|
||||
llm = LLMClient()
|
||||
content = llm.chat("qwen3.5-flash", [{"role": "user", "content": "Hello"}])
|
||||
print(llm.usage)
|
||||
"""
|
||||
|
||||
IMAGE_MODEL = "qwen3-vl-plus"
|
||||
TEXT_MODEL = "qwen3.5-flash-2026-02-23"
|
||||
TIMEOUT = 120
|
||||
MAX_RETRIES = 3
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
*,
|
||||
base_url: str = "https://dashscope.aliyuncs.com/compatible-mode/v1",
|
||||
timeout: int | None = None,
|
||||
):
|
||||
key = os.environ.get("DASHSCOPE_API_KEY", "")
|
||||
if not key:
|
||||
raise ValueError("DASHSCOPE_API_KEY environment variable is not set.")
|
||||
self._client = OpenAI(api_key=key, base_url=base_url)
|
||||
self._timeout = timeout or self.TIMEOUT
|
||||
self._prompt_tokens = 0
|
||||
self._completion_tokens = 0
|
||||
|
||||
@property
|
||||
def usage(self) -> dict:
|
||||
"""Return accumulated token counts as ``{prompt, completion, total}``."""
|
||||
return {
|
||||
"prompt_tokens": self._prompt_tokens,
|
||||
"completion_tokens": self._completion_tokens,
|
||||
"total_tokens": self._prompt_tokens + self._completion_tokens,
|
||||
}
|
||||
|
||||
@staticmethod
|
||||
def estimate_tokens(text: str) -> int:
|
||||
"""Quick token estimate. CJK ≈1.7/token, others ≈3.0/token."""
|
||||
cjk = sum(1 for c in text if '一' <= c <= '鿿' or ' ' <= c <= '〿')
|
||||
other = len(text) - cjk
|
||||
return max(1, int(cjk / 1.7 + other / 3.0))
|
||||
|
||||
@staticmethod
|
||||
def estimate_image_tokens() -> int:
|
||||
"""Fixed estimate for one vision-model image (~500 tokens)."""
|
||||
return 500
|
||||
|
||||
def chat(
|
||||
self, model: str, messages: list[dict], *, timeout: int | None = None,
|
||||
response_format: dict | None = None,
|
||||
) -> str:
|
||||
"""Send a chat completion request and return the response content.
|
||||
|
||||
Automatically retries on failure and accumulates token usage.
|
||||
"""
|
||||
label = f"chat({model})"
|
||||
|
||||
def _call():
|
||||
t0 = time.time()
|
||||
kwargs = dict(model=model, messages=messages, timeout=timeout or self._timeout)
|
||||
if response_format is not None:
|
||||
kwargs["response_format"] = response_format
|
||||
kwargs["temperature"] = 0
|
||||
resp = self._client.chat.completions.create(**kwargs)
|
||||
content = resp.choices[0].message.content
|
||||
usg = resp.usage
|
||||
if usg:
|
||||
self._prompt_tokens += usg.prompt_tokens
|
||||
self._completion_tokens += usg.completion_tokens
|
||||
elapsed = time.time() - t0
|
||||
logger.info("%s: %d chars in %.1fs", label, len(content) if content else 0, elapsed)
|
||||
if not content:
|
||||
raise RuntimeError("Empty response from LLM")
|
||||
return content
|
||||
|
||||
return self._retry(_call, label)
|
||||
|
||||
def _retry(self, fn, label: str) -> str:
|
||||
"""Call *fn()* with exponential-backoff retry."""
|
||||
last_error: Optional[Exception] = None
|
||||
for attempt in range(self.MAX_RETRIES):
|
||||
try:
|
||||
return fn()
|
||||
except Exception as e:
|
||||
last_error = e
|
||||
logger.warning(
|
||||
"%s error (attempt %d/%d): %s",
|
||||
label, attempt + 1, self.MAX_RETRIES, e,
|
||||
)
|
||||
if attempt < self.MAX_RETRIES - 1:
|
||||
time.sleep(2 ** attempt)
|
||||
raise RuntimeError(f"{label}: all retries exhausted") from last_error
|
||||
+191
@@ -0,0 +1,191 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Add missing features to an existing ``_ir.json``.
|
||||
|
||||
Converts a natural-language description of missing features into structured
|
||||
IR entries via LLM, then appends them to the IR file.
|
||||
|
||||
Usage::
|
||||
|
||||
python scripts/add_missing.py "遗漏功能描述" --ir output/<basename>_ir.json
|
||||
python scripts/add_missing.py -f missing.txt --ir output/<basename>_ir.json
|
||||
"""
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import logging
|
||||
import os
|
||||
import sys
|
||||
|
||||
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
|
||||
|
||||
from LLM import LLMClient
|
||||
|
||||
logging.basicConfig(
|
||||
level=logging.INFO,
|
||||
format="%(asctime)s [%(levelname)s] %(message)s",
|
||||
datefmt="%Y-%m-%d %H:%M:%S",
|
||||
)
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# Prompt
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
PROMPT = """你是一个需求文档分析助手。用户发现已生成的功能点列表中有遗漏，请将以下描述中提到的遗漏功能点转换为标准JSON格式。
|
||||
|
||||
## 遗漏功能描述
|
||||
|
||||
{description}
|
||||
|
||||
## JSON格式要求
|
||||
|
||||
每个功能点输出为：
|
||||
|
||||
{
|
||||
"function": "功能名称",
|
||||
"source": {
|
||||
"section": "章节名（如果描述中提到了章节信息则填写，否则填"未知"）",
|
||||
"location": "原文位置描述（如果描述中提供了则填写，否则填"用户补充"）"
|
||||
},
|
||||
"trigger": {
|
||||
"type": "AND或者OR",
|
||||
"conditions": [
|
||||
"触发条件1",
|
||||
"触发条件2"
|
||||
]
|
||||
},
|
||||
"actions": {
|
||||
"场景/角色": [
|
||||
"动作1",
|
||||
"动作2"
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
## 输出原则
|
||||
|
||||
1. 将描述中的每个独立功能点拆分为一个独立的JSON对象
|
||||
2. 如果描述中包含多个功能点，输出一个JSON数组
|
||||
3. 没有 trigger 或 actions 的字段直接**省略**，不要写 null 或空列表/空对象
|
||||
4. 尽量从描述中推断 source.section 和 source.location
|
||||
5. 功能名称应简洁明确，概括该功能的核心行为
|
||||
6. 直接输出纯JSON，不要用 ```json 代码块包裹
|
||||
"""
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# Helpers
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
def _parse_llm_response(raw: str) -> list | dict | str | None:
|
||||
"""Parse JSON from LLM response, handling markdown code fences."""
|
||||
if raw is None:
|
||||
return None
|
||||
stripped = raw.strip()
|
||||
if stripped.startswith("```"):
|
||||
nl = stripped.find("\n")
|
||||
stripped = stripped[nl + 1:] if nl != -1 else stripped[3:]
|
||||
if stripped.endswith("```"):
|
||||
stripped = stripped[:-3]
|
||||
try:
|
||||
return json.loads(stripped)
|
||||
except json.JSONDecodeError:
|
||||
logger.warning("Failed to parse JSON, returning raw text")
|
||||
return raw
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# Core
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
def add_missing(description: str, ir_path: str) -> list[dict]:
|
||||
"""Convert missing feature description to IR entries and append to IR file.
|
||||
|
||||
Returns the newly added entries.
|
||||
"""
|
||||
llm = LLMClient()
|
||||
|
||||
prompt = PROMPT.replace("{description}", description)
|
||||
|
||||
logger.info("Sending missing features to LLM...")
|
||||
try:
|
||||
raw = llm.chat(
|
||||
model=LLMClient.TEXT_MODEL,
|
||||
messages=[{"role": "user", "content": prompt}],
|
||||
response_format={"type": "json_object"},
|
||||
)
|
||||
logger.info("Response: %d chars", len(raw))
|
||||
except RuntimeError as e:
|
||||
logger.error("LLM call failed: %s", e)
|
||||
sys.exit(1)
|
||||
|
||||
parsed = _parse_llm_response(raw)
|
||||
if isinstance(parsed, list):
|
||||
new_entries = parsed
|
||||
elif isinstance(parsed, dict):
|
||||
new_entries = [parsed]
|
||||
else:
|
||||
logger.error("Unparseable response")
|
||||
sys.exit(1)
|
||||
|
||||
# Load existing IR or start fresh
|
||||
if os.path.exists(ir_path):
|
||||
with open(ir_path, "r", encoding="utf-8") as f:
|
||||
existing = json.load(f)
|
||||
logger.info("Loaded existing IR: %d entries", len(existing))
|
||||
else:
|
||||
existing = []
|
||||
logger.info("No existing IR file, creating new")
|
||||
|
||||
existing.extend(new_entries)
|
||||
|
||||
# Write back
|
||||
os.makedirs(os.path.dirname(ir_path) or ".", exist_ok=True)
|
||||
with open(ir_path, "w", encoding="utf-8") as f:
|
||||
json.dump(existing, f, ensure_ascii=False, indent=2)
|
||||
|
||||
logger.info("Saved: %s (%d entries, +%d new)", ir_path, len(existing), len(new_entries))
|
||||
|
||||
usg = llm.usage
|
||||
logger.info("Tokens: %d prompt + %d completion = %d total",
|
||||
usg["prompt_tokens"], usg["completion_tokens"], usg["total_tokens"])
|
||||
|
||||
return new_entries
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# CLI
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
if __name__ == "__main__":
|
||||
parser = argparse.ArgumentParser(
|
||||
description="Add missing features to an IR JSON file via LLM.",
|
||||
)
|
||||
parser.add_argument(
|
||||
"description", nargs="?", metavar="description",
|
||||
help="Natural-language description of missing features",
|
||||
)
|
||||
parser.add_argument(
|
||||
"-f", "--file", metavar="file.txt",
|
||||
help="Read description from a file instead of command line",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--ir", default="output/ir.json", metavar="ir.json",
|
||||
help="Path to the IR JSON file (default: output/ir.json)",
|
||||
)
|
||||
|
||||
args = parser.parse_args()
|
||||
|
||||
if args.file:
|
||||
with open(args.file, "r", encoding="utf-8") as f:
|
||||
description = f.read().strip()
|
||||
elif args.description:
|
||||
description = args.description
|
||||
else:
|
||||
parser.error("Either provide a description string or use -f to read from a file")
|
||||
|
||||
if not description:
|
||||
parser.error("Description is empty")
|
||||
|
||||
add_missing(description, args.ir)
|
||||
+42
@@ -0,0 +1,42 @@
|
||||
---
|
||||
name: 解决方案应用技能
|
||||
description: 应用用户提供的冲突解决方案以创建合并了更正的更新文档表示。
|
||||
---
|
||||
|
||||
# 解决方案应用技能
|
||||
|
||||
## 概述
|
||||
|
||||
此技能采用用户提供的冲突解决方案并将它们应用到解析的文档中，创建一个协调文本和视觉内容差异的校正版本。它生成包含应用更正的更新文档表示。
|
||||
|
||||
## 功能
|
||||
|
||||
该技能：
|
||||
- 接受用户提供的冲突解决方案决策
|
||||
- 支持多种解决方案类型："以图像为准"、"以文字为准"、"两处都保留"或自定义文本
|
||||
- 更新解析的文档结构以合并解决方案决策
|
||||
- 创建文档表示的校正版本，包含应用的更改
|
||||
- 维护所有应用更正的源可追溯性
|
||||
- 向输出添加包含更正指令的resolved_conflicts数组
|
||||
|
||||
## 输入要求
|
||||
|
||||
- 解析文档JSON文件的路径（带有已识别的冲突）
|
||||
- 包含用户决策的解决方案JSON文件的路径
|
||||
- 可选输出目录规范
|
||||
- 解决方案JSON应包含具有以下内容的对象：
|
||||
- `conflict_id`：冲突数组中的冲突索引
|
||||
- `resolution`：决策类型（"以图片为准"、"以文字为准"、"两处都保留"）或自定义文本
|
||||
- `custom_text`：解决方案的可选自定义文本
|
||||
|
||||
## 输出
|
||||
|
||||
该技能生成一个结构化JSON文件，文件名为输入文档的基本名称后跟'_updated.json'，包含：
|
||||
- 包含应用更正的原始文档结构
|
||||
- 详细说明应用更改的`resolved_conflicts`数组
|
||||
- 关于每个冲突类型和应用更正的信息
|
||||
- 用于可追溯性的源跟踪信息
|
||||
|
||||
## 集成点
|
||||
|
||||
此技能消耗冲突检测技能的输出（带冲突的文档）和用户提供的解决方案。其输出被IR生成技能使用以创建最终结构化表示。
|
||||
Executable
+144
@@ -0,0 +1,144 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Apply user resolutions to ``_parsed.json`` using ``_conflicts.json``.
|
||||
|
||||
Usage::
|
||||
|
||||
python scripts/apply_resolutions.py <parsed.json> --resolutions <resolutions.json> [--output-dir DIR]
|
||||
|
||||
The *resolutions.json* file is created by the agent after user arbitration.
|
||||
Each resolution maps a conflict_id to a decision.
|
||||
|
||||
Resolution format (``resolutions.json``)::
|
||||
|
||||
[
|
||||
{
|
||||
"conflict_id": 0, // 0-based index into conflicts array
|
||||
"resolution": "以文字为准",
|
||||
"custom_text": null
|
||||
}
|
||||
]
|
||||
|
||||
Outputs ``<basename>_updated.json`` — identical to *parsed.json* plus a
|
||||
``resolved_conflicts`` top-level array with correction instructions for the IR generator.
|
||||
"""
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import logging
|
||||
import os
|
||||
|
||||
logging.basicConfig(
|
||||
level=logging.INFO,
|
||||
format="%(asctime)s [%(levelname)s] %(message)s",
|
||||
datefmt="%Y-%m-%d %H:%M:%S",
|
||||
)
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
def apply_resolutions(
|
||||
parsed_path: str,
|
||||
resolutions_path: str,
|
||||
output_dir: str | None = None,
|
||||
) -> dict:
|
||||
"""Load *parsed.json*, apply resolutions, write *updated.json*."""
|
||||
with open(parsed_path, "r", encoding="utf-8") as f:
|
||||
data = json.load(f)
|
||||
|
||||
with open(resolutions_path, "r", encoding="utf-8") as f:
|
||||
resolutions = json.load(f)
|
||||
|
||||
base_dir = os.path.dirname(os.path.abspath(parsed_path))
|
||||
|
||||
# Try to find _conflicts.json alongside parsed.json
|
||||
basename = os.path.splitext(os.path.basename(parsed_path))[0]
|
||||
stem = basename[:-7] if basename.endswith("_parsed") else basename
|
||||
candidate = os.path.join(base_dir, f"{stem}_conflicts.json")
|
||||
|
||||
conflicts = data.get("_conflicts", [])
|
||||
if not conflicts and os.path.isfile(candidate):
|
||||
with open(candidate, "r", encoding="utf-8") as f:
|
||||
conflicts = json.load(f)
|
||||
|
||||
if output_dir is None:
|
||||
output_dir = base_dir
|
||||
os.makedirs(output_dir, exist_ok=True)
|
||||
|
||||
# Build resolved_conflicts with correction instructions for ir_generator
|
||||
resolved = []
|
||||
for res in resolutions:
|
||||
cid = res.get("conflict_id")
|
||||
if cid is None or cid < 0 or cid >= len(conflicts):
|
||||
logger.warning("Invalid conflict_id: %s", cid)
|
||||
continue
|
||||
|
||||
conflict = conflicts[cid]
|
||||
choice = res.get("resolution", "")
|
||||
custom = res.get("custom_text")
|
||||
|
||||
entry = {
|
||||
"conflict_id": cid,
|
||||
"conflict_type": conflict.get("conflict_type"),
|
||||
"section": conflict.get("section", ""),
|
||||
"resolution": choice,
|
||||
}
|
||||
|
||||
# Build a correction instruction string
|
||||
image_val = conflict.get("image_snippet", "")
|
||||
text_val = conflict.get("text_snippet", "")
|
||||
if choice == "以图片为准":
|
||||
entry["correction"] = image_val
|
||||
entry["source"] = "图片"
|
||||
elif choice == "以文字为准":
|
||||
entry["correction"] = text_val
|
||||
entry["source"] = "文字"
|
||||
elif choice == "两处都保留":
|
||||
entry["correction"] = f"{text_val}（另外的观点：{image_val}）"
|
||||
entry["source"] = "两者兼容"
|
||||
elif custom:
|
||||
entry["correction"] = custom
|
||||
entry["source"] = "自定义"
|
||||
logger.info("Conflict %d: custom: %s", cid, custom[:60])
|
||||
else:
|
||||
entry["correction"] = text_val
|
||||
entry["source"] = "文字（默认）"
|
||||
logger.warning("Conflict %d: unknown resolution '%s', defaulting to text", cid, choice)
|
||||
|
||||
logger.info("Conflict %d: %s → %s", cid, choice, entry["source"])
|
||||
|
||||
resolved.append(entry)
|
||||
|
||||
data["resolved_conflicts"] = resolved
|
||||
logger.info("Applied %d resolutions", len(resolved))
|
||||
|
||||
# Write output
|
||||
if basename.endswith("_parsed"):
|
||||
out_name = f"{stem}_updated.json"
|
||||
else:
|
||||
out_name = f"{basename}_updated.json"
|
||||
|
||||
output_path = os.path.join(output_dir, out_name)
|
||||
with open(output_path, "w", encoding="utf-8") as f:
|
||||
json.dump(data, f, ensure_ascii=False, indent=2)
|
||||
|
||||
logger.info("Saved: %s", output_path)
|
||||
return data
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# CLI
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
if __name__ == "__main__":
|
||||
parser = argparse.ArgumentParser(
|
||||
description="Apply user resolutions to parsed.json.",
|
||||
)
|
||||
parser.add_argument("input", metavar="parsed.json",
|
||||
help="Path to _parsed.json")
|
||||
parser.add_argument("--resolutions", "-r", required=True,
|
||||
help="Path to resolutions JSON file")
|
||||
parser.add_argument("--output-dir", default=None,
|
||||
help="Output directory (default: same as input)")
|
||||
|
||||
args = parser.parse_args()
|
||||
apply_resolutions(args.input, args.resolutions, args.output_dir)
|
||||
+207
@@ -0,0 +1,207 @@
|
||||
# 任务：重写 IR Generation，实现高可靠性、可测试的中间表示提取
|
||||
|
||||
## 一、背景与目标
|
||||
|
||||
我们正在为吉利汽车车机测试团队验证 AI 自动生成测试用例的可行性。输入是产品需求文档（PRD），输出是 Robot Framework 测试脚本。
|
||||
|
||||
当前流程中，文档解析（doc_parser）已稳定，能将 Word 文档转为结构化 JSON，并将流程图图片转为精确的 `logic_tree` JSON。但直接基于这份 JSON 生成 IR（中间表示）时，LLM 经常丢失分支，稳定性差。
|
||||
|
||||
现在需要重新实现 IR 生成，采用“分阶段理解 + 程序化校验”架构，彻底解决分支丢失问题。
|
||||
|
||||
**核心原则**：LLM 负责“关联理解”和“受限翻译”，程序负责“确定性校验”和“完整性审计”。
|
||||
|
||||
## 二、已有资产与路径
|
||||
|
||||
- **输入文件**：`C:\Users\peterz\.openclaw\workspace\skills\doc_parser_skill\output\车机娱乐系统禁止功能文档_脱敏 v0.9_v2_updated.json`
|
||||
- 结构：包含 `sections` 数组和 `image_analysis` 数组。
|
||||
- `sections` 每个元素有 `source`（章节标题）、`blocks`（`para` 和 `table`）、`images`（图片引用 ID 列表）。
|
||||
- `image_analysis` 中，流程图类型的图片包含 `logic_tree`，是一个有根、有节点、有分支的树形结构，分支准确。
|
||||
|
||||
- **工作目录**：`C:\Users\peterz\.openclaw\workspace\skills\ir_generation_new_skill`
|
||||
- 请在此目录下创建 Python 脚本、Prompt 模板、测试等。
|
||||
|
||||
- **最终输出文件**：放在 `C:\Users\peterz\.openclaw\workspace\skills\doc_parser_skill\output\`
|
||||
- `ir_final.json`：最终的 IR JSON。
|
||||
- `ir_audit_report.md`：完整性审计报告，供人工快速审查。
|
||||
|
||||
## 三、IR Schema 定义
|
||||
|
||||
最终 IR 是一组**可测试的功能规则**，每条规则包含触发条件、动作、来源锚点，足以让规则引擎自动生成覆盖正常/边界/异常的 Robot Framework 测试用例。
|
||||
|
||||
**目标 Schema（示例）**：
|
||||
|
||||
```json
|
||||
{
|
||||
"feature": "行车娱乐限制",
|
||||
"feature_id": "DRL-001",
|
||||
"rules": [
|
||||
{
|
||||
"rule_id": "DRL-001-SYS-FG-01",
|
||||
"description": "开关开启，系统限制应用在前台，车速≥15km/h 且持续超过5秒且非P档时，应用被打断并退至后台，同时弹出特定 Toast",
|
||||
"priority": "P0",
|
||||
"sources": [
|
||||
{"type": "table", "section": "3.1.1", "row": 2},
|
||||
{"type": "logic_tree", "image_id": "rId16", "node_ids": ["n19","n21","n23","n25","n26"]}
|
||||
],
|
||||
"precondition": {
|
||||
"switch": "开启",
|
||||
"app_type": "系统限制",
|
||||
"app_state": "前台"
|
||||
},
|
||||
"trigger": {
|
||||
"operator": "AND",
|
||||
"conditions": [
|
||||
{"signal": "车速", "operator": ">=", "value": 15, "unit": "km/h"},
|
||||
{"signal": "车速_持续时间", "operator": ">", "value": 5, "unit": "秒"},
|
||||
{"signal": "档位", "operator": "!=", "value": "P"}
|
||||
]
|
||||
},
|
||||
"actions": [
|
||||
{"type": "system", "description": "打断应用前台进程"},
|
||||
{"type": "system", "description": "将应用调入后台"},
|
||||
{"type": "user_interaction", "description": "显示Toast", "content": "在行车状态下无法使用该应用"}
|
||||
]
|
||||
}
|
||||
]
|
||||
}
|
||||
注：具体字段可根据实际文档内容灵活扩充，但必须包含 rule_id、description、sources、trigger、actions。
|
||||
|
||||
## 四、新实现流程（分三个阶段）
|
||||
|
||||
阶段一：宏观语义索引（Semantic Index）
|
||||
目标：让 LLM 一次性阅读整个文档，生成一份语义索引，识别出所有功能单元（function unit）并建立概念映射，但不提取具体规则细节。
|
||||
|
||||
输入：完整的 xxx_parsed.json（所有 sections + image_analysis）。
|
||||
|
||||
LLM 调用：一次。
|
||||
|
||||
任务要求：
|
||||
|
||||
编写一个 Python 脚本 step1_semantic_index.py，其功能：
|
||||
|
||||
读取输入 JSON 文件。
|
||||
构造 Prompt，要求 LLM 输出如下结构的 JSON：
|
||||
json
|
||||
{
|
||||
"feature_name": "行车娱乐限制",
|
||||
"concepts": [
|
||||
{"name": "行车娱乐限制", "aliases": ["行车娱乐限制", "行车娱乐禁止"], "defined_in": ["3.1", "3.1.1"]},
|
||||
{"name": "系统限制", "aliases": [], "defined_in": ["3.1", "3.1.1"]},
|
||||
...
|
||||
],
|
||||
"function_units": [
|
||||
{
|
||||
"unit_id": "FU-001",
|
||||
"name": "系统限制-前台-行车打断",
|
||||
"description": "当开关开启、应用为系统限制类型且处于前台时，满足车速和档位条件后，系统打断应用并显示Toast",
|
||||
"sources": [
|
||||
{"section": "3.1.1", "type": "table", "row": 2, "text_snippet": "打断：车速≥15km/h...退至后台"},
|
||||
{"image_id": "rId16", "logic_tree_nodes": ["n19","n21","n23","n25","n26"]},
|
||||
{"image_id": "rId17", "logic_tree_nodes": ["n1","n2","n3"]}
|
||||
]
|
||||
},
|
||||
...
|
||||
]
|
||||
}
|
||||
|
||||
Prompt 中必须强调：功能单元应覆盖文档描述的所有主要行为，特别是图片逻辑树中的决策路径。不允许遗漏分支。
|
||||
调用 LLM（可以使用 anthropic.Anthropic 或内部 API），将结果保存为 semantic_index.json。
|
||||
自检测试：编写 test_step1.py，读取 semantic_index.json 并验证：
|
||||
所有 function_units 的 sources 中引用的 image_id 必须存在于原输入的 image_analysis 中。
|
||||
每个 function_unit 至少引用一张图片或一段文字。
|
||||
所有 logic_tree_nodes 引用的节点 ID 必须在对应 logic_tree 的节点 id 中存在。
|
||||
无空的 unit_id 或 name。
|
||||
迭代指引：如果测试失败，分析原因，调整 Prompt（增加更严格的结构化输出指令、Few-shot 示例）并重新运行，直到全部通过。
|
||||
|
||||
## 阶段二：逐功能单元 IR 提取
|
||||
目标：对阶段一产出的每个功能单元，准备一个仅含相关上下文的数据包，让 LLM 据此填充详细的 IR 规则。
|
||||
|
||||
输入：semantic_index.json 和原始文档 JSON。
|
||||
|
||||
LLM 调用：每个功能单元一次，可并行。
|
||||
|
||||
任务要求：
|
||||
|
||||
编写 step2_ir_extraction.py：
|
||||
|
||||
加载 semantic_index.json 和原始文档 JSON。
|
||||
为每个 function_unit 构造一个“精准上下文包”：
|
||||
从原始文档中提取该单元引用的具体段落、表格行、以及完整的 logic_tree（如果引用了节点，则提取相关子树；简单起见可以提取整棵树）。
|
||||
上下文包示例结构：
|
||||
json
|
||||
{
|
||||
"unit_id": "FU-001",
|
||||
"unit_name": "系统限制-前台-行车打断",
|
||||
"texts": ["打断：车速≥15km/h且持续5秒后，将目标应用/功能退至后台或暂停对应功能..."],
|
||||
"tables": [{"headers":["功能","功能详细说明"], "rows":[...]}],
|
||||
"logic_trees": [
|
||||
{"image_id": "rId16", "tree": {...}},
|
||||
{"image_id": "rId17", "tree": {...}}
|
||||
]
|
||||
}
|
||||
构造 Prompt，要求 LLM 输出一个或多个符合 IR Schema 的规则 JSON 对象（数组 rules）。
|
||||
强制要求：
|
||||
每条规则的 sources 必须包含引用的逻辑树节点 ID 列表和文本来源。
|
||||
触发条件必须精确，包含运算符和数值（如 >=15），不可模糊。
|
||||
动作必须明确区分系统行为和用户可见交互。
|
||||
如果文档中存在矛盾（如图片和文字冲突），请优先采用图片逻辑树，并在 description 中注明差异。
|
||||
提供 IR Schema 和 1-2 个 Few-shot 示例。
|
||||
将所有功能单元产出的规则数组合并，保存为 ir_fragments.json，格式：[ { "unit_id": "FU-001", "rules": [...] }, ... ]
|
||||
自检测试：编写 test_step2.py：
|
||||
验证每个 ir_fragment 的 rules 非空。
|
||||
验证每条规则的 sources 中有逻辑树节点引用。
|
||||
验证 trigger.conditions 中的每个条件均有 signal、operator、value。
|
||||
检查是否有重复的 rule_id（可暂时用 unit_id + 序号，但最终合并时会处理）。
|
||||
迭代指引：如果某些片段规则为空或条件缺失，检查是否为上下文包裁剪太狠（丢失了关键文本），优化提取逻辑或调整 Prompt，再次运行。
|
||||
|
||||
阶段三：确定性合并与完整性校验
|
||||
目标：用程序逻辑将多个片段的规则去重、合并，并基于逻辑树节点覆盖率生成审计报告，告知人工哪里可能遗漏。
|
||||
|
||||
LLM 调用：无（或用极简 LLM 做最终文案规范化）。
|
||||
|
||||
任务要求：
|
||||
|
||||
编写 step3_merge_and_audit.py：
|
||||
|
||||
加载 ir_fragments.json 和原始文档 JSON（含所有 logic_tree 节点）。
|
||||
合并去重：
|
||||
按 trigger 和 actions 的语义相似度进行规则合并。简化方案：如果两条规则的触发条件和动作完全相同（值比较），只保留一条，合并 sources。
|
||||
合并后重新分配稳定的 rule_id（如 DRL-001-SYS-FG-01）。
|
||||
完整性审计（生成 ir_audit_report.md）：
|
||||
逻辑树节点覆盖率：遍历所有 logic_tree 中的每个 decision 和 action 节点，检查是否有至少一条规则的 sources 引用了该节点。列出未被覆盖的节点及其所在的图片 ID。
|
||||
表格枚举覆盖：识别所有表格中列出“应用类型”、“限制方式”的枚举值，检查这些值是否出现在规则的 precondition 或条件中。
|
||||
全局开关覆盖：检查所有涉及开关状态的规则，是否完整覆盖了“开启”和“关闭”两种状态。
|
||||
报告格式：Markdown 表格，列出检查项、状态（✅/⚠️/❌）、详情。
|
||||
合并后的最终 IR 保存为 ir_final.json。
|
||||
人类检查提示：审计报告最上方用醒目的文字说明：“请人工审查以下 ⚠️ 和 ❌ 项，确认是文档遗漏还是 IR 提取遗漏。如无需修改，则在对应项后标注‘已确认’。”
|
||||
编写 test_step3.py，验证：
|
||||
|
||||
ir_final.json 中无重复 rule_id。
|
||||
|
||||
所有 rule_id 符合命名规范（DRL-001-...）。
|
||||
|
||||
ir_audit_report.md 文件存在且包含覆盖率统计。
|
||||
|
||||
迭代指引：如果审计报告显示大量节点未覆盖，需回溯检查阶段二的提取质量，可能是某些功能单元未被识别，或提取时丢失了节点引用。可补充遗漏的单元，重新提取片段后再次合并审计。
|
||||
|
||||
## 五、总体执行顺序与依赖
|
||||
首先实现并运行 step1_semantic_index.py，通过测试。
|
||||
|
||||
接着运行 step2_ir_extraction.py，通过测试。
|
||||
|
||||
最后运行 step3_merge_and_audit.py，生成最终交付物。
|
||||
|
||||
所有脚本应可重复执行，且运行 python main.py（如有）可一键完成所有步骤。
|
||||
|
||||
六、开发与协作要求
|
||||
所有 Python 代码需有清晰的注释，关键函数有 docstring。
|
||||
|
||||
Prompt 模板可以单独作为 .txt 或 .md 文件存放，便于调试。
|
||||
|
||||
测试脚本独立，并输出明确的通过/失败信息。
|
||||
|
||||
如果某个阶段需要人类帮助（例如需要确认某个功能划分是否合理），请在该阶段输出清晰的问题并暂停，等待人类回复后继续。
|
||||
|
||||
你将在工作目录 C:\Users\peterz\.openclaw\workspace\skills\ir_generation_new_skill 中完成所有工作。
|
||||
|
||||
现在开始执行。首先请通读输入文件（路径已给出），理解其结构，然后着手实现阶段一。
|
||||
Executable
+121
@@ -0,0 +1,121 @@
|
||||
"""
|
||||
Shared configuration for the IR Generation pipeline.
|
||||
Reads API keys from a secrets.yaml file, falling back to environment variables.
|
||||
"""
|
||||
|
||||
import os
|
||||
import json
|
||||
import yaml
|
||||
|
||||
# ---- Paths ----
|
||||
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
|
||||
WORKSPACE_DIR = os.path.dirname(BASE_DIR)
|
||||
DOC_PARSER_OUTPUT = os.path.join(WORKSPACE_DIR, "doc_parser_skill", "output")
|
||||
PROMPTS_DIR = os.path.join(BASE_DIR, "prompts")
|
||||
TESTS_DIR = os.path.join(BASE_DIR, "tests")
|
||||
OUTPUT_DIR = os.path.join(BASE_DIR, "output")
|
||||
|
||||
# Input file (the parsed PRD JSON)
|
||||
_DEFAULT_INPUT = os.path.join(
|
||||
DOC_PARSER_OUTPUT,
|
||||
"车机娱乐系统禁止功能文档_脱敏 v0.9_v2_updated.json",
|
||||
)
|
||||
INPUT_JSON = os.environ.get("IR_INPUT_JSON", _DEFAULT_INPUT)
|
||||
|
||||
|
||||
def set_input_file(path: str) -> None:
|
||||
"""Override the default input JSON path."""
|
||||
global INPUT_JSON
|
||||
INPUT_JSON = path
|
||||
|
||||
# Secrets file (shared with workspace-document-analyzer)
|
||||
# .openclaw/workspace/skills/ir_generation_new_skill -> .openclaw/workspace-document-analyzer
|
||||
OPENCLAW_HOME = os.path.dirname(os.path.dirname(WORKSPACE_DIR))
|
||||
SECRETS_YAML = os.path.join(
|
||||
OPENCLAW_HOME, "workspace-document-analyzer", "config", "secrets.yaml",
|
||||
)
|
||||
|
||||
# Intermediate outputs
|
||||
SEMANTIC_INDEX_JSON = os.path.join(OUTPUT_DIR, "semantic_index.json")
|
||||
IR_FRAGMENTS_JSON = os.path.join(OUTPUT_DIR, "ir_fragments.json")
|
||||
|
||||
# Final deliverables (placed in doc_parser output per spec)
|
||||
IR_FINAL_JSON = os.path.join(DOC_PARSER_OUTPUT, "ir_final.json")
|
||||
IR_AUDIT_REPORT_MD = os.path.join(DOC_PARSER_OUTPUT, "ir_audit_report.md")
|
||||
|
||||
# ---- LLM API ----
|
||||
# Choose provider: "deepseek" | "dashscope"
|
||||
LLM_PROVIDER = os.environ.get("IR_PROVIDER", "deepseek")
|
||||
|
||||
# Model names per provider
|
||||
PROVIDER_MODELS = {
|
||||
"deepseek": os.environ.get("IR_MODEL", "deepseek-v4-pro"),
|
||||
"dashscope": os.environ.get("IR_MODEL", "qwen-max"),
|
||||
}
|
||||
MODEL_NAME = PROVIDER_MODELS.get(LLM_PROVIDER, PROVIDER_MODELS["deepseek"])
|
||||
|
||||
# Maximum tokens for LLM responses
|
||||
MAX_TOKENS = int(os.environ.get("IR_MAX_TOKENS", "16000"))
|
||||
TEMPERATURE = float(os.environ.get("IR_TEMPERATURE", "0.1"))
|
||||
|
||||
|
||||
def _load_secrets() -> dict[str, dict[str, str]]:
|
||||
"""Load provider credentials from secrets.yaml.
|
||||
|
||||
Returns a dict like: {"deepseek": {"apiKey": "...", "baseUrl": "..."}, ...}
|
||||
"""
|
||||
if os.path.isfile(SECRETS_YAML):
|
||||
with open(SECRETS_YAML, "r", encoding="utf-8") as f:
|
||||
return yaml.safe_load(f) or {}
|
||||
return {}
|
||||
|
||||
|
||||
def _get_provider_config(provider: str) -> dict[str, str]:
|
||||
"""Get {apiKey, baseUrl} for a provider from secrets, with env-var fallback."""
|
||||
secrets = _load_secrets()
|
||||
entry = secrets.get(provider, {})
|
||||
|
||||
env_prefix = provider.upper()
|
||||
api_key = (
|
||||
os.environ.get(f"{env_prefix}_API_KEY")
|
||||
or entry.get("apiKey", "")
|
||||
)
|
||||
base_url = (
|
||||
os.environ.get(f"{env_prefix}_BASE_URL")
|
||||
or entry.get("baseUrl", "https://api.deepseek.com/v1")
|
||||
)
|
||||
|
||||
if not api_key:
|
||||
raise RuntimeError(
|
||||
f"No API key found for provider '{provider}'. "
|
||||
f"Check {SECRETS_YAML} or set {env_prefix}_API_KEY."
|
||||
)
|
||||
return {"apiKey": api_key, "baseUrl": base_url}
|
||||
|
||||
|
||||
def llm_client():
|
||||
"""Return an OpenAI-compatible client configured from secrets.yaml."""
|
||||
from openai import OpenAI
|
||||
|
||||
cfg = _get_provider_config(LLM_PROVIDER)
|
||||
return OpenAI(base_url=cfg["baseUrl"], api_key=cfg["apiKey"])
|
||||
|
||||
|
||||
def load_input_document(path: str | None = None) -> dict:
|
||||
"""Load the parsed PRD JSON document."""
|
||||
path = path or INPUT_JSON
|
||||
with open(path, "r", encoding="utf-8") as f:
|
||||
return json.load(f)
|
||||
|
||||
|
||||
def save_json(data, path: str) -> None:
|
||||
"""Save data as formatted JSON."""
|
||||
os.makedirs(os.path.dirname(path), exist_ok=True)
|
||||
with open(path, "w", encoding="utf-8") as f:
|
||||
json.dump(data, f, ensure_ascii=False, indent=2)
|
||||
|
||||
|
||||
def load_json(path: str) -> dict:
|
||||
"""Load a JSON file."""
|
||||
with open(path, "r", encoding="utf-8") as f:
|
||||
return json.load(f)
|
||||
Executable
+134
@@ -0,0 +1,134 @@
|
||||
"""
|
||||
IR Generation Pipeline Orchestrator.
|
||||
|
||||
Run all three stages sequentially:
|
||||
python main.py [--skip-step1] [--skip-step2] [--skip-step3] [--test-only]
|
||||
|
||||
The pipeline reads the parsed PRD JSON from doc_parser and produces:
|
||||
- ir_final.json: the final IR rules
|
||||
- ir_audit_report.md: completeness audit report for human review
|
||||
"""
|
||||
|
||||
import argparse
|
||||
import os
|
||||
import subprocess
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
BASE_DIR = Path(__file__).parent
|
||||
|
||||
|
||||
def _subprocess_env(extra: dict | None = None) -> dict:
|
||||
"""Build environment dict for subprocesses, carrying forward overrides."""
|
||||
env = os.environ.copy()
|
||||
env.update(extra or {})
|
||||
return env
|
||||
|
||||
|
||||
def run_step(script_name: str, description: str, extra_env: dict | None = None) -> bool:
|
||||
"""Run a single pipeline step script, return True if it succeeded."""
|
||||
print(f"\n{'#' * 60}")
|
||||
print(f"# {description}")
|
||||
print(f"{'#' * 60}")
|
||||
script_path = BASE_DIR / script_name
|
||||
if not script_path.exists():
|
||||
print(f"错误: 脚本不存在 {script_path}")
|
||||
return False
|
||||
result = subprocess.run(
|
||||
[sys.executable, str(script_path)],
|
||||
cwd=str(BASE_DIR),
|
||||
env=_subprocess_env(extra_env),
|
||||
)
|
||||
return result.returncode == 0
|
||||
|
||||
|
||||
def run_test(test_name: str, description: str, extra_env: dict | None = None) -> bool:
|
||||
"""Run a test script, return True if all tests passed."""
|
||||
print(f"\n{'='*60}")
|
||||
print(f"测试: {description}")
|
||||
print(f"{'='*60}")
|
||||
test_path = BASE_DIR / "tests" / test_name
|
||||
if not test_path.exists():
|
||||
print(f"错误: 测试脚本不存在 {test_path}")
|
||||
return False
|
||||
result = subprocess.run(
|
||||
[sys.executable, str(test_path)],
|
||||
cwd=str(BASE_DIR),
|
||||
env=_subprocess_env(extra_env),
|
||||
)
|
||||
return result.returncode == 0
|
||||
|
||||
|
||||
def main():
|
||||
parser = argparse.ArgumentParser(description="IR Generation Pipeline")
|
||||
parser.add_argument("--skip-step1", action="store_true", help="跳过阶段一（语义索引）")
|
||||
parser.add_argument("--skip-step2", action="store_true", help="跳过阶段二（IR 提取）")
|
||||
parser.add_argument("--skip-step3", action="store_true", help="跳过阶段三（合并与审计）")
|
||||
parser.add_argument("--test-only", action="store_true", help="仅运行测试，不调用 LLM")
|
||||
parser.add_argument(
|
||||
"--input", "-i", type=str, default=None,
|
||||
help="输入 JSON 文件路径（覆盖默认的 doc_parser 输出）"
|
||||
)
|
||||
parser.add_argument(
|
||||
"--provider", "-p", type=str, default=None,
|
||||
help="LLM provider: deepseek | dashscope（覆盖 IR_PROVIDER 环境变量）"
|
||||
)
|
||||
args = parser.parse_args()
|
||||
|
||||
# Build extra env vars for subprocesses
|
||||
extra_env = {}
|
||||
if args.input:
|
||||
extra_env["IR_INPUT_JSON"] = args.input
|
||||
print(f"输入文件: {args.input}")
|
||||
if args.provider:
|
||||
extra_env["IR_PROVIDER"] = args.provider
|
||||
print(f"LLM Provider: {args.provider}")
|
||||
|
||||
if args.test_only:
|
||||
all_ok = True
|
||||
all_ok &= run_test("test_step1.py", "Step 1 验证", extra_env)
|
||||
all_ok &= run_test("test_step2.py", "Step 2 验证", extra_env)
|
||||
all_ok &= run_test("test_step3.py", "Step 3 验证", extra_env)
|
||||
sys.exit(0 if all_ok else 1)
|
||||
|
||||
failures = []
|
||||
|
||||
# Stage 1
|
||||
if not args.skip_step1:
|
||||
ok = run_step("step1_semantic_index.py", "阶段一：宏观语义索引", extra_env)
|
||||
if not ok:
|
||||
failures.append("阶段一")
|
||||
print("\n阶段一失败，停止流水线。修复后重试。")
|
||||
sys.exit(1)
|
||||
run_test("test_step1.py", "Step 1 验证", extra_env)
|
||||
|
||||
# Stage 2
|
||||
if not args.skip_step2:
|
||||
ok = run_step("step2_ir_extraction.py", "阶段二：逐功能单元 IR 提取", extra_env)
|
||||
if not ok:
|
||||
failures.append("阶段二")
|
||||
print("\n阶段二失败，停止流水线。修复后重试。")
|
||||
sys.exit(1)
|
||||
run_test("test_step2.py", "Step 2 验证", extra_env)
|
||||
|
||||
# Stage 3
|
||||
if not args.skip_step3:
|
||||
ok = run_step("step3_merge_and_audit.py", "阶段三：确定性合并与完整性校验", extra_env)
|
||||
if not ok:
|
||||
failures.append("阶段三")
|
||||
sys.exit(1)
|
||||
run_test("test_step3.py", "Step 3 验证", extra_env)
|
||||
|
||||
if failures:
|
||||
print(f"\n失败阶段: {', '.join(failures)}")
|
||||
sys.exit(1)
|
||||
|
||||
print(f"\n{'='*60}")
|
||||
print("流水线全部完成!")
|
||||
print(f"最终 IR: config.IR_FINAL_JSON")
|
||||
print(f"审计报告: config.IR_AUDIT_REPORT_MD")
|
||||
print(f"{'='*60}")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@@ -0,0 +1,69 @@
|
||||
你是吉利汽车车机系统（XX Auto）的产品需求分析师。你的任务是从行车娱乐限制功能 PRD 文档中提取"语义索引"——一份结构化的功能清单，而不是逐字翻译。
|
||||
|
||||
## 文档结构说明
|
||||
|
||||
下面是一份 Word 文档的解析结果，包含：
|
||||
|
||||
1. **sections**：按章节组织的混合内容（段落 + 表格），每个 section 有 `source`（章节标题）、`blocks`（`para` 文本段落和 `table` 结构表格）、`images`（引用的图片 ID 列表）
|
||||
2. **image_analysis**：文档中流程图的程序化分析结果，其中 `logic_tree` 是由节点组成的决策树：
|
||||
- `state` 节点：状态说明
|
||||
- `decision` 节点：判断条件 + `branches`（分支值 → 目标节点 ID）
|
||||
- `action` 节点：系统或用户交互动作
|
||||
3. **resolved_conflicts**：文档中图文冲突的仲裁结果，明确指出应以文字还是图片为准
|
||||
|
||||
## 文档全文
|
||||
|
||||
{document_json}
|
||||
|
||||
## 你的任务
|
||||
|
||||
阅读整份文档后，输出一份 **语义索引 JSON**，包含：
|
||||
|
||||
### 1. feature_name
|
||||
功能名称，如"行车娱乐限制"
|
||||
|
||||
### 2. concepts
|
||||
文档中定义或使用的关键概念列表。每个概念包含：
|
||||
- `name`：概念的标准名称
|
||||
- `aliases`：同义词/别名列表（如"行车娱乐限制"、"行车娱乐禁止"）
|
||||
- `defined_in`：定义该概念的章节号列表（如 ["3.1", "3.1.1"]）
|
||||
|
||||
应识别的概念类型包括但不限于：功能名称、应用类型（系统限制、SDK限制、其他应用）、限制方式（打断、禁止、暂停）、触发条件（车速、档位、持续时间）、开关状态等。
|
||||
|
||||
### 3. function_units
|
||||
文档中描述的所有主要功能行为的列表。**每个 function_unit 对应一条完整的"如果...则..."规则**。每个 function unit 包含：
|
||||
|
||||
- `unit_id`：唯一标识，格式 "FU-001", "FU-002"...
|
||||
- `name`：简短名称，如"系统限制-前台-行车打断"
|
||||
- `description`：1-3 句描述该规则的行为
|
||||
- `sources`：该规则在文档中的来源锚点列表，每项包含：
|
||||
- `section`：章节号
|
||||
- `type`：来源类型，`"table"` 或 `"para"` 或 `"logic_tree"`
|
||||
- `row`：如果是表格行（从 1 开始）
|
||||
- `text_snippet`：前 200 字的关键文字
|
||||
- `image_id`：如果是逻辑树来源，填写图片 rId
|
||||
- `logic_tree_nodes`：如果是逻辑树来源，列出相关节点 ID 列表
|
||||
|
||||
## 关键要求
|
||||
|
||||
1. **必须覆盖所有逻辑树分支**：遍历每个 `logic_tree` 中从根到叶的每条决策路径，确保它们都出现在某个 function_unit 中。逻辑树中的每个 `decision` 节点及其分支、每个 `action` 节点都必须被至少一个 function_unit 引用。
|
||||
|
||||
2. **必须覆盖表格中的所有规则**：表格中列出的每种"限制方法"、"限制规则"都要有对应的 function_unit。
|
||||
|
||||
3. **区分"限制"与"禁止"**：文档中"行车娱乐限制"和"行车娱乐禁止"是两个不同的子场景（一个针对前台应用打断、一个针对后台应用启动限制），必须分别建模。
|
||||
|
||||
4. **区分不同应用类型**：系统限制、SDK 限制、其他应用的行为路径不同。
|
||||
|
||||
5. **包含开关状态**：开关"开启"和"关闭"两种状态下的行为都要覆盖。
|
||||
|
||||
6. **如果 resolved_conflicts 中以图片为准**，则优先按逻辑树的路径描述行为；**如果以文字为准**，则优先按表格文字描述。
|
||||
|
||||
## 输出格式
|
||||
|
||||
**只输出 JSON，不要有 markdown 代码块标记或其他文字**：
|
||||
|
||||
{
|
||||
"feature_name": "...",
|
||||
"concepts": [ ... ],
|
||||
"function_units": [ ... ]
|
||||
}
|
||||
+162
@@ -0,0 +1,162 @@
|
||||
你是吉利汽车车机系统的需求分析专家。你的任务是基于给定的精准上下文包，为单个功能单元（Function Unit）提取详细的 **IR 规则（Intermediate Representation Rule）**。
|
||||
|
||||
## 上下文
|
||||
|
||||
下面是一个功能单元的精准上下文包，包含了从原始需求文档中提取的相关文字、表格和逻辑树：
|
||||
|
||||
### 功能单元概要
|
||||
- **unit_id**: {unit_id}
|
||||
- **unit_name**: {unit_name}
|
||||
- **unit_description**: {unit_description}
|
||||
|
||||
### 相关文字段落
|
||||
{texts}
|
||||
|
||||
### 相关表格
|
||||
{tables}
|
||||
|
||||
### 相关逻辑树
|
||||
{logic_trees}
|
||||
|
||||
### 图文冲突仲裁（如有）
|
||||
{resolved_conflicts}
|
||||
|
||||
## IR Schema
|
||||
|
||||
你需要为这个功能单元输出一个 **规则数组（rules）**。每条规则遵循以下 schema：
|
||||
|
||||
```json
|
||||
{{
|
||||
"rule_id": "{unit_id}-SYS-FG-01",
|
||||
"description": "用完整的中文自然语言描述该规则的触发条件和行为，一句话概括",
|
||||
"priority": "P0",
|
||||
"sources": [
|
||||
{{"type": "table", "section": "3.1.1", "row": 2, "text_snippet": "打断：车速≥15km/h..."}},
|
||||
{{"type": "logic_tree", "image_id": "rId16", "node_ids": ["n19", "n21", "n23", "n25", "n26"]}}
|
||||
],
|
||||
"precondition": {{
|
||||
"switch": "开启",
|
||||
"app_type": "系统限制",
|
||||
"app_state": "前台"
|
||||
}},
|
||||
"trigger": {{
|
||||
"operator": "AND",
|
||||
"conditions": [
|
||||
{{"signal": "车速", "operator": ">=", "value": 15, "unit": "km/h"}},
|
||||
{{"signal": "车速_持续时间", "operator": ">", "value": 5, "unit": "秒"}},
|
||||
{{"signal": "档位", "operator": "!=", "value": "P"}}
|
||||
]
|
||||
}},
|
||||
"actions": [
|
||||
{{"type": "system", "description": "打断应用前台进程"}},
|
||||
{{"type": "system", "description": "将应用调入后台"}},
|
||||
{{"type": "user_interaction", "description": "显示Toast", "content": "在行车状态下无法使用该应用"}}
|
||||
]
|
||||
}}
|
||||
```
|
||||
|
||||
### 字段说明（必读）
|
||||
|
||||
1. **rule_id**: 格式为 `{unit_id}-类型-序号`，其中类型可以是 SYS（系统行为）、UI（用户交互）、SDK（SDK 限制）。序号从 01 开始。
|
||||
2. **description**: 完整但简洁地描述整个规则，包括前置条件、触发条件和所有动作。用中文。
|
||||
3. **priority**: P0（核心安全规则）、P1（重要规则）、P2（边界情况）。
|
||||
4. **sources**: 每条规则必须列出所有数据来源，包括：
|
||||
- 引用的表格行（section, row, text_snippet）
|
||||
- 引用的逻辑树节点 ID 列表（image_id, node_ids）。**注意：node_ids 必须列举该规则在逻辑树中经历的所有 decision 和 action 节点。**
|
||||
5. **precondition**: 规则生效的前置状态条件（开关状态、应用类型、应用前后台状态等）。可以是空对象 `{{}}` 如果无条件。
|
||||
6. **trigger**: 触发条件对象，包含：
|
||||
- `operator`: 条件组合方式，`"AND"` 或 `"OR"`
|
||||
- `conditions`: 条件数组，每个条件必须有 `signal`（信号名）、`operator`（比较运算符）、`value`（数值或字符串）。如果有单位，加 `unit` 字段。
|
||||
如果触发器是瞬时事件（如用户点击），使用 `event` 字段代替 `conditions`。
|
||||
7. **actions**: 每个动作必须有 `type`（`"system"` 或 `"user_interaction"`）和 `description`。用户可见交互（Toast、弹窗、语音播报）必须用 `"user_interaction"` 类型，并包含 `content` 字段。
|
||||
|
||||
## Few-shot 示例
|
||||
|
||||
### 示例 1：行车娱乐限制（前台打断）
|
||||
|
||||
**输入上下文**：开关开启，系统限制类应用在前台，车速≥15km/h且持续>5秒且非P档时，打断应用并显示Toast。
|
||||
|
||||
**期望输出**：
|
||||
|
||||
```json
|
||||
{{
|
||||
"rule_id": "FU-001-SYS-01",
|
||||
"description": "开关开启时，系统限制类应用在前台，当车速≥15km/h且持续超过5秒且非P档时，系统打断应用前台进程、将应用调入后台，并弹出Toast提示'在行车状态下无法使用该应用'",
|
||||
"priority": "P0",
|
||||
"sources": [
|
||||
{{"type": "table", "section": "3.1.1", "row": 2, "text_snippet": "行车娱乐限制：目标应用/功能处于前台时 ○ 打断：车速≥15km/h且持续5秒后..."}},
|
||||
{{"type": "logic_tree", "image_id": "rId16", "node_ids": ["n2", "n8", "n9", "n11", "n13", "n19", "n21", "n23", "n25", "n26"]}}
|
||||
],
|
||||
"precondition": {{
|
||||
"switch": "开启",
|
||||
"app_type": "系统限制",
|
||||
"app_state": "前台"
|
||||
}},
|
||||
"trigger": {{
|
||||
"operator": "AND",
|
||||
"conditions": [
|
||||
{{"signal": "车速", "operator": ">=", "value": 15, "unit": "km/h"}},
|
||||
{{"signal": "车速_持续时间", "operator": ">", "value": 5, "unit": "秒"}},
|
||||
{{"signal": "档位", "operator": "!=", "value": "P"}}
|
||||
]
|
||||
}},
|
||||
"actions": [
|
||||
{{"type": "system", "description": "打断应用前台进程"}},
|
||||
{{"type": "system", "description": "将应用调入后台"}},
|
||||
{{"type": "user_interaction", "description": "显示Toast", "content": "在行车状态下无法使用该应用"}}
|
||||
]
|
||||
}}
|
||||
```
|
||||
|
||||
### 示例 2：行车娱乐禁止（后台启动拦截）
|
||||
|
||||
**输入上下文**：开关开启，应用在后台且非前台，非P档时阻止应用启动，并提示。
|
||||
|
||||
**期望输出**：
|
||||
|
||||
```json
|
||||
{{
|
||||
"rule_id": "FU-002-SYS-01",
|
||||
"description": "开关开启时，目标应用处于后台，当档位非P档时，限制应用启动，并弹出Toast提示'请在P挡时使用该功能/应用'",
|
||||
"priority": "P0",
|
||||
"sources": [
|
||||
{{"type": "table", "section": "3.1.1", "row": 2, "text_snippet": "行车娱乐禁止：目标应用/功能处于后台时 ○ 限制：非P挡时，限制目标应用/功能启用..."}},
|
||||
{{"type": "logic_tree", "image_id": "rId17", "node_ids": ["n5", "n6"]}}
|
||||
],
|
||||
"precondition": {{
|
||||
"switch": "开启",
|
||||
"app_state": "后台"
|
||||
}},
|
||||
"trigger": {{
|
||||
"operator": "AND",
|
||||
"conditions": [
|
||||
{{"signal": "应用请求启动", "operator": "==", "value": true}}
|
||||
]
|
||||
}},
|
||||
"actions": [
|
||||
{{"type": "system", "description": "限制应用/功能启用"}},
|
||||
{{"type": "user_interaction", "description": "显示Toast", "content": "请在P挡时使用该功能/应用"}}
|
||||
]
|
||||
}}
|
||||
```
|
||||
|
||||
## 关键要求
|
||||
|
||||
1. **信号和数值必须精确**：不要写"车速超过阈值"，必须写 `"车速 >= 15 km/h"`。
|
||||
2. **条件必须完整**：如果文档说"车速≥15km/h 且持续超过5秒 且非P档"，这三个条件必须全部出现在 trigger.conditions 中。
|
||||
3. **逻辑树节点必须追踪**：在 sources 中列出该规则在逻辑树中经历的所有 decision 节点和 action 节点。这样做是为了后续审计（检查逻辑树覆盖率）。
|
||||
4. **优先图片逻辑树**：如果文字和图片存在矛盾，优先采用逻辑树中的路径，但保留文字作为补充参考（将两者都列入 sources）。
|
||||
5. **动作类型区分**：系统行为（打断进程、限制启动）用 `"system"`，用户可见交互（Toast、弹窗、语音）用 `"user_interaction"`。
|
||||
6. **多条规则**：如果一个功能单元包含多个独立的行为分支（如正常情况+异常情况），输出多条规则分别描述。规则之间通过 precondition 或 trigger 的条件值来区分。
|
||||
7. **开关关闭状态**：如果功能单元涉及开关，也要考虑开关关闭时的行为。开关关闭时所有限制失效，这也是一条规则。
|
||||
|
||||
## 输出格式
|
||||
|
||||
**只输出 JSON 数组，不要有任何其他文字或 markdown 标记**：
|
||||
|
||||
[
|
||||
{{ ... }},
|
||||
{{ ... }}
|
||||
]
|
||||
|
||||
注意：即使只有一个规则，也必须用数组格式 `[...]`。
|
||||
+200
@@ -0,0 +1,200 @@
|
||||
"""
|
||||
Stage 1: Semantic Index Generation.
|
||||
|
||||
Reads the full parsed PRD JSON, calls the LLM once to produce a semantic index
|
||||
identifying all function units, concepts, and their document sources.
|
||||
|
||||
Output: output/semantic_index.json
|
||||
"""
|
||||
|
||||
import json
|
||||
import re
|
||||
import sys
|
||||
import time
|
||||
from pathlib import Path
|
||||
|
||||
import config
|
||||
|
||||
|
||||
def format_document_for_prompt(doc: dict) -> str:
|
||||
"""Render the full parsed document as a readable string for the LLM prompt."""
|
||||
lines = []
|
||||
|
||||
# ---- Sections ----
|
||||
lines.append("=== SECTIONS ===")
|
||||
for i, section in enumerate(doc.get("sections", [])):
|
||||
source = section.get("source", f"(无标题-章节{i})")
|
||||
lines.append(f"\n--- Section: {source} ---")
|
||||
|
||||
for block in section.get("blocks", []):
|
||||
if block["type"] == "para":
|
||||
lines.append(f"[段落 {block['index']}] {block['text']}")
|
||||
elif block["type"] == "table":
|
||||
lines.append(f"[表格 {block.get('table', '?')}]")
|
||||
headers = block.get("headers", [])
|
||||
lines.append(f" 表头: {' | '.join(headers)}")
|
||||
for row in block.get("rows", []):
|
||||
cols = row.get("columns", [])
|
||||
cell_texts = []
|
||||
for c in cols:
|
||||
cell_texts.append(f"[行{c.get('row','?')}]{c.get('name','')}: {c.get('text','')}")
|
||||
lines.append(f" {'; '.join(cell_texts)}")
|
||||
|
||||
images = section.get("images", [])
|
||||
if images:
|
||||
lines.append(f" 图片引用: {', '.join(images)}")
|
||||
|
||||
# ---- Image Analysis ----
|
||||
lines.append("\n\n=== IMAGE_ANALYSIS (流程图逻辑树) ===")
|
||||
for img in doc.get("image_analysis", []):
|
||||
rid = img.get("rid", "?")
|
||||
img_type = img.get("type", "?")
|
||||
lines.append(f"\n--- Image: {rid} (type={img_type}) ---")
|
||||
lines.append(f" 描述: {img.get('description', '')[:300]}")
|
||||
|
||||
lt = img.get("logic_tree")
|
||||
if lt:
|
||||
lines.append(f" 逻辑树根节点: {lt.get('root', '?')}")
|
||||
lines.append(" 节点详情:")
|
||||
for node in lt.get("nodes", []):
|
||||
nid = node.get("id", "?")
|
||||
ntype = node.get("type", "?")
|
||||
desc = node.get("description", "") or node.get("condition", "")
|
||||
lines.append(f" [{ntype}] {nid}: {desc}")
|
||||
branches = node.get("branches", [])
|
||||
if branches:
|
||||
for br in branches:
|
||||
lines.append(f" → {br['value']} → {br['target']}")
|
||||
|
||||
# ---- Resolved Conflicts ----
|
||||
conflicts = doc.get("resolved_conflicts", [])
|
||||
if conflicts:
|
||||
lines.append("\n\n=== RESOLVED_CONFLICTS (图文冲突仲裁) ===")
|
||||
for c in conflicts:
|
||||
lines.append(
|
||||
f" [{c.get('conflict_type','?')}] {c.get('section','?')}: "
|
||||
f"以{c.get('source','?')}为准 — {c.get('correction','')}"
|
||||
)
|
||||
|
||||
return "\n".join(lines)
|
||||
|
||||
|
||||
def build_prompt(doc: dict) -> str:
|
||||
"""Load the prompt template and inject the formatted document."""
|
||||
template_path = Path(config.PROMPTS_DIR) / "step1_semantic_index.txt"
|
||||
template = template_path.read_text(encoding="utf-8")
|
||||
|
||||
formatted_doc = format_document_for_prompt(doc)
|
||||
prompt = template.replace("{document_json}", formatted_doc)
|
||||
return prompt
|
||||
|
||||
|
||||
def extract_json_from_response(text: str) -> str:
|
||||
"""Robustly extract JSON from LLM response, handling markdown fences."""
|
||||
# Try ```json ... ``` first
|
||||
m = re.search(r"```(?:json)?\s*([\s\S]*?)```", text)
|
||||
if m:
|
||||
return m.group(1).strip()
|
||||
|
||||
# Try to find the outermost { ... }
|
||||
start = text.find("{")
|
||||
if start == -1:
|
||||
raise ValueError("No JSON object found in LLM response")
|
||||
|
||||
depth = 0
|
||||
for i in range(start, len(text)):
|
||||
if text[i] == "{":
|
||||
depth += 1
|
||||
elif text[i] == "}":
|
||||
depth -= 1
|
||||
if depth == 0:
|
||||
return text[start : i + 1]
|
||||
|
||||
raise ValueError("Unclosed JSON object in LLM response")
|
||||
|
||||
|
||||
def call_llm(prompt: str, max_retries: int = 2) -> dict:
|
||||
"""Send the prompt to the LLM and return the parsed semantic index JSON."""
|
||||
client = config.llm_client()
|
||||
|
||||
for attempt in range(max_retries + 1):
|
||||
print(f" LLM 调用 (尝试 {attempt + 1}/{max_retries + 1})...", flush=True)
|
||||
try:
|
||||
resp = client.chat.completions.create(
|
||||
model=config.MODEL_NAME,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个精确的 JSON 输出引擎。只输出合法的 JSON，不输出任何其他文字。"},
|
||||
{"role": "user", "content": prompt},
|
||||
],
|
||||
temperature=config.TEMPERATURE,
|
||||
max_tokens=config.MAX_TOKENS,
|
||||
)
|
||||
content = resp.choices[0].message.content
|
||||
if content is None:
|
||||
raise RuntimeError("LLM returned empty response")
|
||||
|
||||
json_str = extract_json_from_response(content)
|
||||
result = json.loads(json_str)
|
||||
|
||||
# Quick structural validation
|
||||
_validate_schema(result)
|
||||
return result
|
||||
|
||||
except (json.JSONDecodeError, ValueError) as e:
|
||||
print(f" JSON 解析失败: {e}")
|
||||
if attempt < max_retries:
|
||||
time.sleep(2)
|
||||
else:
|
||||
raise RuntimeError(f"无法从 LLM 响应中解析 JSON: {e}") from e
|
||||
|
||||
|
||||
def _validate_schema(data: dict) -> None:
|
||||
"""Validate the top-level schema of the semantic index."""
|
||||
if "feature_name" not in data:
|
||||
raise ValueError("semantic_index 缺少 'feature_name' 字段")
|
||||
if "function_units" not in data:
|
||||
raise ValueError("semantic_index 缺少 'function_units' 字段")
|
||||
|
||||
units = data["function_units"]
|
||||
if not isinstance(units, list) or len(units) == 0:
|
||||
raise ValueError("function_units 必须是非空数组")
|
||||
|
||||
for i, fu in enumerate(units):
|
||||
if not fu.get("unit_id"):
|
||||
raise ValueError(f"function_unit[{i}] 缺少 unit_id")
|
||||
if not fu.get("name"):
|
||||
raise ValueError(f"function_unit[{i}] 缺少 name")
|
||||
|
||||
|
||||
def main():
|
||||
print("=" * 60)
|
||||
print("阶段一：宏观语义索引 (Semantic Index)")
|
||||
print("=" * 60)
|
||||
|
||||
# 1. Load input
|
||||
print(f"\n[1/4] 加载输入文档: {config.INPUT_JSON}")
|
||||
doc = config.load_input_document()
|
||||
print(f" 已加载 {len(doc.get('sections', []))} 个 section, "
|
||||
f"{len(doc.get('image_analysis', []))} 张图片分析")
|
||||
|
||||
# 2. Build prompt
|
||||
print(f"\n[2/4] 构造 Prompt...")
|
||||
prompt = build_prompt(doc)
|
||||
print(f" Prompt 长度: {len(prompt)} 字符 (~{len(prompt)//3} tokens)")
|
||||
|
||||
# 3. Call LLM
|
||||
print(f"\n[3/4] 调用 LLM ({config.MODEL_NAME})...")
|
||||
semantic_index = call_llm(prompt)
|
||||
|
||||
# 4. Save result
|
||||
print(f"\n[4/4] 保存语义索引: {config.SEMANTIC_INDEX_JSON}")
|
||||
config.save_json(semantic_index, config.SEMANTIC_INDEX_JSON)
|
||||
|
||||
n_concepts = len(semantic_index.get("concepts", []))
|
||||
n_units = len(semantic_index.get("function_units", []))
|
||||
print(f"\n完成! 提取了 {n_concepts} 个概念, {n_units} 个功能单元.")
|
||||
print(f"输出: {config.SEMANTIC_INDEX_JSON}")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
+410
@@ -0,0 +1,410 @@
|
||||
"""
|
||||
Stage 2: Per Function Unit IR Extraction.
|
||||
|
||||
For each function unit from the semantic index, constructs a precision context
|
||||
package and calls the LLM to extract detailed IR rules.
|
||||
|
||||
Runs multiple LLM calls in parallel (up to MAX_CONCURRENCY).
|
||||
|
||||
Output: output/ir_fragments.json
|
||||
"""
|
||||
|
||||
import concurrent.futures
|
||||
import json
|
||||
import re
|
||||
import sys
|
||||
import time
|
||||
from pathlib import Path
|
||||
|
||||
import config
|
||||
|
||||
|
||||
MAX_CONCURRENCY = 3 # Max parallel LLM calls
|
||||
|
||||
|
||||
def load_semantic_index() -> dict:
|
||||
"""Load the semantic index from Stage 1."""
|
||||
return config.load_json(config.SEMANTIC_INDEX_JSON)
|
||||
|
||||
|
||||
def build_document_lookup(doc: dict):
|
||||
"""Build lookup structures for fast context extraction from the document."""
|
||||
|
||||
# sections_by_source: "3.1.1" -> section dict
|
||||
sections_by_source = {}
|
||||
for section in doc.get("sections", []):
|
||||
source = section.get("source", "")
|
||||
# Normalize: extract leading number like "3.1.1"
|
||||
parts = source.split()
|
||||
if parts:
|
||||
key = parts[0].strip()
|
||||
sections_by_source[key] = section
|
||||
|
||||
# image_by_rid: "rId16" -> image_analysis entry
|
||||
image_by_rid = {}
|
||||
for img in doc.get("image_analysis", []):
|
||||
rid = img.get("rid", "")
|
||||
if rid:
|
||||
image_by_rid[rid] = img
|
||||
|
||||
# Conflicts indexed by section
|
||||
conflicts_by_section = {}
|
||||
for c in doc.get("resolved_conflicts", []):
|
||||
section = c.get("section", "")
|
||||
key = section.split()[0] if section else ""
|
||||
conflicts_by_section.setdefault(key, []).append(c)
|
||||
|
||||
return sections_by_source, image_by_rid, conflicts_by_section
|
||||
|
||||
|
||||
def extract_context_package(
|
||||
fu: dict, doc: dict, sections_by_source: dict, image_by_rid: dict,
|
||||
conflicts_by_section: dict
|
||||
) -> dict:
|
||||
"""Build a precision context package for a single function unit."""
|
||||
texts = []
|
||||
tables = []
|
||||
logic_trees = []
|
||||
seen_sections = set()
|
||||
seen_images = set()
|
||||
|
||||
for src in fu.get("sources", []):
|
||||
src_type = src.get("type", "")
|
||||
section_key = src.get("section", "").split()[0] if src.get("section") else ""
|
||||
|
||||
# --- Text source ---
|
||||
if src_type in ("table", "para") and section_key:
|
||||
if section_key in seen_sections:
|
||||
continue
|
||||
seen_sections.add(section_key)
|
||||
|
||||
section = sections_by_source.get(section_key)
|
||||
if section is None:
|
||||
# Fuzzy match by prefix
|
||||
for key in sections_by_source:
|
||||
if key.startswith(section_key):
|
||||
section = sections_by_source[key]
|
||||
break
|
||||
|
||||
if section:
|
||||
for block in section.get("blocks", []):
|
||||
if block["type"] == "para":
|
||||
texts.append({
|
||||
"section": section_key,
|
||||
"text": block["text"]
|
||||
})
|
||||
elif block["type"] == "table":
|
||||
row_num = src.get("row") if src_type == "table" else None
|
||||
if row_num is not None:
|
||||
# Extract only the specific row
|
||||
matching_rows = []
|
||||
for r in block.get("rows", []):
|
||||
for c in r.get("columns", []):
|
||||
if c.get("row") == row_num:
|
||||
matching_rows.append({
|
||||
"headers": block.get("headers", []),
|
||||
"cells": {
|
||||
col["name"]: col["text"]
|
||||
for col in r["columns"]
|
||||
},
|
||||
"row": row_num
|
||||
})
|
||||
break
|
||||
tables.append({
|
||||
"section": section_key,
|
||||
"headers": block.get("headers", []),
|
||||
"rows": matching_rows,
|
||||
"all_rows": [
|
||||
{
|
||||
"row": col.get("row"),
|
||||
"name": col.get("name"),
|
||||
"text": col.get("text")
|
||||
}
|
||||
for row in block.get("rows", [])
|
||||
for col in row.get("columns", [])
|
||||
]
|
||||
})
|
||||
else:
|
||||
# Include full table
|
||||
tables.append({
|
||||
"section": section_key,
|
||||
"headers": block.get("headers", []),
|
||||
"all_rows": [
|
||||
{
|
||||
"row": col.get("row"),
|
||||
"name": col.get("name"),
|
||||
"text": col.get("text")
|
||||
}
|
||||
for row in block.get("rows", [])
|
||||
for col in row.get("columns", [])
|
||||
]
|
||||
})
|
||||
|
||||
# --- Logic tree source ---
|
||||
if src_type == "logic_tree":
|
||||
image_id = src.get("image_id", "")
|
||||
if not image_id or image_id in seen_images:
|
||||
continue
|
||||
seen_images.add(image_id)
|
||||
|
||||
img = image_by_rid.get(image_id)
|
||||
if img:
|
||||
lt = img.get("logic_tree")
|
||||
if lt:
|
||||
logic_trees.append({
|
||||
"image_id": image_id,
|
||||
"description": img.get("description", ""),
|
||||
"tree": lt
|
||||
})
|
||||
|
||||
# Include relevant resolved conflicts
|
||||
relevant_conflicts = []
|
||||
for section_key in seen_sections:
|
||||
for c in conflicts_by_section.get(section_key, []):
|
||||
relevant_conflicts.append(c)
|
||||
|
||||
return {
|
||||
"unit_id": fu["unit_id"],
|
||||
"unit_name": fu.get("name", ""),
|
||||
"unit_description": fu.get("description", ""),
|
||||
"texts": texts,
|
||||
"tables": tables,
|
||||
"logic_trees": logic_trees,
|
||||
"resolved_conflicts": relevant_conflicts
|
||||
}
|
||||
|
||||
|
||||
def format_context_package(pkg: dict) -> str:
|
||||
"""Format a context package as a readable string for the prompt."""
|
||||
parts = []
|
||||
|
||||
# Texts
|
||||
parts.append("【文字段落】")
|
||||
for i, t in enumerate(pkg.get("texts", [])):
|
||||
parts.append(f"[{t.get('section', '?')}] {t.get('text', '')}")
|
||||
if not pkg.get("texts"):
|
||||
parts.append("（无）")
|
||||
|
||||
# Tables
|
||||
parts.append("\n【表格数据】")
|
||||
for i, tbl in enumerate(pkg.get("tables", [])):
|
||||
parts.append(f"表格 {i+1} (section={tbl.get('section', '?')})")
|
||||
headers = tbl.get("headers", [])
|
||||
parts.append(f" 表头: {headers}")
|
||||
parts.append(" 全部行数据:")
|
||||
for row in tbl.get("all_rows", []):
|
||||
parts.append(
|
||||
f" 行{row.get('row','?')}[{row.get('name','?')}]: {row.get('text','')}"
|
||||
)
|
||||
# Highlight matched rows if any
|
||||
matched = tbl.get("rows", [])
|
||||
if matched:
|
||||
parts.append(" <重点关注行>:")
|
||||
for mr in matched:
|
||||
parts.append(f" 行{mr.get('row','?')}: {mr.get('cells', {})}")
|
||||
if not pkg.get("tables"):
|
||||
parts.append("（无）")
|
||||
|
||||
# Logic trees
|
||||
parts.append("\n【逻辑树】")
|
||||
for i, lt in enumerate(pkg.get("logic_trees", [])):
|
||||
parts.append(f"逻辑树 {i+1} (image_id={lt.get('image_id', '?')})")
|
||||
parts.append(f" 描述: {lt.get('description', '')[:200]}")
|
||||
tree = lt.get("tree", {})
|
||||
parts.append(f" 根: {tree.get('root', '?')}")
|
||||
parts.append(" 节点:")
|
||||
for node in tree.get("nodes", []):
|
||||
nid = node.get("id", "?")
|
||||
ntype = node.get("type", "?")
|
||||
desc = node.get("description", "") or node.get("condition", "")
|
||||
parts.append(f" [{ntype}] {nid}: {desc}")
|
||||
for br in node.get("branches", []):
|
||||
parts.append(f" → {br['value']} → {br['target']}")
|
||||
if not pkg.get("logic_trees"):
|
||||
parts.append("（无）")
|
||||
|
||||
# Conflicts
|
||||
conflicts = pkg.get("resolved_conflicts", [])
|
||||
if conflicts:
|
||||
parts.append("\n【图文冲突仲裁】")
|
||||
for c in conflicts:
|
||||
parts.append(
|
||||
f" [{c.get('conflict_type', '?')}] 以{c.get('source', '?')}为准: "
|
||||
f"{c.get('correction', '')}"
|
||||
)
|
||||
|
||||
return "\n".join(parts)
|
||||
|
||||
|
||||
def _escape_json_for_format(s: str) -> str:
|
||||
"""Escape curly braces in a JSON string for use with str.format()."""
|
||||
return s.replace("{", "{{").replace("}", "}}")
|
||||
|
||||
|
||||
def build_prompt(pkg: dict) -> str:
|
||||
"""Build the LLM prompt for a single function unit."""
|
||||
template_path = Path(config.PROMPTS_DIR) / "step2_ir_extraction.txt"
|
||||
template = template_path.read_text(encoding="utf-8")
|
||||
|
||||
prompt = template.format(
|
||||
unit_id=pkg["unit_id"],
|
||||
unit_name=pkg["unit_name"],
|
||||
unit_description=pkg["unit_description"],
|
||||
texts=_escape_json_for_format(
|
||||
json.dumps(pkg.get("texts", []), ensure_ascii=False, indent=2)
|
||||
),
|
||||
tables=_escape_json_for_format(
|
||||
json.dumps(pkg.get("tables", []), ensure_ascii=False, indent=2)
|
||||
),
|
||||
logic_trees=_escape_json_for_format(
|
||||
json.dumps(pkg.get("logic_trees", []), ensure_ascii=False, indent=2)
|
||||
),
|
||||
resolved_conflicts=_escape_json_for_format(
|
||||
json.dumps(pkg.get("resolved_conflicts", []), ensure_ascii=False, indent=2)
|
||||
),
|
||||
)
|
||||
return prompt
|
||||
|
||||
|
||||
def extract_json_from_response(text: str) -> str:
|
||||
"""Extract JSON array from LLM response."""
|
||||
m = re.search(r"```(?:json)?\s*(\[[\s\S]*?\])\s*```", text)
|
||||
if m:
|
||||
return m.group(1).strip()
|
||||
|
||||
# Find outermost [ ... ]
|
||||
start = text.find("[")
|
||||
if start == -1:
|
||||
raise ValueError("No JSON array found in LLM response")
|
||||
|
||||
depth = 0
|
||||
for i in range(start, len(text)):
|
||||
if text[i] == "[":
|
||||
depth += 1
|
||||
elif text[i] == "]":
|
||||
depth -= 1
|
||||
if depth == 0:
|
||||
return text[start : i + 1]
|
||||
|
||||
raise ValueError("Unclosed JSON array in LLM response")
|
||||
|
||||
|
||||
def extract_rules_for_unit(pkg: dict, max_retries: int = 2) -> list[dict]:
|
||||
"""Call LLM for one function unit, return its IR rules."""
|
||||
prompt = build_prompt(pkg)
|
||||
client = config.llm_client()
|
||||
|
||||
for attempt in range(max_retries + 1):
|
||||
try:
|
||||
resp = client.chat.completions.create(
|
||||
model=config.MODEL_NAME,
|
||||
messages=[
|
||||
{
|
||||
"role": "system",
|
||||
"content": "你是一个精确的 JSON 输出引擎。只输出合法的 JSON 数组，不输出任何其他文字。"
|
||||
},
|
||||
{"role": "user", "content": prompt},
|
||||
],
|
||||
temperature=config.TEMPERATURE,
|
||||
max_tokens=config.MAX_TOKENS,
|
||||
)
|
||||
content = resp.choices[0].message.content
|
||||
if content is None:
|
||||
raise RuntimeError("LLM returned empty response")
|
||||
|
||||
json_str = extract_json_from_response(content)
|
||||
rules = json.loads(json_str)
|
||||
if not isinstance(rules, list):
|
||||
raise ValueError(f"Expected JSON array, got {type(rules).__name__}")
|
||||
|
||||
return rules
|
||||
|
||||
except (json.JSONDecodeError, ValueError) as e:
|
||||
print(f" JSON 解析失败 (尝试 {attempt + 1}): {e}")
|
||||
if attempt < max_retries:
|
||||
time.sleep(2)
|
||||
|
||||
return []
|
||||
|
||||
|
||||
def extract_all_rules(
|
||||
semantic_index: dict, doc: dict
|
||||
) -> list[dict]:
|
||||
"""Extract IR rules for all function units. Runs in parallel up to MAX_CONCURRENCY."""
|
||||
sections_by_source, image_by_rid, conflicts_by_section = build_document_lookup(doc)
|
||||
function_units = semantic_index.get("function_units", [])
|
||||
|
||||
print(f" 共 {len(function_units)} 个功能单元待处理")
|
||||
print(f" 最大并发: {MAX_CONCURRENCY}")
|
||||
|
||||
# Build context packages (serial — fast)
|
||||
packages = []
|
||||
for fu in function_units:
|
||||
pkg = extract_context_package(
|
||||
fu, doc, sections_by_source, image_by_rid, conflicts_by_section
|
||||
)
|
||||
packages.append(pkg)
|
||||
|
||||
# Run LLM calls in parallel
|
||||
fragments = []
|
||||
with concurrent.futures.ThreadPoolExecutor(max_workers=MAX_CONCURRENCY) as executor:
|
||||
futures = {}
|
||||
for i, pkg in enumerate(packages):
|
||||
future = executor.submit(extract_rules_for_unit, pkg)
|
||||
futures[future] = (i, pkg["unit_id"], pkg["unit_name"])
|
||||
|
||||
for future in concurrent.futures.as_completed(futures):
|
||||
i, uid, uname = futures[future]
|
||||
try:
|
||||
rules = future.result()
|
||||
fragments.append({
|
||||
"unit_id": uid,
|
||||
"unit_name": uname,
|
||||
"rules": rules
|
||||
})
|
||||
print(f" ✅ {uid} ({uname}): {len(rules)} 条规则")
|
||||
except Exception as e:
|
||||
print(f" ❌ {uid} ({uname}): 失败 — {e}")
|
||||
fragments.append({
|
||||
"unit_id": uid,
|
||||
"unit_name": uname,
|
||||
"rules": [],
|
||||
"error": str(e)
|
||||
})
|
||||
|
||||
# Sort by unit_id to maintain stable ordering
|
||||
fragments.sort(key=lambda f: f["unit_id"])
|
||||
return fragments
|
||||
|
||||
|
||||
def main():
|
||||
print("=" * 60)
|
||||
print("阶段二：逐功能单元 IR 提取")
|
||||
print("=" * 60)
|
||||
|
||||
# 1. Load inputs
|
||||
print(f"\n[1/3] 加载输入...")
|
||||
semantic_index = load_semantic_index()
|
||||
doc = config.load_input_document()
|
||||
n_units = len(semantic_index.get("function_units", []))
|
||||
print(f" 语义索引: {n_units} 个功能单元")
|
||||
|
||||
# 2. Extract rules
|
||||
print(f"\n[2/3] 逐单元提取 IR 规则...")
|
||||
fragments = extract_all_rules(semantic_index, doc)
|
||||
|
||||
# 3. Save
|
||||
print(f"\n[3/3] 保存 IR 片段...")
|
||||
config.save_json(fragments, config.IR_FRAGMENTS_JSON)
|
||||
|
||||
total_rules = sum(len(f["rules"]) for f in fragments)
|
||||
failed_units = [f for f in fragments if f.get("error")]
|
||||
print(f"\n完成! {len(fragments)} 个功能单元, 共 {total_rules} 条规则")
|
||||
if failed_units:
|
||||
print(f" ⚠️ {len(failed_units)} 个单元提取失败: "
|
||||
f"{[f['unit_id'] for f in failed_units]}")
|
||||
print(f"输出: {config.IR_FRAGMENTS_JSON}")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
+474
@@ -0,0 +1,474 @@
|
||||
"""
|
||||
Stage 3: Deterministic Merge & Completeness Audit.
|
||||
|
||||
- Merges IR rule fragments, deduplicating by trigger+actions similarity.
|
||||
- Reassigns stable rule_ids.
|
||||
- Generates an audit report covering:
|
||||
1. Logic tree node coverage
|
||||
2. Table enumeration coverage
|
||||
3. Global switch state coverage
|
||||
|
||||
Outputs:
|
||||
- ir_final.json (in doc_parser output per spec)
|
||||
- ir_audit_report.md (in doc_parser output)
|
||||
"""
|
||||
|
||||
import json
|
||||
import hashlib
|
||||
import sys
|
||||
from collections import defaultdict
|
||||
from pathlib import Path
|
||||
|
||||
import config
|
||||
|
||||
|
||||
PASS = "[PASS]"
|
||||
WARN = "[WARN]"
|
||||
FAIL = "[FAIL]"
|
||||
|
||||
|
||||
def load_fragments() -> list[dict]:
|
||||
"""Load IR fragments from Stage 2."""
|
||||
return config.load_json(config.IR_FRAGMENTS_JSON)
|
||||
|
||||
|
||||
def load_semantic_index() -> dict:
|
||||
"""Load semantic index from Stage 1."""
|
||||
return config.load_json(config.SEMANTIC_INDEX_JSON)
|
||||
|
||||
|
||||
def rule_signature(rule: dict) -> str:
|
||||
"""Generate a dedup signature from trigger + actions.
|
||||
|
||||
Two rules with identical trigger conditions and actions produce
|
||||
the same signature and should be merged.
|
||||
"""
|
||||
trigger = rule.get("trigger", {})
|
||||
actions = rule.get("actions", [])
|
||||
|
||||
# Normalize: sort conditions by signal name for stability
|
||||
conditions = sorted(trigger.get("conditions", []), key=lambda c: c.get("signal", ""))
|
||||
|
||||
# Sort actions by description
|
||||
sorted_actions = sorted(actions, key=lambda a: a.get("description", ""))
|
||||
|
||||
sig_data = {
|
||||
"conditions": conditions,
|
||||
"actions": sorted_actions,
|
||||
}
|
||||
sig_json = json.dumps(sig_data, ensure_ascii=False, sort_keys=True)
|
||||
return hashlib.sha256(sig_json.encode()).hexdigest()[:16]
|
||||
|
||||
|
||||
def merge_rules(fragments: list[dict]) -> list[dict]:
|
||||
"""Merge rules across all fragments, deduplicating by trigger+actions."""
|
||||
signature_map: dict[str, dict] = {}
|
||||
order = []
|
||||
|
||||
for fragment in fragments:
|
||||
for rule in fragment.get("rules", []):
|
||||
sig = rule_signature(rule)
|
||||
|
||||
if sig in signature_map:
|
||||
# Merge sources
|
||||
existing = signature_map[sig]
|
||||
existing_sources = existing.setdefault("sources", [])
|
||||
for src in rule.get("sources", []):
|
||||
if src not in existing_sources:
|
||||
existing_sources.append(src)
|
||||
# Use the more detailed description
|
||||
if len(rule.get("description", "")) > len(existing.get("description", "")):
|
||||
existing["description"] = rule["description"]
|
||||
else:
|
||||
signature_map[sig] = dict(rule)
|
||||
order.append(sig)
|
||||
|
||||
merged = [signature_map[sig] for sig in order]
|
||||
print(f" 合并前: {sum(len(f.get('rules', [])) for f in fragments)} 条规则")
|
||||
print(f" 合并后: {len(merged)} 条规则")
|
||||
return merged
|
||||
|
||||
|
||||
def assign_rule_ids(rules: list[dict], feature_id: str = "DRL-001") -> list[dict]:
|
||||
"""Reassign stable rule_ids based on type and sequence."""
|
||||
type_counters = defaultdict(int)
|
||||
|
||||
for rule in rules:
|
||||
# Determine type from the first action's type
|
||||
actions = rule.get("actions", [])
|
||||
if any(a.get("type") == "user_interaction" for a in actions) and \
|
||||
not any(a.get("type") == "system" for a in actions):
|
||||
rtype = "UI"
|
||||
elif any("SDK" in str(a) for a in actions):
|
||||
rtype = "SDK"
|
||||
else:
|
||||
rtype = "SYS"
|
||||
|
||||
type_counters[rtype] += 1
|
||||
seq = type_counters[rtype]
|
||||
rule["rule_id"] = f"{feature_id}-{rtype}-FG-{seq:02d}"
|
||||
|
||||
# Also generate top-level feature metadata
|
||||
return rules
|
||||
|
||||
|
||||
def find_all_logic_tree_nodes(doc: dict) -> dict[str, list[dict]]:
|
||||
"""Return {image_id: [all nodes]} for all logic trees."""
|
||||
result = {}
|
||||
for img in doc.get("image_analysis", []):
|
||||
lt = img.get("logic_tree")
|
||||
rid = img.get("rid", "")
|
||||
if lt and rid:
|
||||
result[rid] = lt.get("nodes", [])
|
||||
return result
|
||||
|
||||
|
||||
def find_referenced_nodes(rules: list[dict]) -> dict[str, set[str]]:
|
||||
"""Return {image_id: {referenced node ids}} across all rules."""
|
||||
referenced = defaultdict(set)
|
||||
for rule in rules:
|
||||
for src in rule.get("sources", []):
|
||||
if src.get("type") == "logic_tree":
|
||||
image_id = src.get("image_id", "")
|
||||
for nid in src.get("node_ids", []):
|
||||
referenced[image_id].add(nid)
|
||||
return dict(referenced)
|
||||
|
||||
|
||||
def audit_logic_tree_coverage(
|
||||
doc: dict, rules: list[dict]
|
||||
) -> list[dict]:
|
||||
"""Generate coverage statistics for logic tree nodes."""
|
||||
all_nodes = find_all_logic_tree_nodes(doc)
|
||||
referenced = find_referenced_nodes(rules)
|
||||
results = []
|
||||
|
||||
for image_id, nodes in all_nodes.items():
|
||||
ref_set = referenced.get(image_id, set())
|
||||
|
||||
decision_nodes = [n for n in nodes if n["type"] == "decision"]
|
||||
action_nodes = [n for n in nodes if n["type"] == "action"]
|
||||
state_nodes = [n for n in nodes if n["type"] == "state"]
|
||||
|
||||
decisions_covered = [n for n in decision_nodes if n["id"] in ref_set]
|
||||
actions_covered = [n for n in action_nodes if n["id"] in ref_set]
|
||||
decisions_uncovered = [n for n in decision_nodes if n["id"] not in ref_set]
|
||||
actions_uncovered = [n for n in action_nodes if n["id"] not in ref_set]
|
||||
|
||||
total_checkable = len(decision_nodes) + len(action_nodes)
|
||||
total_covered = len(decisions_covered) + len(actions_covered)
|
||||
coverage = (total_covered / total_checkable * 100) if total_checkable > 0 else 100
|
||||
|
||||
status = PASS if coverage >= 95 else (WARN if coverage >= 70 else FAIL)
|
||||
detail_parts = [f"{total_covered}/{total_checkable} decision+action 节点被引用"]
|
||||
|
||||
if decisions_uncovered:
|
||||
detail_parts.append(
|
||||
f"未覆盖的 decision: {[n['id'] + ': ' + n.get('condition','')[:40] for n in decisions_uncovered]}"
|
||||
)
|
||||
if actions_uncovered:
|
||||
detail_parts.append(
|
||||
f"未覆盖的 action: {[n['id'] + ': ' + n.get('description','')[:40] for n in actions_uncovered]}"
|
||||
)
|
||||
|
||||
results.append({
|
||||
"check": f"逻辑树 {image_id} 节点覆盖率",
|
||||
"status": status,
|
||||
"coverage_pct": round(coverage, 1),
|
||||
"detail": "; ".join(detail_parts),
|
||||
"image_id": image_id,
|
||||
"uncovered_decisions": decisions_uncovered,
|
||||
"uncovered_actions": actions_uncovered,
|
||||
})
|
||||
|
||||
return results
|
||||
|
||||
|
||||
def find_table_enums(doc: dict) -> list[dict]:
|
||||
"""Find enumerated values in tables (e.g., app types, limit methods)."""
|
||||
enums = []
|
||||
|
||||
for section in doc.get("sections", []):
|
||||
for block in section.get("blocks", []):
|
||||
if block["type"] != "table":
|
||||
continue
|
||||
headers = block.get("headers", [])
|
||||
if not headers:
|
||||
continue
|
||||
|
||||
# Look for the "功能" / "功能详细说明" table pattern (key-value pairs)
|
||||
if "功能" in headers and "功能详细说明" in headers:
|
||||
for row in block.get("rows", []):
|
||||
cols = row.get("columns", [])
|
||||
key_col = next((c for c in cols if c.get("name") == "功能"), None)
|
||||
val_col = next(
|
||||
(c for c in cols if c.get("name") == "功能详细说明"), None
|
||||
)
|
||||
if key_col and val_col:
|
||||
enums.append({
|
||||
"section": section.get("source", ""),
|
||||
"row": key_col.get("row"),
|
||||
"key": key_col.get("text", ""),
|
||||
"value": val_col.get("text", ""),
|
||||
})
|
||||
else:
|
||||
# Generic table: record first column values as potential enum
|
||||
first_col_name = headers[0] if headers else ""
|
||||
values = []
|
||||
for row in block.get("rows", []):
|
||||
for col in row.get("columns", []):
|
||||
if col.get("name") == first_col_name:
|
||||
values.append(col.get("text", ""))
|
||||
if values:
|
||||
enums.append({
|
||||
"section": section.get("source", ""),
|
||||
"column": first_col_name,
|
||||
"values": values,
|
||||
})
|
||||
|
||||
return enums
|
||||
|
||||
|
||||
def audit_table_enums(rules: list[dict], doc: dict) -> list[dict]:
|
||||
"""Check if key enumerated values appear in rule preconditions."""
|
||||
results = []
|
||||
table_enums = find_table_enums(doc)
|
||||
|
||||
# Collect all rule precondition fields and their values
|
||||
rule_preconditions = []
|
||||
for rule in rules:
|
||||
precond = rule.get("precondition", {})
|
||||
rule_preconditions.append(precond)
|
||||
|
||||
# Check specific enum categories
|
||||
app_types = {"系统限制", "SDK限制", "其他应用"}
|
||||
switch_states = {"开启", "关闭"}
|
||||
app_states = {"前台", "后台"}
|
||||
|
||||
# App type coverage
|
||||
found_app_types = set()
|
||||
for precond in rule_preconditions:
|
||||
at = precond.get("app_type", "")
|
||||
if at:
|
||||
found_app_types.add(at)
|
||||
missing_types = app_types - found_app_types
|
||||
results.append({
|
||||
"check": "应用类型枚举覆盖",
|
||||
"status": PASS if not missing_types else WARN,
|
||||
"detail": f"已覆盖: {found_app_types or '无'}"
|
||||
+ (f"; 未覆盖: {missing_types}" if missing_types else ""),
|
||||
})
|
||||
|
||||
# App state coverage
|
||||
found_states = set()
|
||||
for precond in rule_preconditions:
|
||||
st = precond.get("app_state", "")
|
||||
if st:
|
||||
found_states.add(st)
|
||||
missing_states = app_states - found_states
|
||||
results.append({
|
||||
"check": "应用前后台状态覆盖",
|
||||
"status": PASS if not missing_states else WARN,
|
||||
"detail": f"已覆盖: {found_states or '无'}"
|
||||
+ (f"; 未覆盖: {missing_states}" if missing_states else ""),
|
||||
})
|
||||
|
||||
# Trigger signal coverage (check each table enum key appears)
|
||||
trigger_signals = set()
|
||||
for rule in rules:
|
||||
for cond in rule.get("trigger", {}).get("conditions", []):
|
||||
signal = cond.get("signal", "")
|
||||
if signal:
|
||||
trigger_signals.add(signal)
|
||||
|
||||
# Check if key concepts from the doc appear in signals
|
||||
key_signals = {"车速", "档位", "车速_持续时间", "应用请求启动"}
|
||||
missing_signals = key_signals - trigger_signals
|
||||
results.append({
|
||||
"check": "触发信号覆盖（车速/档位/持续时间/启动请求）",
|
||||
"status": PASS if not missing_signals else WARN,
|
||||
"detail": f"已覆盖信号: {sorted(trigger_signals)}"
|
||||
+ (f"; 未覆盖: {missing_signals}" if missing_signals else ""),
|
||||
})
|
||||
|
||||
return results
|
||||
|
||||
|
||||
def audit_switch_coverage(rules: list[dict]) -> list[dict]:
|
||||
"""Check that rules cover both switch ON and OFF states."""
|
||||
switch_on = False
|
||||
switch_off = False
|
||||
switch_rules = []
|
||||
|
||||
for rule in rules:
|
||||
precond = rule.get("precondition", {})
|
||||
sw = precond.get("switch", "")
|
||||
if sw == "开启":
|
||||
switch_on = True
|
||||
switch_rules.append(rule.get("rule_id", "?"))
|
||||
elif sw == "关闭":
|
||||
switch_off = True
|
||||
switch_rules.append(rule.get("rule_id", "?"))
|
||||
|
||||
status = PASS
|
||||
detail_parts = []
|
||||
if switch_on:
|
||||
detail_parts.append(f"开关=开启: 有规则覆盖")
|
||||
else:
|
||||
detail_parts.append(f"开关=开启: 未找到规则")
|
||||
status = FAIL
|
||||
if switch_off:
|
||||
detail_parts.append(f"开关=关闭: 有规则覆盖")
|
||||
else:
|
||||
detail_parts.append(f"开关=关闭: 未找到规则")
|
||||
status = FAIL
|
||||
|
||||
return [{
|
||||
"check": "开关状态完整性（开启/关闭）",
|
||||
"status": status,
|
||||
"detail": "; ".join(detail_parts),
|
||||
}]
|
||||
|
||||
|
||||
def generate_audit_report(
|
||||
rules: list[dict],
|
||||
doc: dict,
|
||||
feature_name: str,
|
||||
lt_results: list[dict],
|
||||
enum_results: list[dict],
|
||||
switch_results: list[dict],
|
||||
) -> str:
|
||||
"""Generate ir_audit_report.md in Markdown format."""
|
||||
|
||||
lines = []
|
||||
lines.append(f"# IR 完整性审计报告")
|
||||
lines.append(f"")
|
||||
lines.append(f"**功能**: {feature_name}")
|
||||
lines.append(f"**规则总数**: {len(rules)}")
|
||||
lines.append(f"**生成时间**: {__import__('datetime').datetime.now().isoformat()}")
|
||||
lines.append(f"")
|
||||
|
||||
# Human review notice
|
||||
lines.append(f"> ⚠️ **重要**: 请人工审查以下 ⚠️ 和 ❌ 项，确认是文档遗漏还是 IR 提取遗漏。")
|
||||
lines.append(f'> 如无需修改，在对应项后标注 **"已确认"**。')
|
||||
lines.append(f"")
|
||||
|
||||
# ---- Logic Tree Coverage ----
|
||||
lines.append(f"## 1. 逻辑树节点覆盖率")
|
||||
lines.append(f"")
|
||||
lines.append(f"| 图片 ID | 覆盖率 | 状态 | 详情 |")
|
||||
lines.append(f"|---------|--------|------|------|")
|
||||
for r in lt_results:
|
||||
lines.append(
|
||||
f"| {r['image_id']} | {r['coverage_pct']}% | {r['status']} | {r['detail']} |"
|
||||
)
|
||||
lines.append(f"")
|
||||
|
||||
# Uncovered node details
|
||||
for r in lt_results:
|
||||
if r["uncovered_decisions"] or r["uncovered_actions"]:
|
||||
lines.append(f"### {r['image_id']} 未覆盖节点详情")
|
||||
lines.append(f"")
|
||||
for n in r.get("uncovered_decisions", []):
|
||||
lines.append(f"- **Decision** `{n['id']}`: {n.get('condition', '?')}")
|
||||
for n in r.get("uncovered_actions", []):
|
||||
lines.append(f"- **Action** `{n['id']}`: {n.get('description', '?')}")
|
||||
lines.append(f"")
|
||||
|
||||
# ---- Table Enumeration Coverage ----
|
||||
lines.append(f"## 2. 表格枚举覆盖")
|
||||
lines.append(f"")
|
||||
lines.append(f"| 检查项 | 状态 | 详情 |")
|
||||
lines.append(f"|--------|------|------|")
|
||||
for r in enum_results:
|
||||
lines.append(f"| {r['check']} | {r['status']} | {r['detail']} |")
|
||||
lines.append(f"")
|
||||
|
||||
# ---- Switch Coverage ----
|
||||
lines.append(f"## 3. 全局开关状态覆盖")
|
||||
lines.append(f"")
|
||||
lines.append(f"| 检查项 | 状态 | 详情 |")
|
||||
lines.append(f"|--------|------|------|")
|
||||
for r in switch_results:
|
||||
lines.append(f"| {r['check']} | {r['status']} | {r['detail']} |")
|
||||
lines.append(f"")
|
||||
|
||||
# ---- Rule Summary ----
|
||||
lines.append(f"## 4. 规则清单")
|
||||
lines.append(f"")
|
||||
lines.append(f"| rule_id | Priority | 简述 |")
|
||||
lines.append(f"|---------|----------|------|")
|
||||
for rule in rules:
|
||||
desc = rule.get("description", "")[:80]
|
||||
lines.append(f"| {rule.get('rule_id', '?')} | {rule.get('priority', '?')} | {desc} |")
|
||||
lines.append(f"")
|
||||
|
||||
return "\n".join(lines)
|
||||
|
||||
|
||||
def main():
|
||||
print("=" * 60)
|
||||
print("阶段三：确定性合并与完整性校验")
|
||||
print("=" * 60)
|
||||
|
||||
# 1. Load inputs
|
||||
print(f"\n[1/5] 加载输入...")
|
||||
fragments = load_fragments()
|
||||
doc = config.load_input_document()
|
||||
semantic_index = load_semantic_index()
|
||||
feature_name = semantic_index.get("feature_name", "行车娱乐限制")
|
||||
feature_id = "DRL-001"
|
||||
print(f" 功能: {feature_name} ({feature_id})")
|
||||
print(f" 片段数: {len(fragments)}")
|
||||
|
||||
# 2. Merge rules
|
||||
print(f"\n[2/5] 合并去重...")
|
||||
merged_rules = merge_rules(fragments)
|
||||
|
||||
# 3. Reassign rule IDs
|
||||
print(f"\n[3/5] 重分配 rule_id...")
|
||||
final_rules = assign_rule_ids(merged_rules, feature_id)
|
||||
print(f" 已分配 {len(final_rules)} 个稳定 ID")
|
||||
|
||||
# Collect top-level metadata
|
||||
ir_final = {
|
||||
"feature": feature_name,
|
||||
"feature_id": feature_id,
|
||||
"rules": final_rules,
|
||||
}
|
||||
|
||||
# Save ir_final.json
|
||||
print(f"\n[4/5] 生成审计报告...")
|
||||
lt_results = audit_logic_tree_coverage(doc, final_rules)
|
||||
enum_results = audit_table_enums(final_rules, doc)
|
||||
switch_results = audit_switch_coverage(final_rules)
|
||||
|
||||
report = generate_audit_report(
|
||||
final_rules, doc, feature_name,
|
||||
lt_results, enum_results, switch_results
|
||||
)
|
||||
|
||||
# 5. Save outputs
|
||||
print(f"\n[5/5] 保存输出...")
|
||||
config.save_json(ir_final, config.IR_FINAL_JSON)
|
||||
print(f" IR: {config.IR_FINAL_JSON}")
|
||||
|
||||
with open(config.IR_AUDIT_REPORT_MD, "w", encoding="utf-8") as f:
|
||||
f.write(report)
|
||||
print(f" 审计报告: {config.IR_AUDIT_REPORT_MD}")
|
||||
|
||||
# Print quick summary
|
||||
print(f"\n完成!")
|
||||
issue_count = sum(
|
||||
1 for r in lt_results + enum_results + switch_results
|
||||
if r["status"] in (WARN, FAIL)
|
||||
)
|
||||
print(f" 规则: {len(final_rules)} 条")
|
||||
print(f" 审计问题: {issue_count} 个需要人工审查")
|
||||
|
||||
if issue_count > 0:
|
||||
print(f"\n 请查看 {config.IR_AUDIT_REPORT_MD} 并审查标记项。")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
+238
@@ -0,0 +1,238 @@
|
||||
"""
|
||||
Tests for Stage 1 (Semantic Index).
|
||||
|
||||
Validates that the generated semantic_index.json meets all completeness
|
||||
and structural requirements.
|
||||
"""
|
||||
|
||||
import json
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
# Allow running from project root or tests/ directory
|
||||
sys.path.insert(0, str(Path(__file__).parent.parent))
|
||||
import config
|
||||
|
||||
|
||||
PASS = "[PASS]"
|
||||
FAIL = "[FAIL]"
|
||||
WARN = "[WARN]"
|
||||
|
||||
|
||||
def load_inputs():
|
||||
"""Load semantic_index.json and the original parsed document."""
|
||||
try:
|
||||
si = config.load_json(config.SEMANTIC_INDEX_JSON)
|
||||
except FileNotFoundError:
|
||||
print(f"{FAIL} semantic_index.json 未找到: {config.SEMANTIC_INDEX_JSON}")
|
||||
print(" 请先运行 step1_semantic_index.py")
|
||||
sys.exit(1)
|
||||
|
||||
doc = config.load_input_document()
|
||||
return si, doc
|
||||
|
||||
|
||||
def build_image_index(doc: dict) -> dict[str, dict]:
|
||||
"""Build lookup: image rId -> image_analysis entry."""
|
||||
idx = {}
|
||||
for img in doc.get("image_analysis", []):
|
||||
rid = img.get("rid", "")
|
||||
if rid:
|
||||
idx[rid] = img
|
||||
return idx
|
||||
|
||||
|
||||
def build_logic_tree_node_index(doc: dict) -> dict[str, set[str]]:
|
||||
"""Build lookup: image rId -> set of all node IDs in that logic_tree."""
|
||||
idx = {}
|
||||
for img in doc.get("image_analysis", []):
|
||||
rid = img.get("rid", "")
|
||||
lt = img.get("logic_tree")
|
||||
if lt and rid:
|
||||
node_ids = {n["id"] for n in lt.get("nodes", [])}
|
||||
idx[rid] = node_ids
|
||||
return idx
|
||||
|
||||
|
||||
def check_unit_ids(units: list[dict]) -> list[str]:
|
||||
"""Check that every function_unit has a non-empty unit_id and name."""
|
||||
errors = []
|
||||
seen_ids = set()
|
||||
for i, fu in enumerate(units):
|
||||
uid = fu.get("unit_id", "")
|
||||
name = fu.get("name", "")
|
||||
if not uid:
|
||||
errors.append(f"function_unit[{i}]: unit_id 为空")
|
||||
elif uid in seen_ids:
|
||||
errors.append(f"function_unit[{i}]: unit_id '{uid}' 重复")
|
||||
seen_ids.add(uid)
|
||||
if not name:
|
||||
errors.append(f"function_unit[{i}] ({uid}): name 为空")
|
||||
return errors
|
||||
|
||||
|
||||
def check_sources_exist(
|
||||
units: list[dict], image_index: dict[str, dict], node_index: dict[str, set[str]]
|
||||
) -> list[str]:
|
||||
"""Check that all source references point to real content."""
|
||||
errors = []
|
||||
|
||||
for fu in units:
|
||||
uid = fu.get("unit_id", "?")
|
||||
sources = fu.get("sources", [])
|
||||
if not sources:
|
||||
errors.append(f"{uid}: sources 为空，必须至少引用一张图片或一段文字")
|
||||
continue
|
||||
|
||||
has_text = False
|
||||
has_image = False
|
||||
|
||||
for j, src in enumerate(sources):
|
||||
src_type = src.get("type", "")
|
||||
if src_type in ("table", "para"):
|
||||
has_text = True
|
||||
section = src.get("section", "")
|
||||
if not section:
|
||||
errors.append(f"{uid}.sources[{j}]: 缺少 section")
|
||||
elif src_type == "logic_tree":
|
||||
has_image = True
|
||||
image_id = src.get("image_id", "")
|
||||
if not image_id:
|
||||
errors.append(f"{uid}.sources[{j}]: logic_tree 缺少 image_id")
|
||||
continue
|
||||
|
||||
# Check image exists
|
||||
if image_id not in image_index:
|
||||
errors.append(
|
||||
f"{uid}.sources[{j}]: image_id '{image_id}' "
|
||||
f"在 image_analysis 中不存在"
|
||||
)
|
||||
continue
|
||||
|
||||
# Check logic_tree_nodes if provided
|
||||
node_ids = src.get("logic_tree_nodes", [])
|
||||
if node_ids and image_id in node_index:
|
||||
valid_nodes = node_index[image_id]
|
||||
for nid in node_ids:
|
||||
if nid not in valid_nodes:
|
||||
errors.append(
|
||||
f"{uid}.sources[{j}]: 节点 '{nid}' 在 "
|
||||
f"{image_id} 的逻辑树中不存在"
|
||||
)
|
||||
elif not node_ids:
|
||||
errors.append(
|
||||
f"{uid}.sources[{j}]: logic_tree 类型但未提供 logic_tree_nodes"
|
||||
)
|
||||
|
||||
if not has_text and not has_image:
|
||||
errors.append(f"{uid}: 必须至少引用一个文本或图片来源")
|
||||
|
||||
return errors
|
||||
|
||||
|
||||
def check_logic_tree_coverage(
|
||||
units: list[dict], node_index: dict[str, set[str]]
|
||||
) -> list[str]:
|
||||
"""Check that decision and action nodes in logic trees are covered."""
|
||||
warnings = []
|
||||
|
||||
for image_id, all_nodes in node_index.items():
|
||||
# Collect all nodes referenced across all function_units for this image
|
||||
referenced = set()
|
||||
for fu in units:
|
||||
for src in fu.get("sources", []):
|
||||
if src.get("image_id") == image_id:
|
||||
for nid in src.get("logic_tree_nodes", []):
|
||||
referenced.add(nid)
|
||||
|
||||
uncovered = all_nodes - referenced
|
||||
if uncovered:
|
||||
# Get node types from the document
|
||||
doc = config.load_input_document()
|
||||
node_types = {}
|
||||
for img in doc.get("image_analysis", []):
|
||||
if img.get("rid") == image_id:
|
||||
lt = img.get("logic_tree", {})
|
||||
for n in lt.get("nodes", []):
|
||||
node_types[n["id"]] = n.get("type", "?")
|
||||
break
|
||||
|
||||
decision_action_uncovered = [
|
||||
n for n in uncovered if node_types.get(n) in ("decision", "action")
|
||||
]
|
||||
if decision_action_uncovered:
|
||||
warnings.append(
|
||||
f"{image_id}: {len(decision_action_uncovered)} 个 "
|
||||
f"decision/action 节点未被引用: {decision_action_uncovered}"
|
||||
)
|
||||
|
||||
return warnings
|
||||
|
||||
|
||||
def run_all_tests():
|
||||
print("=" * 60)
|
||||
print("Step 1 自检测试")
|
||||
print("=" * 60)
|
||||
|
||||
si, doc = load_inputs()
|
||||
units = si.get("function_units", [])
|
||||
image_index = build_image_index(doc)
|
||||
node_index = build_logic_tree_node_index(doc)
|
||||
|
||||
all_errors = []
|
||||
all_warnings = []
|
||||
|
||||
# Test 1: unit_id and name validity
|
||||
errors = check_unit_ids(units)
|
||||
if errors:
|
||||
print(f"\n{FAIL} unit_id/name 检查: {len(errors)} 个错误")
|
||||
for e in errors:
|
||||
print(f" - {e}")
|
||||
all_errors.extend(errors)
|
||||
else:
|
||||
print(f"\n{PASS} unit_id/name 检查: 全部通过 ({len(units)} 个功能单元)")
|
||||
|
||||
# Test 2: source references exist
|
||||
errors = check_sources_exist(units, image_index, node_index)
|
||||
if errors:
|
||||
print(f"\n{FAIL} 来源引用检查: {len(errors)} 个错误")
|
||||
for e in errors:
|
||||
print(f" - {e}")
|
||||
all_errors.extend(errors)
|
||||
else:
|
||||
print(f"\n{PASS} 来源引用检查: 全部通过")
|
||||
|
||||
# Test 3: Logic tree coverage
|
||||
warnings = check_logic_tree_coverage(units, node_index)
|
||||
if warnings:
|
||||
print(f"\n{WARN} 逻辑树节点覆盖率: {len(warnings)} 个警告")
|
||||
for w in warnings:
|
||||
print(f" - {w}")
|
||||
all_warnings.extend(warnings)
|
||||
else:
|
||||
print(f"\n{PASS} 逻辑树节点覆盖率: 全部通过")
|
||||
|
||||
# Summary
|
||||
print(f"\n{'='*60}")
|
||||
total_failures = len(all_errors)
|
||||
total_warnings = len(all_warnings)
|
||||
|
||||
if total_failures == 0 and total_warnings == 0:
|
||||
print(f"{PASS} 所有测试通过!")
|
||||
elif total_failures == 0:
|
||||
print(f"{WARN} 全部通过但有 {total_warnings} 个警告")
|
||||
else:
|
||||
print(f"{FAIL} 测试失败: {total_failures} 个错误, {total_warnings} 个警告")
|
||||
print("\n请检查 LLM 输出质量，可能需要调整 Prompt 并重新运行 step1_semantic_index.py")
|
||||
|
||||
print(f"\n统计:")
|
||||
print(f" 功能单元数: {len(units)}")
|
||||
print(f" 概念数: {len(si.get('concepts', []))}")
|
||||
print(f" 逻辑树图片数: {len(node_index)}")
|
||||
|
||||
return total_failures == 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
success = run_all_tests()
|
||||
sys.exit(0 if success else 1)
|
||||
+228
@@ -0,0 +1,228 @@
|
||||
"""
|
||||
Tests for Stage 2 (IR Extraction).
|
||||
|
||||
Validates that ir_fragments.json meets quality and structural requirements:
|
||||
- All fragments have non-empty rules
|
||||
- All rules have source references with logic tree nodes
|
||||
- All trigger conditions have signal/operator/value
|
||||
- No duplicate rule_ids (across all fragments)
|
||||
"""
|
||||
|
||||
import json
|
||||
import sys
|
||||
from pathlib import Path
|
||||
from collections import Counter
|
||||
|
||||
sys.path.insert(0, str(Path(__file__).parent.parent))
|
||||
import config
|
||||
|
||||
|
||||
PASS = "[PASS]"
|
||||
FAIL = "[FAIL]"
|
||||
WARN = "[WARN]"
|
||||
|
||||
|
||||
def load_fragments():
|
||||
"""Load ir_fragments.json."""
|
||||
try:
|
||||
return config.load_json(config.IR_FRAGMENTS_JSON)
|
||||
except FileNotFoundError:
|
||||
print(f"{FAIL} ir_fragments.json 未找到: {config.IR_FRAGMENTS_JSON}")
|
||||
print(" 请先运行 step2_ir_extraction.py")
|
||||
sys.exit(1)
|
||||
|
||||
|
||||
def check_non_empty_rules(fragments: list[dict]) -> list[str]:
|
||||
"""Every fragment must have at least one rule."""
|
||||
errors = []
|
||||
for f in fragments:
|
||||
uid = f.get("unit_id", "?")
|
||||
rules = f.get("rules", [])
|
||||
if not rules:
|
||||
if f.get("error"):
|
||||
errors.append(f"{uid}: 提取失败 — {f['error']}")
|
||||
else:
|
||||
errors.append(f"{uid}: rules 为空")
|
||||
return errors
|
||||
|
||||
|
||||
def check_sources_have_logic_tree_nodes(fragments: list[dict]) -> list[str]:
|
||||
"""Every rule should reference at least one logic tree node in its sources."""
|
||||
errors = []
|
||||
for f in fragments:
|
||||
uid = f.get("unit_id", "?")
|
||||
for j, rule in enumerate(f.get("rules", [])):
|
||||
rid = rule.get("rule_id", f"rule[{j}]")
|
||||
sources = rule.get("sources", [])
|
||||
has_logic_tree = any(
|
||||
src.get("type") == "logic_tree" and src.get("node_ids")
|
||||
for src in sources
|
||||
)
|
||||
# NOTE: Some rules might only reference text (e.g., switch-off rules),
|
||||
# so we flag as warning rather than error
|
||||
if not has_logic_tree:
|
||||
has_text = any(
|
||||
src.get("type") in ("table", "para") for src in sources
|
||||
)
|
||||
if not has_text:
|
||||
errors.append(f"{rid}: sources 中既无逻辑树引用也无文字引用")
|
||||
return errors
|
||||
|
||||
|
||||
def check_trigger_conditions(fragments: list[dict]) -> list[str]:
|
||||
"""Every trigger condition must have signal, operator, value."""
|
||||
errors = []
|
||||
for f in fragments:
|
||||
uid = f.get("unit_id", "?")
|
||||
for j, rule in enumerate(f.get("rules", [])):
|
||||
rid = rule.get("rule_id", f"rule[{j}]")
|
||||
trigger = rule.get("trigger", {})
|
||||
conditions = trigger.get("conditions", [])
|
||||
|
||||
# Check if trigger uses 'event' instead of conditions
|
||||
if trigger.get("event") is not None:
|
||||
continue # event-based trigger is valid
|
||||
|
||||
for k, cond in enumerate(conditions):
|
||||
signal = cond.get("signal", "")
|
||||
operator = cond.get("operator", "")
|
||||
# value can be 0, False, "", so check with 'in'
|
||||
has_value = "value" in cond
|
||||
|
||||
if not signal:
|
||||
errors.append(f"{rid}.condition[{k}]: 缺少 signal")
|
||||
if not operator:
|
||||
errors.append(f"{rid}.condition[{k}]: 缺少 operator")
|
||||
if not has_value:
|
||||
errors.append(f"{rid}.condition[{k}]: 缺少 value")
|
||||
|
||||
return errors
|
||||
|
||||
|
||||
def check_duplicate_rule_ids(fragments: list[dict]) -> list[str]:
|
||||
"""Check for duplicate rule_ids across all fragments."""
|
||||
all_rule_ids = []
|
||||
for f in fragments:
|
||||
for rule in f.get("rules", []):
|
||||
rid = rule.get("rule_id", "")
|
||||
if rid:
|
||||
all_rule_ids.append(rid)
|
||||
|
||||
duplicates = [rid for rid, count in Counter(all_rule_ids).items() if count > 1]
|
||||
errors = []
|
||||
if duplicates:
|
||||
errors.append(f"重复 rule_id: {duplicates}")
|
||||
return errors
|
||||
|
||||
|
||||
def check_action_types(fragments: list[dict]) -> list[str]:
|
||||
"""Verify that actions have valid types."""
|
||||
valid_types = {"system", "user_interaction"}
|
||||
errors = []
|
||||
for f in fragments:
|
||||
for j, rule in enumerate(f.get("rules", [])):
|
||||
rid = rule.get("rule_id", f"rule[{j}]")
|
||||
for k, action in enumerate(rule.get("actions", [])):
|
||||
atype = action.get("type", "")
|
||||
if atype not in valid_types:
|
||||
errors.append(
|
||||
f"{rid}.action[{k}]: type='{atype}' 无效, "
|
||||
f"应为 {valid_types}"
|
||||
)
|
||||
if atype == "user_interaction" and "content" not in action:
|
||||
errors.append(
|
||||
f"{rid}.action[{k}]: user_interaction 类型缺少 content 字段"
|
||||
)
|
||||
return errors
|
||||
|
||||
|
||||
def run_all_tests():
|
||||
print("=" * 60)
|
||||
print("Step 2 自检测试")
|
||||
print("=" * 60)
|
||||
|
||||
fragments = load_fragments()
|
||||
all_errors = []
|
||||
total_units = len(fragments)
|
||||
total_rules = sum(len(f.get("rules", [])) for f in fragments)
|
||||
|
||||
# Test 1: Non-empty rules
|
||||
errors = check_non_empty_rules(fragments)
|
||||
if errors:
|
||||
print(f"\n{FAIL} 非空规则检查: {len(errors)} 个错误")
|
||||
for e in errors:
|
||||
print(f" - {e}")
|
||||
all_errors.extend(errors)
|
||||
else:
|
||||
print(f"\n{PASS} 非空规则检查: 全部通过 ({total_units} 个片段)")
|
||||
|
||||
# Test 2: Sources have logic tree references
|
||||
errors = check_sources_have_logic_tree_nodes(fragments)
|
||||
if errors:
|
||||
print(f"\n{FAIL} 来源节点引用: {len(errors)} 个规则缺少来源引用")
|
||||
for e in errors[:10]: # Show first 10
|
||||
print(f" - {e}")
|
||||
if len(errors) > 10:
|
||||
print(f" ... 还有 {len(errors) - 10} 个")
|
||||
all_errors.extend(errors)
|
||||
else:
|
||||
print(f"\n{PASS} 来源节点引用: 全部通过")
|
||||
|
||||
# Test 3: Trigger conditions completeness
|
||||
errors = check_trigger_conditions(fragments)
|
||||
if errors:
|
||||
print(f"\n{FAIL} 触发条件完整性: {len(errors)} 个条件不完整")
|
||||
for e in errors[:10]:
|
||||
print(f" - {e}")
|
||||
if len(errors) > 10:
|
||||
print(f" ... 还有 {len(errors) - 10} 个")
|
||||
all_errors.extend(errors)
|
||||
else:
|
||||
print(f"\n{PASS} 触发条件完整性: 全部通过")
|
||||
|
||||
# Test 4: No duplicate rule_ids
|
||||
errors = check_duplicate_rule_ids(fragments)
|
||||
if errors:
|
||||
print(f"\n{FAIL} rule_id 唯一性: 发现重复")
|
||||
for e in errors:
|
||||
print(f" - {e}")
|
||||
all_errors.extend(errors)
|
||||
else:
|
||||
print(f"\n{PASS} rule_id 唯一性: 全部通过")
|
||||
|
||||
# Test 5: Valid action types
|
||||
errors = check_action_types(fragments)
|
||||
if errors:
|
||||
print(f"\n{FAIL} 动作类型检查: {len(errors)} 个问题")
|
||||
for e in errors[:10]:
|
||||
print(f" - {e}")
|
||||
all_errors.extend(errors)
|
||||
else:
|
||||
print(f"\n{PASS} 动作类型检查: 全部通过")
|
||||
|
||||
# Summary
|
||||
print(f"\n{'='*60}")
|
||||
total_failures = len(all_errors)
|
||||
|
||||
if total_failures == 0:
|
||||
print(f"{PASS} 所有测试通过!")
|
||||
else:
|
||||
print(f"{FAIL} 测试失败: {total_failures} 个错误")
|
||||
print("\n建议:")
|
||||
print(" 1. 检查 ir_fragments.json 中出错的规则")
|
||||
print(" 2. 如果某些功能单元的规则为空，检查上下文包是否丢失了关键信息")
|
||||
print(" 3. 调整 Prompt（prompts/step2_ir_extraction.txt）或上下文提取逻辑后重新运行")
|
||||
|
||||
print(f"\n统计:")
|
||||
print(f" 功能单元数: {total_units}")
|
||||
print(f" 规则总数: {total_rules}")
|
||||
error_units = sum(1 for f in fragments if f.get("error"))
|
||||
if error_units:
|
||||
print(f" 提取失败的单元: {error_units}")
|
||||
|
||||
return total_failures == 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
success = run_all_tests()
|
||||
sys.exit(0 if success else 1)
|
||||
+191
@@ -0,0 +1,191 @@
|
||||
"""
|
||||
Tests for Stage 3 (Merge & Audit).
|
||||
|
||||
Validates:
|
||||
- ir_final.json exists and is well-formed
|
||||
- No duplicate rule_ids
|
||||
- All rule_ids follow naming convention
|
||||
- ir_audit_report.md exists and contains required sections
|
||||
"""
|
||||
|
||||
import re
|
||||
import sys
|
||||
from pathlib import Path
|
||||
from collections import Counter
|
||||
|
||||
sys.path.insert(0, str(Path(__file__).parent.parent))
|
||||
import config
|
||||
|
||||
|
||||
PASS = "[PASS]"
|
||||
FAIL = "[FAIL]"
|
||||
WARN = "[WARN]"
|
||||
|
||||
|
||||
def load_ir_final():
|
||||
"""Load ir_final.json."""
|
||||
try:
|
||||
return config.load_json(config.IR_FINAL_JSON)
|
||||
except FileNotFoundError:
|
||||
print(f"{FAIL} ir_final.json 未找到: {config.IR_FINAL_JSON}")
|
||||
print(" 请先运行 step3_merge_and_audit.py")
|
||||
sys.exit(1)
|
||||
|
||||
|
||||
def load_audit_report():
|
||||
"""Load ir_audit_report.md if it exists."""
|
||||
try:
|
||||
with open(config.IR_AUDIT_REPORT_MD, "r", encoding="utf-8") as f:
|
||||
return f.read()
|
||||
except FileNotFoundError:
|
||||
print(f"{FAIL} ir_audit_report.md 未找到: {config.IR_AUDIT_REPORT_MD}")
|
||||
print(" 请先运行 step3_merge_and_audit.py")
|
||||
sys.exit(1)
|
||||
|
||||
|
||||
def check_rule_ids(ir: dict) -> list[str]:
|
||||
"""Check for duplicate rule_ids and naming convention."""
|
||||
errors = []
|
||||
rules = ir.get("rules", [])
|
||||
rule_ids = [r.get("rule_id", "") for r in rules]
|
||||
|
||||
# No duplicates
|
||||
duplicates = [rid for rid, count in Counter(rule_ids).items() if count > 1]
|
||||
if duplicates:
|
||||
errors.append(f"重复 rule_id: {duplicates}")
|
||||
|
||||
# Naming convention: FEATURE-TYPE-CAT-NN
|
||||
pattern = re.compile(r"^[A-Z]+-\d{3}-(SYS|UI|SDK)-FG-\d{2}$")
|
||||
for rid in rule_ids:
|
||||
if rid and not pattern.match(rid):
|
||||
errors.append(f"rule_id 命名不规范: '{rid}' (期望格式: DRL-001-SYS-FG-01)")
|
||||
|
||||
return errors
|
||||
|
||||
|
||||
def check_top_level_structure(ir: dict) -> list[str]:
|
||||
"""Check that ir_final has the required top-level fields."""
|
||||
errors = []
|
||||
for field in ["feature", "feature_id", "rules"]:
|
||||
if field not in ir:
|
||||
errors.append(f"ir_final 缺少顶层字段: {field}")
|
||||
|
||||
if not isinstance(ir.get("rules"), list):
|
||||
errors.append("ir_final.rules 必须是数组")
|
||||
elif len(ir["rules"]) == 0:
|
||||
errors.append("ir_final.rules 为空")
|
||||
|
||||
return errors
|
||||
|
||||
|
||||
def check_rule_completeness(rules: list[dict]) -> list[str]:
|
||||
"""Check each rule has all required fields."""
|
||||
errors = []
|
||||
required_fields = ["rule_id", "description", "priority", "sources", "trigger", "actions"]
|
||||
for i, rule in enumerate(rules):
|
||||
rid = rule.get("rule_id", f"rule[{i}]")
|
||||
for field in required_fields:
|
||||
if field not in rule:
|
||||
errors.append(f"{rid}: 缺少字段 '{field}'")
|
||||
# sources must be non-empty
|
||||
if not rule.get("sources"):
|
||||
errors.append(f"{rid}: sources 为空")
|
||||
# actions must be non-empty
|
||||
if not rule.get("actions"):
|
||||
errors.append(f"{rid}: actions 为空")
|
||||
return errors
|
||||
|
||||
|
||||
def check_audit_report(report: str) -> list[str]:
|
||||
"""Check audit report has required sections."""
|
||||
errors = []
|
||||
|
||||
required_sections = [
|
||||
"逻辑树节点覆盖率",
|
||||
"表格枚举覆盖",
|
||||
"开关状态",
|
||||
]
|
||||
for section in required_sections:
|
||||
if section not in report:
|
||||
errors.append(f"审计报告缺少章节: {section}")
|
||||
|
||||
# Should have at least one coverage percentage
|
||||
if "覆盖率" not in report and "%" not in report:
|
||||
errors.append("审计报告中未找到覆盖率统计")
|
||||
|
||||
# Should have the human review notice
|
||||
if "人工审查" not in report:
|
||||
errors.append("审计报告缺少人工审查提示")
|
||||
|
||||
return errors
|
||||
|
||||
|
||||
def run_all_tests():
|
||||
print("=" * 60)
|
||||
print("Step 3 自检测试")
|
||||
print("=" * 60)
|
||||
|
||||
ir = load_ir_final()
|
||||
report = load_audit_report()
|
||||
rules = ir.get("rules", [])
|
||||
all_errors = []
|
||||
|
||||
# Test 1: Top-level structure
|
||||
errors = check_top_level_structure(ir)
|
||||
if errors:
|
||||
print(f"\n{FAIL} 顶层结构检查: {len(errors)} 个错误")
|
||||
for e in errors:
|
||||
print(f" - {e}")
|
||||
all_errors.extend(errors)
|
||||
else:
|
||||
print(f"\n{PASS} 顶层结构检查: 通过 (feature={ir.get('feature')}, feature_id={ir.get('feature_id')})")
|
||||
|
||||
# Test 2: rule_id uniqueness and naming
|
||||
errors = check_rule_ids(ir)
|
||||
if errors:
|
||||
print(f"\n{FAIL} rule_id 检查: {len(errors)} 个错误")
|
||||
for e in errors:
|
||||
print(f" - {e}")
|
||||
all_errors.extend(errors)
|
||||
else:
|
||||
print(f"\n{PASS} rule_id 检查: 全部通过 ({len(rules)} 个唯一 ID)")
|
||||
|
||||
# Test 3: Rule field completeness
|
||||
errors = check_rule_completeness(rules)
|
||||
if errors:
|
||||
print(f"\n{FAIL} 规则字段完整性: {len(errors)} 个错误")
|
||||
for e in errors:
|
||||
print(f" - {e}")
|
||||
all_errors.extend(errors)
|
||||
else:
|
||||
print(f"\n{PASS} 规则字段完整性: 全部通过")
|
||||
|
||||
# Test 4: Audit report content
|
||||
errors = check_audit_report(report)
|
||||
if errors:
|
||||
print(f"\n{FAIL} 审计报告检查: {len(errors)} 个错误")
|
||||
for e in errors:
|
||||
print(f" - {e}")
|
||||
all_errors.extend(errors)
|
||||
else:
|
||||
print(f"\n{PASS} 审计报告检查: 全部通过")
|
||||
|
||||
# Summary
|
||||
print(f"\n{'='*60}")
|
||||
total_failures = len(all_errors)
|
||||
|
||||
if total_failures == 0:
|
||||
print(f"{PASS} 所有测试通过!")
|
||||
print(f"\n最终交付物:")
|
||||
print(f" - {config.IR_FINAL_JSON} ({len(rules)} 条规则)")
|
||||
print(f" - {config.IR_AUDIT_REPORT_MD}")
|
||||
else:
|
||||
print(f"{FAIL} 测试失败: {total_failures} 个错误")
|
||||
print("\n建议: 检查 ir_fragments.json 和合并逻辑，修复问题后重新运行 step3_merge_and_audit.py")
|
||||
|
||||
return total_failures == 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
success = run_all_tests()
|
||||
sys.exit(0 if success else 1)
|
||||
Reference in New Issue
Block a user