docs: Phase1 里程碑1 实现设计(骨架+数据模型+config 加载)

This commit is contained in:
lhl
2026-08-08 15:16:19 +08:00
commit 56da6c9917
2 changed files with 230 additions and 0 deletions
+31
View File
@@ -0,0 +1,31 @@
# ---- 环境与密钥 ----
.env
.env.*
# ---- 运行时数据 ----
data/
# ---- Python 构建产物 ----
__pycache__/
*.py[cod]
*.egg-info/
dist/
build/
.eggs/
# ---- 测试缓存 ----
.pytest_cache/
.coverage
htmlcov/
# ---- 虚拟环境 ----
.venv/
venv/
# ---- 工具内部操作产物(不入库) ----
conversations/
.aura/
# ---- IDE / 编辑器 ----
.vscode/
.idea/
@@ -0,0 +1,199 @@
# 设计文档:Phase 1 里程碑 1(项目骨架 + 数据模型 + 配置加载)
> 版本: v1.0 | 日期: 2026-08-08 | 状态: 已批准
>
> 本文档是 `docs/implementation-plan.md` 阶段 1(项目基盘)**里程碑 1** 的实现设计,覆盖任务 1.1(项目结构/打包)与 1.2(共通数据模型定义),并前置 config 加载(对应 `docs/config-design.md` 的落地)。
---
## 1. 目标与范围
**本里程碑交付**
- 可运行的 Python 项目骨架(src 布局 + pyproject 打包 + pytest 就绪)
- 全部业务数据模型 `data_models.py`design §3 + §9.4
- 配置加载 `config.py`app / inference / rag 三 yaml + .env + 环境变量)
**不包含**(后续里程碑):FileReader / CodeParser / ImageAnalyzer / InferenceEngine / ToolExecutor / 状态机 / 记忆 / 可观测性。
**完成门槛**`pytest` 全绿;`pip install -e .``import genesis` 可用。
---
## 2. 项目结构与打包
```
Genesis/
├── pyproject.toml
├── src/genesis/
│ ├── __init__.py # __version__ = "0.1.0"
│ ├── data_models.py
│ └── config.py
├── tests/
│ ├── __init__.py
│ ├── fixtures/
│ │ ├── app_min.yaml
│ │ ├── inference_min.yaml
│ │ ├── rag_min.yaml
│ │ └── .env.test
│ ├── test_data_models.py
│ └── test_config.py
├── .gitignore
└── README.md # 空壳占位(README 在成果物阶段补全)
```
### pyproject.toml 要点
```toml
[build-system]
requires = ["setuptools>=69"]
build-backend = "setuptools.build_meta"
[project]
name = "genesis"
version = "0.1.0"
requires-python = ">=3.11"
dependencies = [
"pydantic>=2.13",
"pydantic-settings>=2.15",
"pyyaml>=6.0",
"openpyxl>=3.1",
"python-docx>=1.0",
]
[project.optional-dependencies]
dev = ["pytest>=8.0"]
[tool.setuptools.packages.find]
where = ["src"]
[tool.pytest.ini_options]
testpaths = ["tests"]
```
### .gitignore 要点
`.env``data/``*.egg-info/``__pycache__/``.pytest_cache/``.venv/`
---
## 3. data_models.py
**约定**
- 文件顶部 `from __future__ import annotations`design §9.4.6P0-1 落地)
- 全部 `@dataclass`,字段名/默认值与 design §9.4 一一对应
- 不做自写序列化;`dataclasses.asdict` 即可 JSON 化(后续里程碑按需)
- runtime 层类型(ChatResult 等)不在本文件,避免重复定义
### 3.1 枚举
```python
class SheetType(Enum): FUNCTION / SCREEN / REPORT / DATABASE / INTERFACE / BATCH / MASTER / GENERIC
class ElementType(Enum): FUNCTION="機能" / SCREEN="画面" / REPORT="帳票" / DB="DB" / IF="IF" / BATCH="バッチ"
class RelationType(Enum): USE="利用" / REFER="参照" / UPDATE="更新" / OUTPUT="输出" / INPUT="输入" / DEPEND="依赖"
class Confidence(Enum): HIGH="high" / MEDIUM="medium" / LOW="low"
class ExtractionMethod(Enum): OPENPYXL="openpyxl" / LLM_FROM_FREE_TEXT="llm_from_free_text"
```
### 3.2 基础类型
```python
Provenance(file_name, sheet_name, row:int, column:str, column_header:str)
CellFormatting(strikethrough=False, font_color=None, bg_color=None)
CellComment(author, text, source_uri)
CellValue(value, provenance, formatting=None, comment=None)
```
### 3.3 Parser 输出类型
```python
ExcelTable(name, detected_type: SheetType, extraction_method: str, headers: list[str], rows: list[dict[str, CellValue]])
ChapterMarker(type: str, name: str, level: int)
ParsedTemplate(file_name, sections: list[ChapterMarker], placeholders: dict[str,str], styles: dict)
RuleDocument(file_name, category, markdown_content, source_path, file_type, hash)
ImageAnalysis(image_ref, description, confidence, source_uri, sheet_name, anchor_cell, status, nearby_text="")
ControllerInfo(name, class_name, path, base_path, endpoints, source_uri)
ServiceInfo(name, class_name, path, methods, source_uri)
EntityInfo(name, class_name, path, table_name, fields, source_uri)
EndpointInfo(method, path, controller, description, source_uri)
ExistingSystemInfo(controller_layer, service_layer, entity_layer, api_endpoints, source_path)
```
### 3.4 工具层与汇聚
```python
UnifiedDocument(file_name, file_type, source_path, content_type,
tables=None, sheet_names=None, paragraphs=None, slides=None, text=None, encoding=None)
CodeStructure(root_path, language, modules, classes,
controllers, services, entities, endpoints, raw_imports)
ImageDescription(image_ref, description, objects, ocr_text, confidence, model)
StructuredSource(tables: list[ExcelTable], template: ParsedTemplate,
rule_docs: list[RuleDocument], image_analyses: list[ImageAnalysis],
existing_system: ExistingSystemInfo|None, comments: list[CellComment])
```
---
## 4. config.py
**设计依据**config-design §1(优先级:环境变量 > yaml > 默认值)、§7(校验与脱敏)。
### 4.1 结构
```python
# 三个 yaml 对应模型
class ServerConfig(BaseModel): max_upload_mb=100; allowed_extensions=[...]
class AppConfig(BaseModel): app; server: ServerConfig; session; paths; task_queue
class InferenceConfig(BaseModel): models(primary/fallback/vision); llm_calls; structured_output; prompt_registry
class RagConfig(BaseModel): embedding; vector_store; chunking; retrieval
# 根 Settings
class Settings(BaseSettings):
model_config = SettingsConfigDict(env_prefix="GENESIS_", env_file=".env", extra="ignore")
app: AppConfig
inference: InferenceConfig
rag: RagConfig
@classmethod
def from_dir(cls, config_dir: Path) -> Settings: ... # 读三 yaml 构造;缺失→默认
def get_redacted(self) -> dict: ... # key/secret/token → "***"
```
### 4.2 行为约定
- yaml 文件缺失/字段缺失 → 用模型默认值兜底,不阻塞启动
- 非法类型/越界值 → pydantic 校验失败并抛出带原因的异常(config-design §7
- API Key 等敏感项从 `.env`/环境变量读,**不入 yaml**;缺失不阻塞启动
- `get_redacted()``/api/settings` 脱敏展示
---
## 5. 测试策略
### test_data_models.py
- 枚举成员与值断言(SheetType 8 个、ElementType 等)
- 关键 dataclass 可用关键字实例化(覆盖**前向引用**ExcelTable 引用 SheetType、StructuredSource 引用各 Parser 类型)
- 默认值生效(CellFormatting、ImageAnalysis.nearby_text
- `dataclasses.asdict` 序列化冒烟
### test_config.py
- fixtures 最小 yaml → Settings.from_dir 构造 → 字段映射断言
- 最小 yaml(几乎空)→ 默认值兜底
- 环境变量覆盖(monkeypatch `GENESIS_APP__SERVER__MAX_UPLOAD_MB` 等)
- `get_redacted()` 脱敏(api_key/secret/token → ***
---
## 6. 里程碑完成门槛
- [ ] `pip install -e ".[dev]"` 成功,`import genesis` 可用
- [ ] `pytest` 全绿
- [ ] 数据模型字段与 design §9.4 一一对应(无遗漏/无多余)
- [ ] config 三 yaml + env + 脱敏行为符合 config-design §1/§7
---
## 7. 提交策略
- git init 后:设计文档单独 commit → 骨架/打包 commit → data_models + 测试 commit → config + 测试 commit
- 每个 commit 前跑 pytest 确保绿