117 lines
4.0 KiB
Markdown
117 lines
4.0 KiB
Markdown
# 规则 Schema 工作流检查摘要 v0.1
|
||
|
||
## Completed
|
||
|
||
- 新增 GPT 规划本地化协议。
|
||
- 新增模型抽取规则。
|
||
- 新增模型卡 Markdown contract。
|
||
- 新增模型抽取 workflow。
|
||
- 新增机器模型索引 `models/model_index.json`。
|
||
- 新增人读模型卡索引 `cards/card_index.md`。
|
||
- 新增权威机器模型卡 schema `schemas/model_card.schema.json`。
|
||
- 新增 `schemas/model_index.schema.json`。
|
||
- 新增 `schemas/card_index.schema.json`。
|
||
- 为 source article、source excerpt、regression case schema 补充中文 `description`。
|
||
- 为 regression case schema 增加 `case_type` 枚举：`positive`、`boundary`、`misuse`。
|
||
- 升级 `scripts/validate_model_library.py`，增加本地 contract 子集、index、回归覆盖检查。
|
||
- 新增 `scripts/check_card_contract.py`。
|
||
- 新增 `scripts/run_selector_demo.py`。
|
||
- 新增和更新单元测试。
|
||
- 更新 README、workflow、schema/scripts/selector/cards/models 目录说明。
|
||
|
||
## Files Changed
|
||
|
||
Created:
|
||
|
||
- `docs/MODEL_CARD_CONTRACT.md`
|
||
- `docs/MODEL_EXTRACTION_WORKFLOW.md`
|
||
- `schemas/model_card.schema.json`
|
||
- `schemas/model_index.schema.json`
|
||
- `schemas/card_index.schema.json`
|
||
- `models/model_index.json`
|
||
- `cards/card_index.md`
|
||
- `scripts/check_card_contract.py`
|
||
- `scripts/run_selector_demo.py`
|
||
- `tests/test_card_contract.py`
|
||
- `reports/规则Schema工作流检查摘要_v0.1.md`
|
||
|
||
Modified:
|
||
|
||
- `README.md`
|
||
- `docs/WORKFLOW.md`
|
||
- `schemas/README.md`
|
||
- `schemas/source_article.schema.json`
|
||
- `schemas/source_excerpt.schema.json`
|
||
- `schemas/regression_case.schema.json`
|
||
- `scripts/README.md`
|
||
- `scripts/validate_model_library.py`
|
||
- `selector/README.md`
|
||
- `models/README.md`
|
||
- `cards/README.md`
|
||
- `tests/test_validate_model_library.py`
|
||
- `reports/validation_report.md`
|
||
|
||
## Checks Run
|
||
|
||
```powershell
|
||
python -m unittest discover -s tests -p "test*.py" -v
|
||
```
|
||
|
||
Result:
|
||
|
||
- PASS
|
||
- 9 tests run.
|
||
|
||
```powershell
|
||
python scripts\check_card_contract.py
|
||
```
|
||
|
||
Result:
|
||
|
||
- FAIL as expected.
|
||
- Current `cards/qpi.md` and `cards/intellectual_archaeology.md` are still `draft_pre_contract` and do not yet contain all required headings.
|
||
- `cards/README.md` is now ignored by the checker.
|
||
|
||
```powershell
|
||
python scripts\validate_model_library.py
|
||
```
|
||
|
||
Result:
|
||
|
||
- FAIL as expected.
|
||
- Current `models/qpi.model.json` and `models/intellectual_archaeology.model.json` are still `draft_pre_contract`.
|
||
- Main failures are missing recommended/required GPT-plan fields, invalid enum values, invalid `selection_priority` range, and incomplete `stability_profile`.
|
||
|
||
```powershell
|
||
python scripts\run_selector_demo.py
|
||
```
|
||
|
||
Result:
|
||
|
||
- PASS.
|
||
- Demo outputs recommended model IDs, scores, reasons, and routing notes.
|
||
- Scores are low because current model JSON has not yet been repaired with trigger keywords and 1-10 priority values.
|
||
|
||
## Known Limits
|
||
|
||
- `scripts/validate_model_library.py` implements the local schema subset required by current contracts. It does not implement a complete JSON Schema draft 2020-12 engine.
|
||
- Current model/card content is intentionally not repaired in this pass.
|
||
- `models/model_index.json` and `cards/card_index.md` mark both sample models as `draft_pre_contract`.
|
||
- Selector demo works, but its output quality depends on model JSON fields that are not yet repaired.
|
||
- `card_index.schema.json` defines card index metadata, but `cards/card_index.md` itself is Markdown and is checked by local tooling rather than full JSON Schema.
|
||
|
||
## Owner Review Checklist
|
||
|
||
- GPT 规划是否已正确本地化？
|
||
- card contract 是否覆盖模型卡样例？
|
||
- model schema 是否覆盖机器 JSON 样例？
|
||
- index 设计是否满足当前和未来扩展？
|
||
- 工作流 gate 是否足以防止直接跳内容？
|
||
- 工具需求是否足够但不过度？
|
||
- 是否接受 `tests/regression_cases.json` 继续作为合并式 regression case 文件？
|
||
- 是否接受 `model_spec.schema.json` 暂时保留为 pre-contract legacy schema？
|
||
|
||
## Do Not Proceed Before Owner Confirms
|
||
|
||
Do not repair QPI or Intellectual Archaeology content until this foundation is reviewed or the owner explicitly authorizes content repair.
|