Keep raw open-music assets out of normal git history

Constraint: The user wants real datasets added locally and potentially pushed, which would make ordinary git history fragile without LFS guardrails
Rejected: Download first and retrofit tracking later | Risks oversized commits and inconsistent reproducibility rules
Confidence: high
Scope-risk: narrow
Directive: Route all future raw corpus archives and audio under acr-engine/data/raw through LFS unless a smaller manifest-only alternative is explicitly chosen
Tested: git lfs version; git check-attr filter -- acr-engine/data/raw/fma_small_audio/example.wav; git check-attr filter -- acr-engine/data/raw/archive.zip
Not-tested: Actual large-file add/push against remote LFS storage remains pending until real dataset files are downloaded

+acr-engine/data/raw/**/*.zip filter=lfs diff=lfs merge=lfs -text
+acr-engine/data/raw/**/*.tar filter=lfs diff=lfs merge=lfs -text
+acr-engine/data/raw/**/*.tar.gz filter=lfs diff=lfs merge=lfs -text
+acr-engine/data/raw/**/*.tgz filter=lfs diff=lfs merge=lfs -text
+acr-engine/data/raw/**/*.wav filter=lfs diff=lfs merge=lfs -text
+acr-engine/data/raw/**/*.mp3 filter=lfs diff=lfs merge=lfs -text
+acr-engine/data/raw/**/*.flac filter=lfs diff=lfs merge=lfs -text
+acr-engine/data/raw/**/*.ogg filter=lfs diff=lfs merge=lfs -text

-# Local Open-Music Drop Zones
+# Raw Open Music Drop Zones
 
 
-Put real downloaded open-music audio files here before running the one-shot smoke flow.
+## One-screen summary
 
 
-## Recommended folders
-- `data/raw/fma_small_audio/`
-- `data/raw/mtg_jamendo_audio/`
+| Dataset | Local directory | Primary use here | Current status | License note |
+|---|---|---|---|---|
+| FMA Small | [fma_small_audio/](./fma_small_audio/) | first real-data smoke + training baseline | not downloaded | track-level terms vary; verify before broader use |
+| MTG-Jamendo | [mtg_jamendo_audio/](./mtg_jamendo_audio/) | retrieval/eval smoke and corpus compatibility checks | not downloaded | common usage path is research-oriented; do **not** treat as commercial-ready by default |
+
+## Recommended order
+
+```mermaid
+flowchart LR
+    A[Download / place local audio] --> B[check-local-ready]
+    B --> C[inspect-local]
+    C --> D[smoke-local]
+```
+
+## Minimal commands
+
+### 1) Verify the folder is actually usable
 
 
-## Next command
-For FMA:
 ```bash
 ```bash
-/usr/local/miniconda3/bin/python src/data/external_adapters.py smoke-local fma data/raw/fma_small_audio --output-root data/external_smoke --eval-ratio 0.2 --query-duration 8.0 --train-epochs 1 --batch-size 2
+/usr/local/miniconda3/bin/python src/data/external_adapters.py check-local-ready fma data/raw/fma_small_audio --eval-ratio 0.2 --query-duration 8.0
+/usr/local/miniconda3/bin/python src/data/external_adapters.py check-local-ready mtg_jamendo data/raw/mtg_jamendo_audio --eval-ratio 0.2 --query-duration 8.0
 ```
 ```
 
 
-For MTG-Jamendo:
+### 2) Run the full smoke pipeline
+
 ```bash
 ```bash
+/usr/local/miniconda3/bin/python src/data/external_adapters.py smoke-local fma data/raw/fma_small_audio --output-root data/external_smoke --eval-ratio 0.2 --query-duration 8.0 --train-epochs 1 --batch-size 2
 /usr/local/miniconda3/bin/python src/data/external_adapters.py smoke-local mtg_jamendo data/raw/mtg_jamendo_audio --output-root data/external_smoke --eval-ratio 0.2 --query-duration 8.0 --train-epochs 1 --batch-size 2
 /usr/local/miniconda3/bin/python src/data/external_adapters.py smoke-local mtg_jamendo data/raw/mtg_jamendo_audio --output-root data/external_smoke --eval-ratio 0.2 --query-duration 8.0 --train-epochs 1 --batch-size 2
 ```
 ```
+
+## Git LFS policy
+
+Large raw archives and audio under `data/raw/` are tracked through Git LFS via [/.gitattributes](../../.gitattributes).
+
+## Download / ingestion policy
+
+1. Prefer **small, verifiable subsets first**.
+2. Keep original archives under `data/raw/` only when they are truly needed for reproducibility.
+3. For MTG-Jamendo, keep the corpus in an **evaluation/research** lane unless a verified commercial-safe subset is explicitly documented.
+4. Record source URL, subset choice, and license evidence in docs before broadening training scope.

@@ -222,6 +222,31 @@
 - 交接信息更适合自动化和长期持续开发
 - 交接信息更适合自动化和长期持续开发
 
 
 
 
+
+### Stage: 原始开放数据 LFS 治理
+
+完成项：
+- 新增根目录 [/.gitattributes](../.gitattributes)
+- 为 `acr-engine/data/raw/` 下的大文件与音频配置 Git LFS 跟踪策略
+- 重写 [acr-engine/data/raw/README.md](../acr-engine/data/raw/README.md)，补充：
+  - 数据落点职责表
+  - `check-local-ready -> smoke-local` 最短路径
+  - 原始数据下载 / LFS 治理策略
+- 补充 [docs/dataset-sources-and-licensing.md](./dataset-sources-and-licensing.md) 的下载 / LFS 治理说明
+
+验证结果：
+- `git lfs version` 成功
+- `git check-attr filter -- acr-engine/data/raw/fma_small_audio/example.wav` 返回 `filter: lfs`
+- `git check-attr filter -- acr-engine/data/raw/archive.zip` 返回 `filter: lfs`
+- 文档已明确区分：
+  - 工程可用性
+  - benchmark 适用性
+  - 商用可部署性
+
+结论：
+- 仓库现在具备承接真实开放音频和压缩包的 LFS 基础设施
+- 后续下载真实数据时，不会直接把大文件塞进普通 git 历史
+
 ### Stage: 真实数据就绪度守门
 ### Stage: 真实数据就绪度守门
 
 
 完成项：
 完成项：

@@ -96,3 +96,35 @@ flowchart LR
 
 
 ## Sources
 ## Sources
 - See [references-and-sources.md](./references-and-sources.md) for the current source map.
 - See [references-and-sources.md](./references-and-sources.md) for the current source map.
+
+
+## Download / LFS governance
+
+### Preferred repository behavior
+
+```mermaid
+flowchart TD
+    A[Upstream dataset source] --> B[Local raw drop zone]
+    B --> C[Git LFS tracked large files]
+    C --> D[check-local-ready]
+    D --> E[prepare-local / smoke-local]
+```
+
+### Current repo policy
+
+| Item | Policy | Reason |
+|---|---|---|
+| `acr-engine/data/raw/**/*.zip` | Git LFS | avoid bloating normal git history |
+| `acr-engine/data/raw/**/*.wav` / `.mp3` / `.flac` / `.ogg` | Git LFS | allow local reproducibility without normal blob explosion |
+| FMA Small | acceptable as first real-data engineering baseline | easiest realistic open music smoke path |
+| MTG-Jamendo | default to research/eval lane | do not assume commercial-safe rights without subset-specific proof |
+
+### Operational note
+
+Even when a dataset is technically downloadable, this project should separate:
+
+- **engineering usability**
+- **benchmark suitability**
+- **commercial deployment suitability**
+
+These are not the same thing.