update: README.md

This commit is contained in:
tao.chen
2026-08-14 19:30:25 +08:00
parent 139f2c02c1
commit 76af0f763c
2 changed files with 328 additions and 100 deletions
+242
View File
@@ -0,0 +1,242 @@
# Handover — 近期变更与待办
面向接手运维 / 二次开发的人。记录最近几次 commit 的动机、改动范围、未尽事项,
以及「线上有旧数据时怎么过渡」。详细架构见 `ARCHITECTURE.md`;本仓库的开发规约
`DEVELOP.md`
---
## 1. 最近 5 个 commit(按时间倒序)
### `139f2c0` — fix: delete errortrash 路径统一 + 删除后 DB 字段更新)
**背景**:删除文件统一改为移动到 trash 桶后,存在三个连锁 bug:
1. **local 模式 trash 路径双重 `data`**`trash_key = f"{item.bucket_name}/{item.object_key}"`
local 模式下 `bucket_name` 是绝对路径(如 `/data/workspace`),最终落到
`data/trash/data/workspace/{ws_id}/{user_id}/...`
2. **s3 模式桶名 / 路径名错位**`.env` 之前是 `S3_WORKSPACE_BUCKET=workspaces`(复数),
代码 default / `PURPOSE_BUCKETS` / `USAGE_TYPE_TO_PURPOSE` 全部是单数 `workspace`
跨环境部署产生 `KeyError: 'workspaces'`(软删时 `object_stores[item.bucket_name]` 抛错)。
3. **软删后 `storage_objects` 路径字段未更新**:只翻 `trash_key` / `object_status` /
`deleted_at` / `is_deleted``bucket_name` / `object_key` / `storage_uri` 仍指向源,
与物理位置不一致,`purge_trash_object` 必须读旧 `trash_key` 才能定位 trash 文件。
**改动**5 个文件,44+/24-):
| 文件 | 改动 |
|---|---|
| `backend/src/backend/services/storage.py` | `trash_key = f"{source_purpose}/{item.object_key}"`;软删成功后更新 `bucket_name` / `object_key` / `storage_uri` 指向 trash 位置 |
| `backend/src/backend/storage_api.py` | `restore_object``item.object_key` 直接定位 trash,按 `/``source_purpose`,写回源桶,更新 DB 字段指回 source;`purge_trash_object``(actual_bucket_name("trash"), item.object_key)` 删除,加 `bucket_name != trash` 的 409 防御 |
| `common/src/common/storage/factory.py` | 引入 `USAGE_TYPE_TO_PURPOSE` 字典(替代 `storage_api.py` 的本地定义) |
| `common/src/common/storage/__init__.py` | 导出 `USAGE_TYPE_TO_PURPOSE` |
| `.env.example` | `S3_WORKSPACE_BUCKET=workspaces``workspace``S3_VERSION_BUCKET=versions``version``S3_RUN_LOG_BUCKET=run-logs``run-log`(统一单数) |
**注意点**
- `workspaces_root()` 函数名 / `WORKSPACES_ROOT` 环境变量名不动(这是 path 概念,不是 bucket 名)。
- `Workspaces` 数据库表名 / `auth.py``"workspaces"` JSON key(API 响应)不动(这些不是桶名)。
- `visibility="workspace"` 是 ACL Literal,不动。
### `2e070fa` — fix: pre check(前端 script 创建/上传前置校验)
`CreateScriptModal.tsx`+25 行)— 文件名 / workspace_id / 类型前置校验。
`ScriptsPage.tsx`+1/ `scriptWorkspaceStore.ts`+19/-10)— store action 增加
`precheck` 步骤,提前拦截无效请求。
### `5c5dec9` — fix: allows_same_name_different_parent
允许同名 script 出现在不同父目录下。`scripts.py`+32/-?)、`services/storage.py`
`storage_api.py` 微调;`test_scripts.py` 新增 57 行回归测试。
### `5d49ff5` — fix: file upload error(大重写)
修复 file upload 路径上一连串 schema / API / 上传流程的不一致。改动面最大
14 个文件,872+/604-):
- 新增索引 / 调整唯一约束(去掉了 scripts 和 data_resources 的旧 unique index
- 把多个小 migration 合并进 `e1f2a3b4c5d6_rebuild_baseline.py` 一份大 baseline
- `resources.py` / `scripts.py` 上传流程对齐
- `test_resources.py` / `test_scripts.py` 大量回归
- `API.md` 增加 3 行
### `c65d6dc` — fix: build error
`backend/Dockerfile` 单行修复(镜像构建报错)。
---
## 2. Trash 修复详解(commit `139f2c0`
### 2.1 修复后的语义
- **trash_key**`f"{source_purpose}/{item.object_key}"``source_purpose` 来自
`USAGE_TYPE_TO_PURPOSE[item.usage_type]`,与 `item.bucket_name` 解耦 —
避免 local 模式绝对路径泄漏 + 命名错位风险。
- **软删后 `storage_objects` 行的 7 个字段同步**
| 字段 | 软删前 | 软删后 |
|---|---|---|
| `bucket_name` | 源桶 | trash 桶 |
| `object_key` | 源 key | trash key |
| `trash_key` | NULL | trash key |
| `storage_uri` | 源 URI | trash URI`build_storage_uri` 重建) |
| `object_status` | `available` | `deleted` |
| `is_deleted` | `0` | `1` |
| `deleted_at` | NULL | `<utcnow>` |
- **物理文件落点**
| 模式 | 源文件(被删) | trash 文件(新落) |
|---|---|---|
| S3 | `s3://<source_bucket>/<key>` | `s3://trash/<source_purpose>/<key>` |
| local | `<LOCAL_STORAGE_BASE_DIR>/<source_purpose>/<key>` | `<LOCAL_STORAGE_BASE_DIR>/trash/<source_purpose>/<key>` |
`source_purpose``{workspace, version, run_log}`
### 2.2 `restore_object` 反向流程
```python
data = await object_stores[trash_bucket].get(item.object_key) # 从 trash 读
source_purpose, _, source_key = item.object_key.partition("/") # 拆 source_purpose
target_bucket = actual_bucket_name(source_purpose) # 还原目标桶
await object_stores[target_bucket].put(source_key, data) # 写回源
item.bucket_name = target_bucket
item.object_key = source_key
item.storage_uri = build_storage_uri(target_bucket, source_key)
# trash_key 保留不动 → 未来 reaper 可据此清理 trash 里的 orphan 副本
item.object_status = "available"
item.deleted_at = None
```
**trash 文件不会被 restore 主动删除**(保留现有安全语义:restore 失败时
数据仍在 trash 里有兜底)。未实现的 reaper 未来可扫
`object_status='available' AND trash_key IS NOT NULL` 清理 orphan。
### 2.3 `purge_trash_object` 防御
加 409 防御:`if item.bucket_name != actual_bucket_name("trash"): raise 409`
避免误删源数据(用 `item.object_key` 直接删 trash 文件,不再读旧 `trash_key`)。
### 2.4 USAGE_TYPE_TO_PURPOSE 提升到 common 层
`backend/storage_api.py``backend/services/storage.py` 都依赖这个映射。
提到 `common/storage/factory.py` 后两边都 `from common.storage import USAGE_TYPE_TO_PURPOSE`
避免互相 import 循环。
---
## 3. 线上数据兼容(重要)
修复后**新行为对 DB 里已有的旧行不感知**。如果部署环境之前是
`S3_WORKSPACE_BUCKET=workspaces`(复数)且已上传过文件,
`storage_objects.bucket_name` 列里会有复数残留。新代码 default 是单数,
`object_stores` dict key 也是单数,访问旧行仍会 `KeyError: 'workspaces'`
**部署新代码前必须先跑一次数据修复 SQL**(先 SELECT 看数量,确认是同一个 deployment):
```sql
-- 1. 先看现状(按 backend 区分)
SELECT storage_backend, bucket_name, COUNT(*)
FROM storage_objects
WHERE object_status = 'available'
GROUP BY storage_backend, bucket_name
ORDER BY storage_backend, bucket_name;
-- 2a. S3 部署:把复数残留修正成单数
UPDATE storage_objects
SET bucket_name = 'workspace'
WHERE bucket_name = 'workspaces'
AND object_status = 'available';
-- 2b. local 部署:把绝对路径里的复数残留修正成单数
UPDATE storage_objects
SET bucket_name = '/data/workspace'
WHERE bucket_name = '/data/workspaces'
AND object_status = 'available';
```
> **只改 `available` 行**`deleted` 行已经在 trash 里、bucket_name 会被新代码改写,无需处理。
> 如果同时改 `version` / `run_log` 桶残留,按相同模式跑。
---
## 4. 本地验收清单
```bash
# 1. 语法
uv run python -m compileall common/src backend/src runtime/src schedule/src
# 2. 测试
uv run --package backend pytest backend/tests -q
# 期望:30 passed
# 3. 命名一致性
grep -rn '"workspaces"\|"versions"\|"run-logs"' \
--include='*.py' --include='*.yml' --include='*.env*' \
backend/src common/src docker-compose.yml .env .env.example 2>/dev/null
# 期望:除 `__tablename__ = "workspaces"`DB 表名)和 `auth.py` 的 JSON key 外,零结果
# 4. 单复数对齐
grep -n 'S3_WORKSPACE_BUCKET\|S3_VERSION_BUCKET\|S3_RUN_LOG_BUCKET\|S3_TRASH_BUCKET' \
.env.example
# 期望:全部 = 单数(workspace / version / run-log / trash
```
---
## 5. 回归测试覆盖
| 场景 | 触发方式 | 期望 |
|---|---|---|
| 同名 script 不同父目录 | `test_soft_delete_object_sets_is_deleted` 旁边的新测试 | 两条 row 同名共存 |
| 软删 → 物理文件落到 trash 桶的 `source_purpose/{key}` | 手工:上传 → 删 → `mc ls s3/trash/` | 文件在;源桶里已删 |
| 软删后 `storage_objects` 7 字段同步 | 软删后 `SELECT * FROM storage_objects WHERE storage_object_id=...` | 见 §2.1 表格 |
| `restore_object` 把 trash 文件写回源桶 | 软删 → restore → 查源桶 | 源桶恢复;trash 文件保留(orphan |
| `purge_trash_object` 拒绝误删源数据 | 手工:把 `bucket_name` 改成源桶后调 purge | 409 |
---
## 6. 监控项(建议加)
1. `storage_objects``bucket_name``actual_bucket_name(purpose)` 不一致的比例
(按 `usage_type` 分组 COUNT)。生产环境应长期趋近于 0。
2. `object_status='deleted'``bucket_name != actual_bucket_name('trash')` 的行数 —
应为 0(软删失败回滚残留)。
3. trash 桶目录大小 vs `object_status='deleted'` 行数(孤儿文件比例)。
没实现 reaper 前这个会单调上升。
---
## 7. 已知未实现项
- **Trash reaper**`s3_trash_retention_days` 已声明,但代码没有扫表清理。
`restore_object` 故意保留 trash 文件作 orphan,等 reaper 兜底。
第一次实现建议在 `schedule/src/schedule/` 加一个 tick job
```sql
SELECT storage_object_id FROM storage_objects
WHERE deleted_at < UTC_TIMESTAMP() - INTERVAL :retention_days DAY
AND object_status = 'deleted'
LIMIT 100;
```
然后调 `/internal/v1/admin/trash/purge`。
- **跨 backend trash 兼容**:当前只在 `item.storage_backend == settings.storage_backend`
的行上做物理移动。异构 backend 的旧行(少)软删时只翻 DB 状态,trash 文件实际没移动。
远程 8.153.151.51:8890 应该都是同一 backend,目前无影响。
- **历史 trash_key 兼容读**:新代码用 `item.object_key` 读 trash**不**回退到
`item.trash_key`。如果迁移前有按旧 `trash_key = f"{bucket_name}/{object_key}"`
写入的已删行,`object_key` 还是源 keyrestore 会找不到 trash 文件(404)。
处理:在 `restore_object` 加 fallback — 先按 `item.object_key` 试,失败再按 `item.trash_key`
试一次(兼容窗口期)。建议在 reaper 上线后删除 fallback。
---
## 8. 仍未合并到 develop 的本地改动
```text
M frontend/vite.config.ts (4+/2-)
?? data/ (untracked — 本地 docker volume 残留,可忽略)
```
`frontend/vite.config.ts` 的改动不在本次 trash 修复范围内,等下次 commit 一起提。
+86 -100
View File
@@ -1,29 +1,26 @@
# Model Platform
A self-hosted **Jupyter-based model development platform** that combines
an interactive workspace, a DAG scheduler, an object-storage-backed artifact
store, and per-workspace runtime isolation — all behind a single Nginx
gateway.
自托管的 **Jupyter 模型开发平台**:交互式 workspace、DAG 调度、对象存储工件、
按 workspace 隔离的运行时 — 全部经一个 Nginx 网关对外。
> Stack: React Router SPA · FastAPI · APScheduler · MySQL · S3-compatible
> storage (or local filesystem via `STORAGE_BACKEND=local`) · shared
> Jupyter · FUSE mount via rclone (s3 mode only)
> Single ingress (Nginx :80); all other services are Docker-internal.
> 技术栈:React Router SPA · FastAPI · APScheduler · MySQL · S3 兼容存储
> (或本地文件系统,`STORAGE_BACKEND=local`)· 共享 Jupyter · rclone FUSE 挂载(仅 s3 模式)
> 单一入口(Nginx :80);其他服务只在 Docker 内网互通。
## What it does
## 它做什么
| Capability | Where |
| 能力 | 位置 |
|---|---|
| Workspace-scoped notebook editing with row-level lock | `backend/jupyter.py` + `scripts.is_locked` |
| Authenticated Jupyter routing (browser never sees the runtime token) | `nginx/default.conf` + `auth_request` + `backend/jupyter.py` |
| Object storage for notebooks / scripts / versions / run logs (s3 / local toggle) | `common/storage/` + `backend/scripts.py` |
| DAG-style scheduling: nodes, edges, cron, manual trigger, retries, snapshots | `backend/schedules.py` + `backend/schedule_runs.py` + `schedule/` (5 modules) |
| DAG execution via MySQL Outbox (no Redis, no in-process queues) | `schedule/orchestrator.py` + `schedule/worker.py` |
| Per-workspace Jupyter sub-process pool with asyncio locks | `runtime/process.py` |
| rclone FUSE mount of the workspace bucket into the runtime (s3 mode) | `runtime/mount.py` |
| MySQL-only persistence (26 tables, soft-delete, no foreign keys) | `common/db/models/` |
| workspace notebook 编辑,行级锁 | `backend/jupyter.py` + `scripts.is_locked` |
| Jupyter 鉴权路由(浏览器永远拿不到 runtime token | `nginx/default.conf` + `auth_request` + `backend/jupyter.py` |
| notebook / script / version / run_log 的对象存储(s3 / local 二选一) | `common/storage/` + `backend/scripts.py` |
| DAG 调度:节点、边、cron、手动触发、重试、快照 | `backend/schedules.py` + `backend/schedule_runs.py` + `schedule/`5 个模块) |
| DAG 执行走 MySQL Outbox(无 Redis,无进程内队列) | `schedule/orchestrator.py` + `schedule/worker.py` |
| 每个 workspace 一个 Jupyter 子进程池,配 asyncio | `runtime/process.py` |
| runtime 内 rclone FUSE 把 workspace 桶挂上来(s3 模式) | `runtime/mount.py` |
| MySQL 持久化(26 张表,软删除,无外键) | `common/db/models/` |
## Architecture at a glance
## 架构一览
```
┌────────────────────┐
@@ -66,69 +63,67 @@ gateway.
└───────────────────────────────┘
```
Object storage is selectable via `STORAGE_BACKEND` (s3 | local). In s3 mode
the 4 purpose-named buckets (`workspaces` / `versions` / `run-logs` / `trash`)
are S3 buckets; in local mode they're subdirectories of `LOCAL_STORAGE_BASE_DIR`,
shared via the `local-storage` Docker volume. See `DEVELOP.md` §Storage.
对象存储通过 `STORAGE_BACKEND`s3 | local)二选一。s3 模式下 4 个 purpose 命名桶
`workspace` / `version` / `run-log` / `trash`)是独立的 S3 bucketlocal 模式下
`LOCAL_STORAGE_BASE_DIR` 的子目录,通过 Docker volume `local-storage` 共享。
详见 `DEVELOP.md` §存储。
Detailed design lives in `ARCHITECTURE.md`. Implementation deviations and
recent refactors are recorded in `HANDOVER.md`.
详细设计见 `ARCHITECTURE.md`。实现的偏离和近期重构记录在 `HANDOVER.md`
## Repository layout
## 目录结构
```text
frontend/ React Router SPA
backend/ FastAPI: public API + internal storage API
runtime/ Jupyter subprocess manager + rclone FUSE
schedule/ DAG scheduler (5 modules: context/scheduler/
orchestrator/worker/service)
common/ Settings, SQLAlchemy models, storage SDK,
outbox events, jobstore
migrations/ Alembic baseline + per-feature revisions
nginx/ (concept only — see "Container" below)
scripts/ nginx-entrypoint.sh (template renderer)
docker-compose.yml 4 services — web / backend / runtime / schedule
default.conf Nginx template (mounted, rendered at start)
.env.example All env vars consumed by common.config.Settings
backend/ FastAPI:公开 API + 内部存储 API
runtime/ Jupyter 子进程管理 + rclone FUSE
schedule/ DAG 调度器(5 模块:context/scheduler/
orchestrator/worker/service
common/ 配置、SQLAlchemy 模型、存储 SDK
outbox 事件、jobstore
migrations/ Alembic 基线 + 各特性 migration
nginx/ (仅概念 — 见下方「容器」一节)
scripts/ nginx-entrypoint.sh(模板渲染)
docker-compose.yml 4 服务 — web / backend / runtime / schedule
default.conf Nginx 模板(挂载,启动时渲染)
.env.example common.config.Settings 消费的所有环境变量
```
## Containers
## 容器
| Service | Image | Exposed | Purpose |
| 服务 | 镜像 | 暴露 | 用途 |
|---|---|---|---|
| `web` | `nginx:alpine` | host `:8888``:80` | SPA, `/api/` reverse-proxy, `/jupyter/{ws}/` auth_request proxy, `/storage/` S3 passthrough (s3 mode only) |
| `backend` | `Dockerfile` | internal only | DAG CRUD, script CRUD, schedule triggers, `/api/v1/auth/jupyter`, `/internal/v1/*` storage control plane |
| `runtime` | `Dockerfile` | internal only | Per-workspace Jupyter sub-process pool, rclone FUSE mount of `workspaces` bucket (s3 mode) |
| `schedule` | `Dockerfile` | internal only | Cron tick + DAG execution via MySQL Outbox polling |
| `web` | `nginx:alpine` | 宿主机 `:8888``:80` | SPA`/api/` 反向代理、`/jupyter/{ws}/` auth_request 代理、`/storage/` S3 直通(仅 s3 模式) |
| `backend` | `Dockerfile` | 仅内网 | DAG CRUDscript CRUDschedule 触发、`/api/v1/auth/jupyter``/internal/v1/*` 存储控制面 |
| `runtime` | `Dockerfile` | 仅内网 | 每个 workspace 一个 Jupyter 子进程池、rclone FUSE 挂载 `workspace` 桶(s3 模式) |
| `schedule` | `Dockerfile` | 仅内网 | cron tick + DAG 执行(轮询 MySQL Outbox |
The architecture **deliberately has only one host port** (the gateway);
all other services are on the Docker internal network. This is enforced in
`docker-compose.yml` — no `ports:` on backend / runtime / schedule.
架构**故意只暴露一个宿主机端口**(网关);其他服务都在 Docker 内网。
这一点在 `docker-compose.yml` 里强制执行 — backend / runtime / schedule 都没有 `ports:`
## Quick start
## 快速启动
```bash
cp .env.example .env
# Edit .env — at minimum change MYSQL password and (in s3 mode) S3 credentials.
# 编辑 .env — 至少改 MYSQL 密码,以及(s3 模式下)S3 凭据。
# Static check
# 静态检查
uv sync --all-packages
uv run --frozen --package backend python -m compileall -q backend/src common/src
uv run --frozen --package schedule python -m compileall -q schedule/src
uv run --frozen --package runtime python -m compileall -q runtime/src
# Apply schema
# 应用 schema
uv run --frozen --package backend alembic upgrade head
# Bring up the stack
docker compose config # validate
# 启动整套服务
docker compose config # 校验
docker compose up -d --build
docker compose ps
```
Visit `http://localhost:8888`.
访问 `http://localhost:8888`
### Logs
### 日志
```bash
docker compose logs -f backend
@@ -136,47 +131,43 @@ docker compose logs -f schedule
docker compose logs -f runtime
```
### Tear down (keeps MySQL + S3 / local-storage volumes)
### 停服(保留 MySQL + S3 / local-storage 数据卷)
```bash
docker compose down
```
### Wipe data
### 抹数据
```bash
docker compose down -v
```
## Configuration
## 配置
All environment variables are declared once in `common/src/common/config.py`
as a pydantic-settings `Settings` class, with a `@lru_cache` singleton.
Adding a new env var:
所有环境变量在 `common/src/common/config.py` 里用 pydantic-settings 的 `Settings`
类一次性声明,外面套一层 `@lru_cache` 单例。新增环境变量:
1. Add the field to `Settings` in `common/src/common/config.py` (with a
sensible default so dev-env "just works").
2. Add the line to `.env.example` with a comment.
3. Use `settings.<name>` at the call site. Never `os.environ["..."]`.
1. `common/src/common/config.py` `Settings` 里加字段(带合理 default,使 dev 启动不需要设)
2.`.env.example` 加一行带注释
3. 调用点用 `settings.<name>`,永远不要用 `os.environ["..."]`
See `DEVELOP.md` for the full list of variables and their meanings.
完整环境变量列表和含义见 `DEVELOP.md`
## Storage layout
## 存储布局
Four purpose-named buckets. The mapping from `StorageObjects.usage_type`
to bucket is decided in **one place** (`backend/storage_api.py:resolve_bucket`):
4 个 purpose 命名桶。从 `StorageObjects.usage_type` 到桶的映射由
**单一入口**`backend/storage_api.py:resolve_bucket`)决定:
| `usage_type` | Bucket (env var) | Default name |
| `usage_type` | 桶(环境变量) | 默认名 |
|---|---|---|
| `working_copy`, `public_script`, `data_resource`, `snapshot` | `S3_WORKSPACE_BUCKET` | `workspaces` |
| `version_artifact` | `S3_VERSION_BUCKET` | `versions` |
| `run_log`, `run_result` | `S3_RUN_LOG_BUCKET` | `run-logs` |
| (soft-delete target) | `S3_TRASH_BUCKET` | `trash` |
| `working_copy``public_script``data_resource``snapshot` | `S3_WORKSPACE_BUCKET` | `workspace` |
| `version_artifact` | `S3_VERSION_BUCKET` | `version` |
| `run_log``run_result` | `S3_RUN_LOG_BUCKET` | `run-log` |
| (软删除目标) | `S3_TRASH_BUCKET` | `trash` |
In `STORAGE_BACKEND=s3` mode these are 4 separate S3 buckets. In
`STORAGE_BACKEND=local` mode they are 4 subdirectories under
`LOCAL_STORAGE_BASE_DIR` (default `/data`), so the layout above
becomes:
`STORAGE_BACKEND=s3` 模式下是 4 个独立 S3 桶。`STORAGE_BACKEND=local` 模式下
`LOCAL_STORAGE_BASE_DIR`(默认 `/data`)下的 4 个子目录:
```
/data/
@@ -186,36 +177,31 @@ becomes:
└── trash/ # S3_TRASH_BUCKET
```
A workspace's `artifact_bucket` column (when non-null) overrides the
default for that workspace, regardless of `usage_type`useful for
isolating a paying customer to their own bucket.
某个 workspace `artifact_bucket` 列(非 NULL 时)覆盖该 workspace 的默认桶,
无视 `usage_type`适合把付费客户隔离到专属桶。
The object key is a flat two-level path`workspace_id` and a server-
issued `ulid` for the object:
对象 key 是两层扁平路径`workspace_id` 加服务端签发的 `ulid`
```
<workspace_bucket>/<workspace_id>/<ulid>{.<ext>}
```
The file name, extension, content type, and logical path live in the
`StorageObjects` and `Scripts` rows, not in the object key, so the
storage can be re-organised without a database rewrite.
文件名、扩展名、MIME、逻辑路径都放在 `StorageObjects``Scripts` 行里,不进
object key — 重新组织存储不需要重写数据库。
Backend code never writes to the container's local filesystem (except
in `STORAGE_BACKEND=local` mode, where the shared `local-storage` volume
is the canonical store). Schedule Executor stages node artifacts in
`tempfile.TemporaryDirectory()` (auto-cleaned). Only the `runtime`
container keeps a host volume — required by the rclone FUSE mount in
s3 mode, and a no-op pass-through in local mode.
Backend 代码从不写容器本地文件系统(`STORAGE_BACKEND=local` 模式除外,那里共享
`local-storage` volume 就是规范存储)。Schedule Executor 在 `tempfile.TemporaryDirectory()`
里暂存节点工件(自动清理)。只有 `runtime` 容器保留宿主 volume — s3 模式下 rclone FUSE
挂载需要;local 模式下是 no-op 透传。
## Documentation
## 文档
- `README.md` (this file) — quick orientation
- `ARCHITECTURE.md`design diagrams + simplification history
- `HANDOVER.md`implementation deviations, recent refactors, pending work
- `DEVELOP.md`developer guide (env vars, code conventions, common tasks)
- `CLAUDE.md` — agent-facing conventions for the repo
- `README.md`(本文)— 快速导读
- `ARCHITECTURE.md`设计图 + 简化历史
- `HANDOVER.md`实现偏离、近期重构、待办事项
- `DEVELOP.md`开发指南(环境变量、代码规约、常用操作)
- `CLAUDE.md` — agent 面向的本仓库规约
## License
## 许可
Internal.
内部。