chore(mediacrawler): checkpoint local login changes

.gitignore

@@ -179,4 +179,43 @@ agent_zone
 debug_tools
 
 database/*.db
-.omx/
+
+# === auto-added by ensure-repo-hygiene.sh ===
+# 敏感文件
+.env
+.env.*
+!.env.example
+*.key
+*.pem
+credentials.json
+credentials.*
+secrets.*
+
+# Python
+__pycache__/
+*.py[cod]
+*.egg-info/
+.venv/
+venv/
+.pytest_cache/
+.mypy_cache/
+.ruff_cache/
+
+# Node
+node_modules/
+*.log
+npm-debug.log*
+yarn-debug.log*
+.pnpm-debug.log*
+
+# IDE
+.vscode/
+.idea/
+*.swp
+.DS_Store
+
+# Build
+dist/
+build/
+*.egg
+.omx/

docs/business-product-logic.md

@@ -0,0 +1,34 @@
+# MediaCrawler 业务与产品逻辑
+
+## 这是什么
+
+MediaCrawler 是多平台自媒体公开数据采集框架。它的核心价值不是某一个平台的逆向细节，而是把“多平台采集、登录态、存储、代理、反爬”统一成同一套抽象。
+
+## 核心产品逻辑
+
+- 用户通过统一入口选择平台、登录方式、爬取类型，框架再分发到具体平台实现。
+- Playwright/CDP 登录态、代理池和签名逻辑是基础设施层，不该散落在业务脚本里。
+- 内容、评论、创作者信息和媒体文件都被视为同一条采集流水线的输出，而不是互相割裂的脚本。
+
+## 核心业务闭环
+
+1. 读取命令行参数和配置。
+2. 选择平台 crawler 与对应 API client。
+3. 建立浏览器上下文、登录态与代理能力。
+4. 执行搜索 / 详情 / 创作者抓取。
+5. 解析内容、评论、媒体资源。
+6. 写入文件或数据库存储层。
+
+## 当前业务边界
+
+- 这是通用采集框架，不直接承担 Workspace 内某个具体内容业务的最终决策。
+- 多平台统一抽象是第一性设计，不应把平台特例直接回写到入口层。
+- README 和架构文档都强调“学习/研究用途”，不应把它包装成当前 Workspace 的线上生产中台。
+
+## 推荐阅读顺序
+
+1. `README.md`
+2. `docs/项目架构文档.md`
+3. `docs/项目代码结构.md`
+4. `docs/CDP模式使用指南.md`
+

docs/dev-task-closeout-plan-2026-05-20.md

@@ -0,0 +1,58 @@
+# Dev Task Closeout 2026-05-20
+
+## Discovery Gate
+
+### 用户目标
+
+- 完成 stop hook 判定未闭环的 MediaCrawler 工程收尾：清理 dirty tree、提交、push、确认 PR 和验证 gate。
+
+### 任务分类
+
+- C：既有分支收尾与冲突自愈，不新增架构或第三方集成。
+- 调研结论：基于现有分支、Workspace finisher 和 PR validation 收尾；不从零实现。
+- 缺少 codebase map：仓内未发现 `docs/codebase-map.md`，本轮通过 git diff、PR metadata、`uv.lock` 和测试文件临时补足相关面。
+
+### AI 推断的相关面清单
+
+- source / caller: `media_platform/douyin/login.py`, `media_platform/douyin/client.py`, `media_platform/tieba/*`, `cmd_arg/arg.py`
+- tests / fixtures / contracts: `test/test_utils.py`, `tests/test_cmd_arg_tieba.py`, `tests/test_tieba_client_pagination.py`, `tests/test_tieba_extractor.py`
+- docs / 历史决策: `README.md`, `README_en.md`, `docs/business-product-logic.md`, CDP guide 和常见问题文档
+- config / flags / env: `pyproject.toml`, `uv.lock`, `.gitignore`; no env or secrets edited
+- runtime entrypoint / release surface: no local launchd or release entrypoint touched
+- external side effect / online write path: GitHub branch push and PR update only
+
+### 证据
+
+- `codex-dev-task-finisher.py` initially reported dirty tree and missing PR recognition.
+- `gh pr view` confirmed PR #880, fork head branch, and earlier `mergeStateStatus=DIRTY`.
+- `upstream/main` was merged into the task branch to resolve the dirty merge state.
+- `.gitignore` conflict kept both repository hygiene rules and upstream `.omx/`.
+- `uv.lock` conflict was resolved, then stale duplicate tuna package records inconsistent with current `pyproject.toml` default PyPI index were removed so `uv` could parse and normalize the lockfile.
+
+### 不确定项 / 需要用户拍板
+
+- None for local closure.
+- Merge remains subject to upstream repository permissions, review, and checks.
+
+### 验证矩阵
+
+- local: `git diff --check`
+- contract: `uv lock --check`; `uv lock`
+- smoke: `uv run pytest test/test_utils.py tests/test_cmd_arg_tieba.py tests/test_tieba_client_pagination.py tests/test_tieba_extractor.py`
+- online regression: not applicable; no live runtime, launchd, release entrypoint, or external owner write path touched
+- external endstate: branch pushed to `origin/chore/dirty-worktree-checkpoint-20260422`; PR #880 updated
+
+### Compatibility Oracle
+
+- 旧调用方 / caller: existing CLI and platform crawler callers remain on the same public module paths; no caller contract was intentionally changed in this closeout slice.
+- legacy fixture / oracle: existing tests plus upstream-added Tieba tests are the oracle for preserved behavior in this branch.
+- 行为是否变化: no behavior change was introduced by the closeout work itself; only merge conflict resolution, lockfile normalization, hygiene ignore rules, and documentation evidence were added.
+
+## TODO
+
+- [x] Commit dirty worktree changes with precise staged files.
+- [x] Push branch and read back remote branch head.
+- [x] Confirm PR #880 exists and update PR evidence.
+- [x] Resolve upstream merge conflicts and push updated head.
+- [x] Run local contract and smoke validation.
+- [ ] Re-run PR validation and final finisher after this plan file is committed.

media_platform/douyin/login.py

@@ -248,18 +248,19 @@ async def move_slider(self, back_selector: str, gap_selector: str, move_step: in
         element = await self.context_page.query_selector(gap_selector)
         bounding_box = await element.bounding_box() # type: ignore
 
-        await self.context_page.mouse.move(bounding_box["x"] + bounding_box["width"] / 2, # type: ignore
-                                           bounding_box["y"] + bounding_box["height"] / 2) # type: ignore
+        slider_center_x = bounding_box["x"] + bounding_box["width"] / 2 # type: ignore
+        slider_center_y = bounding_box["y"] + bounding_box["height"] / 2 # type: ignore
+        await self.context_page.mouse.move(slider_center_x, slider_center_y)
         # Get x coordinate center position
-        x = bounding_box["x"] + bounding_box["width"] / 2 # type: ignore
+        x = slider_center_x
         # Simulate sliding operation
         await element.hover() # type: ignore
         await self.context_page.mouse.down()
 
         for track in tracks:
             # Loop mouse movement according to trajectory
             # steps controls the ratio of single movement speed, default is 1, meaning the distance moves in 0.1 seconds no matter how far, larger value means slower
-            await self.context_page.mouse.move(x + track, 0, steps=move_step)
+            await self.context_page.mouse.move(x + track, slider_center_y, steps=move_step)
             x += track
         await self.context_page.mouse.up()
 

pyproject.toml

@@ -43,5 +43,5 @@ dependencies = [
 ]
 
 [[tool.uv.index]]
-url = "https://pypi.tuna.tsinghua.edu.cn/simple"
+url = "https://pypi.org/simple"
 default = true