Tags: tmwgsicp/ForgeRSS
Tags
v1.4.1 — Linux CI hotfix Fix import-time crash in generators/social/zsxq/scraper.py on hosts without the Windows-only LOCALAPPDATA env var. The unguarded Path(None) raised TypeError, breaking scripts/run_all.py on Ubuntu CI runners. EDGE_USER_DATA is now constructed only when LOCALAPPDATA is set, and the system-Edge fallback branch checks the sentinel before using it. Windows users with ZSXQ_USE_SYSTEM_EDGE=true are unaffected.
v1.4.0 New platforms (社交媒体 + 公司公告): - B 站 UP 主视频抓取(含 cookie 注入、可选下载) - 小红书用户笔记(接受纯 user_id 或带 xsec_token 的完整 URL) - 知识星球话题(含 PDF/音频/图片附件下载,<群名>_<群ID>/<话题>_<topicID>/ 归档) - 巨潮资讯网公司公告(三种订阅模式:关键词 / 分类 / 全市场最新,CI 可跑) Login state system (全平台统一): - 三层检测:URL 重定向标记 → not_logged_in HTML 标记 → logged_in 关键词 - config/login_keywords.json 配置化,无需改代码 - tools/test_login_check.py 端到端测试 + 自动杀 stale Chrome 进程 - tools/diagnose_login.py 协作诊断(DOM 关键词命中报告) - Profile 复用零配置,杀进程后重启即识别为登录态 Refactor: - 社交平台按目录组织(zhihu/bilibili/xiaohongshu/zsxq/base 等) - scripts/registry.py 18 个 generator 统一注册表 - BaseFeedGenerator: FEED_NAME 校验、SQLite WAL+busy_timeout、--max 真实生效 Fixes: - B 站 selector 更新 (.upload-video-card),视频标题正确解析 - ZSXQ selector 更新 (<app-topic>),附件下载链路完整 - 巨潮 ATTACHMENTS_DIR 从 cwd 相对路径改为基于项目根的绝对路径 - 清理 mojibake(登录态关键词保留正确中文) Open-source posture: - CI 默认关闭 cninfo PDF 下载(避免浪费带宽) - 删除 desktop-only feed 的死快照(CI 永远更新不到) - 微信视频号独立到 weixin_standalone/(不进 commit)
Major improvements to AI coding docs crawling - 316 articles total - Fix Qwen Code: 10 -> 95 articles - Fix Claude Code: 10 -> 95 articles - Fix Cursor: improved coverage and removed duplicates (50 articles) - Add streaming RSS generation for memory efficiency - Add document categorization and sorting - Clean up redundant scripts - Weekly update schedule for documentation feeds
v1.2.0 - AI Documentation Feeds & Streaming RSS Major Features: - AI coding tools documentation (OpenAI Codex, Claude Code, Cursor, Qwen Code) - Memory-efficient streaming RSS generation - Separate workflows: news (6hr) vs docs (weekly) - Force English language for documentation Improvements: - Increase doc limits to 300 pages - Better link discovery for JS-rendered pages - Document categorization and sorting - Language filtering (exclude non-English) Total: 231+ documentation articles
PreviousNext