自动发现学术论文 → 构建本地知识图谱 → 制作视频专栏的开源 skill。
Automatically discover papers (OpenAlex / arXiv), build a local knowledge graph (课题组 / 作者 / 论文 / 论点 / 论据), rank strong research groups by citation relations, render highlighted global & local graph views, and produce a daily Bilibili video column episode (EasySlides deck + TTS narration + gated release pipeline).
每日定时 ─ fetch_openalex / fetch_arxiv ─ build_graph ─ rank_groups ─ column_report
│
┌─────────────────────────────────────────┘
▼
render_graph (global 全局格局 + local 论文地位高亮)
▼
E:\bili\Bili 视频管线(EasySlides deck · IndexTTS2 配音 · Remotion 渲染 · 发布闸门)
▼
episodes.json 记录 ─ LESSONS.md 经验回写 ─ git commit & push(skill 自我进化)
git clone https://github.com/mappedinfo/paper-radar
cd paper-radar
python scripts/fetch_openalex.py --channel reinforcement-learning
python scripts/fetch_arxiv.py --channel reinforcement-learning
python scripts/build_graph.py --channel reinforcement-learning
python scripts/rank_groups.py --channel reinforcement-learning
python scripts/column_report.py --channel reinforcement-learning
python scripts/render_graph.py --channel reinforcement-learning --mode globalPython ≥ 3.10, standard library only (no pip installs needed for the radar
itself). Reports land in reports/, graphs in data/<channel>/graph.json,
renders in data/<channel>/renders/.
https://mappedinfo.github.io/paper-radar — 交互式力导向知识图谱
(拖拽/缩放/搜索/点击查看 2 跳局部关系),节点含 课题组 / 作者 / 论文 /
论点 / 论据。本站只可视化关系、论点与论据,不存储任何论文内容。
数据由 scripts/publish_site.py 在每日例程中刷新并随仓库推送。
Copy channels/reinforcement-learning.json, adjust topic queries, venue
whitelist, and seed groups, then run the same commands with --channel <id>.
- Skill routing:
SKILL.md - Graph schema:
references/graph-schema.md - 每期 PPT 通用大纲与布局:
references/episode-deck-blueprint.md - Daily runbook + skill-evolution protocol:
workflows/daily-column.md
MIT © mappedinfo