name: arxiv-digest
每日从 arXiv 抓取具身智能、表征学习、强化学习方向的最新论文,经过 LLM 筛选和解读后,推送到 Discord 并归档到 Hugo 博客。
使用 scripts/run_daily.py 作为唯一入口:
# 仅生成简报(dry run)
python3 scripts/run_daily.py
# 生成并推送到 Discord
python3 scripts/run_daily.py --publish-discord
# 生成并同步到 Hugo
python3 scripts/run_daily.py --publish-hugo
# 生成并推送两者
python3 scripts/run_daily.py --publish-discord --publish-hugo
# 跳过 LLM 增强(快速测试)
python3 scripts/run_daily.py --skip-enrich
robotdaily.html - 移动端友好的 HTML 卡片robotdaily.md - Markdown 归档版本site/content/ai-daily/YYYY-MM-DD.md每次运行在 output/YYYY-MM-DD/ 目录生成:
candidates.json - 候选论文列表selected.json - 精选论文列表enriched.json - LLM 增强后数据robotdaily.html - HTML 移动端摘要卡片robotdaily.md - Markdown 归档版本manifest.json - 元数据清单环境变量配置在 arxiv-digest/.env:
# LLM 模型
INSIGHT_MODELS=qwen3.5:27b
# 输出目录
ROBOTDAILY_OUTPUT_DIR=/home/zhn/.openclaw/workspace/skills/robdaily/arxiv-digest/output
# Hugo 内容目录
HUGO_CONTENT_DIR=/home/zhn/.openclaw/workspace/skills/robdaily/site/content/ai-daily
# Discord 推送模式:thread | channel | fixed-channel | existing-channel
DISCORD_DELIVERY_MODE=existing-channel
# Discord 配置
DISCORD_ACCOUNT_ID=codex
DISCORD_GUILD_ID=1474063117875937340
DISCORD_TARGET_CHANNEL_ID=1481632217930141697
DISCORD_BOT_TOKEN=your-bot-token-here
# 线程自动归档时间(分钟)
DISCORD_THREAD_AUTO_ARCHIVE_MIN=10080
每天上午 10:30(Asia/Shanghai)自动运行:
30 10 * * * cd /home/zhn/.openclaw/workspace/skills/robdaily/arxiv-digest && python3 scripts/run_daily.py --publish-discord --publish-hugo >> logs/robotdaily.log 2>&1
cat arxiv-digest/logs/robotdaily.log
ls -la arxiv-digest/output/$(date +%Y-%m-%d)/
cd site && hugo --quiet