What you missed is what you didn't need.

AtomBrief is a personal information-metabolism system: RSS capture → keyword hard-filter → AI distillation into 1–2 sentence knowledge atoms → an auto-built weekly report site. Yesterday's digest gets overwritten, not archived. Zero hoarding, zero reading guilt.

Academic literature has its own track — see AtomFeed. AtomBrief handles everything else: tech, tools, ideas, growth.

---

Core promises

• No accumulation — the daily digest is overwritten every day; unread ≠ debt
• Atomization — no long-read guilt; AI distills each item to 1–2 sentences
• Assetization — highlights you keep become a weekly report, pushed to GitHub Pages as a public proof of information taste

System architecture

ReadYou OPML export
        ↓
[Phase 1]  Ingestion & hard filter        keyword whitelist / blacklist
        ↓
[Phase 2]  Daily digest distillation      DeepSeek → knowledge atoms, deduped
        ↓
[Phase 3]  Human-in-the-loop curation     Web UI: swipe away / highlight → atoms.json
        ↓
[Phase 4]  Weekly assetization            AI recomposition → GitHub Push → Pages

Quick start

git clone https://github.com/AtomRearch/AtomBrief.git
cd AtomBrief
python -m venv venv && venv\Scripts\activate    # or source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env                            # fill in DEEPSEEK_API_KEY etc.

Export your subscriptions from ReadYou (or any reader) as OPML to ./data/readyou.opml — a starter list ships at RSSsource/follow.example.opml.

python main.py --mode daily     # fetch + distill today's digest
python main.py --mode serve     # review at http://localhost:8000
python main.py --mode weekly    # compose & push the weekly report

Automation (GitHub Actions)

Two workflows ship in .github/workflows/:

Workflow	Schedule (CST)	Job
daily_digest.yml	18:00 daily	fetch → filter → distill → commit digest
weekly_report.yml	Sunday 20:00	compose weekly report → push → Pages

Setup: add the DEEPSEEK_API_KEY repository secret, enable Read and write workflow permissions, and commit your data/readyou.opml.

Configuration

All runtime configuration lives in .env (see .env.example):

Variable	Purpose
DEEPSEEK_API_KEY	atom distillation & weekly composition
RSS_OPML_PATH	your subscription list
KEYWORDS_WHITELIST / KEYWORDS_BLACKLIST	hard filter (whitelist wins)
GITHUB_TOKEN / GITHUB_REPO	weekly-report push target
QQ_SMTP_* / NOTIFICATION_EMAIL	optional email notification
DAILY_DIGEST_TIME / WEEKLY_REPORT_TIME	local scheduling

Tech stack

Python 3.10+ · feedparser + aiohttp · Flask + vanilla JS · DeepSeek API (V3/R1) · JSON/JSONL storage · GitHub Actions + Pages

More docs

Quick start · Deployment · Weekly site guide · Project summary

Ecosystem

Part of the AtomRearch matrix. Siblings: AtomFeed (academic literature) · AtomEIS (impedance analysis) · AtomSilo (knowledge storage).

License

MIT — see LICENSE.

Philosophy: input as soft rest, output as hard assets.