Practical Scenarios: 5 workflows that make sip useful
This page isn't a feature walkthrough — it gives you real workflows you can copy and run. Every scenario has: the goal, the steps, and "what counts as success."
Scenario 1: Let AI answer only from sources you trust
Goal: When you or your agent researches, it stops citing Sohu, Baijiahao, and other garbage sources.
Steps:
# 1. Add sources you trust (or batch-import via OPML)
sip -d https://blog.hotsouprealm.top/atom.xml
sip --import-opml your-export.xml
# 2. Configure AI (one-time) — pick Embedding/LLM provider, enter API key
sip --init
# 3. Vectorize (turn articles into semantic vectors)
sip --index
# 4. Search! AI only returns articles from your subscribed sources
sip --search "your topic" --jsonSuccess looks like: every --search result traces back to a source in your subscription list (verify with sip -l).
Level up:
- Want someone else's (parents/friends) AI to benefit too? Hand them the sip-rss skill and this system prompt:
Before researching, run sip --search "keyword" --json and only cite the trusted sources it returns;
when summarizing, note each source and link; if unsure, say "not found" instead of making things up.Scenario 2: Track what changed in an important article
Goal: When an important article (policy, announcement, tutorial) is quietly edited, you find out immediately.
Steps:
# 1. Update feeds daily
sip --sync # only updates "due" feeds; or sip --update-all for everything
# 2. See who edited what today
sip --today # the top "today's changes" section lists edited articles and what changed
# 3. Dig into an article's revision history
sip --versions 42 # which versions exist, when they were made
sip --diff 42 # what changed in the last two versions
sip --diff 42 v1 v3 --json # compare two specific versions, structured outputSuccess looks like: --diff outputs added/removed lines with before/after context, not the words "has updates."
After an edit, the title in the TUI gets a
✎mark; pressVon it to view versions directly. This is "version is truth."
Scenario 3: Read what matters in 5 minutes a day
Goal: Don't get drowned by an endless feed — one fixed bowl a day, read and close.
Steps:
# 1. Set update schedules per feed (only fetch when due, no wasted resources)
sip --schedule 1 daily@08:00 # feed 1 updates at 8 AM daily
sip --schedule 2 30m # feed 2 every 30 minutes (e.g. a news feed)
# 2. Opening the TUI silently syncs due feeds
./sip.exe
# 3. Or check today's list from the command line
sip --today # 5 picks + estimated reading time + reasons
sip --today --refresh # if you want to re-roll the bowlSuccess looks like: --today gives you a list of up to 5 articles, each with "why it was picked." Read them all or skip them — your daily intake loop is closed.
Scenario 4: Set up a "peace-of-mind reader" for family members
Goal: Parents/elders open it and read; garbage sources are already blocked outside.
Steps:
# 1. Only subscribe to sources you vetted (CCTV, local weather bureau, medical accounts you trust)
sip --onboarding # built-in recommended templates (add by category)
sip -d https://source-you-vetted/rss
# 2. Articles with too-short summaries prompt a full-text fetch for slow reading
sip --fulltext 12 # fetch original text to local cache
# 3. All they need to remember: open sip → select up/down → press Enter to read
./sip.exe # TUI, full keyboard, no ads, no recommendation feedSuccess looks like: a family member can open the TUI, read one article, and quit without your help.
Privacy boundary: reading records stay in local
readwithhotsoup/— no upload, no account. Telemetry (Sumenia) is off by default and won't bother them in the UI.
Scenario 5: Hang sip into a QQ / WeChat / Discord / Telegram group bot
Goal: You @ the bot in a group chat and it answers using sip, only from sources you subscribe to.
Steps:
# 1. Install OpenClaw (or Cherry Studio) per the bot integration tutorial
# 2. Put sip.exe on PATH, put the sip-rss skill into the agent's skill directory
# 3. Paste the system prompt (with the "run sip --search first" rule)
# 4. @ it in the group@bot any noteworthy RSS updates in the last two days? → runs sip --today --json
@bot find articles about "LLM Agent" → runs sip --search "LLM Agent" --jsonSuccess looks like: the bot returns trusted summaries with source links, and every cited source is in your whitelist.
Full tutorial: Bot Integration.
Scenario 6: The intake loop — cleaner and cleaner over time
Goal: Continuously optimize the feed list itself, so your information stream gets "righter" over time.
Steps:
# 1. See reading facts (sip never concludes for you)
sip --insights # per feed: opened/finished/completion rate/skipped/likes
# 2. Make "decisions" about what bothers you — rules take effect only after you confirm
sip --policy lower_frequency --feed 3 # feed 3 updates too often → lower frequency
sip --policy archive --feed 5 # feed 5 not wanted → archive
sip --policy tag #useful --feed 2 # tag feed 2
# 3. Cross-feed duplicates: confirm after diff, then hide (data kept, reversible)
sip --dedup scan # mark "possibly the same article" — outputs duplicate clusters (representative + members)
sip --dedup hide-cluster <representativeId> # hide the whole cluster in one shot (keep the representative, hide the rest)
sip --dedup undo <key> # change your mind and undo (key shown in list output)Success looks like: --insights has real facts, and after --policy takes effect -l shows the tag/frequency changes. AI never writes rules automatically — every decision is createdBy: user.
FAQ
| Question | Answer |
|---|---|
--search reports "AI not configured" | Run sip --init first; if "no index", run sip --index |
| Summaries too short to enjoy | sip --fulltext <id> fetches full text; --show prefers it automatically |
| Want fixed-time daily updates | sip --schedule <id> daily@HH:mm, fetched only when due |
| New articles didn't make today's list | Today's list is cached (one bowl a day); read new content with sip --grep / --show |
| Migrating from FreshRSS | Export OPML → sip --import-opml file.xml |
Next steps
- All commands at a glance: CLI Mode
- Feature details: Features Overview