CLI Mode

sip -l                  # list all feeds
sip -l 1                # list articles of feed 1 (ID format [list index/real ID])
sip -d https://xxx/rss  # download a new RSS feed
sip -u 1                # update feed 1
sip -a 1                # archive (adds a timestamp)
sip -una 1              # unarchive
sip -r 1                # remove a feed
sip -h                  # help
sip --lang en-US -l     # switch to the English UI

Full-screen reading: sip --show <article-id> opens a sidebar-less full-screen reading interface (Markdown-rendered body). At the bottom it prompts "Press W to enter the full reader · Press Esc to exit" — press W to seamlessly switch into the full TUI (positioned at the current article), or Esc/Q to return to the command line.

AI reading the original: sip --show <article-id> --json prints the article's title/source/link/publish time/author + raw body (no rendering) as JSON to stdout for AI or scripts, e.g. sip --show 42 --json --lang en-US --ignoresafeannouncement. If the article's full text was fetched, the JSON also includes a fulltext field (plain-text body, more complete than the RSS summary — AI should use it first when answering).

Version tracking & Diff:

sip --versions 42            # list all historical versions of article 42 (with status and time)
sip --show 87 --json         # 87 might be the ID of a historical version, still readable
sip --diff 42                # diff the body between the two latest versions
sip --diff 42 v1 v3 --json   # specify two versions, structured output {from, to, changes:[{type,before,after}]}

--versions takes the global article ID from --show/--grep results; each version is an independent DB row with its own ID. When an article has only one version, a notice is printed (exit code 0, not an error). The IDs in -l <feed-id> listings use the [list index/real ID] dual format — take the right-hand real ID when using --show/--versions/--summary.

Parameter Reference

Short	Long	Description
-l	--list	List all feeds; with an ID, list that feed's articles (-l --json / -l 1 --json structured output, includes health status and content quality). ID format [list index/real ID]; commands like --show/--versions/--summary use the right-hand real ID
-d	--download	Download a new RSS feed (http/https prefix optional, auto-completed)
-u	--update	Update a specific feed (by ID)
-a	--archive	Archive the current snapshot (adds a timestamp)
-una	--unarchive	Unarchive (checks for same-name conflicts)
-r	--remove	Remove a feed and all its articles and vectors (add --yes/-y to skip confirmation, for scripts/AI non-interactive use)
--show <id>		Full-screen reading (no sidebar; W enters full TUI, Esc exits); with --json outputs unrendered original JSON to AI/scripts
--versions <id>		List all historical versions of an article (with status and time; --json structured); to read an old version use --show <that version's id>
--diff <id> [vA vB]		Diff an article's body between two versions (default: two latest); --json structured output for AI
--export <id | feed:N | all> [out.md|dir]		Export articles as Markdown (asks before --export-all; --yes skips)
--fulltext <id>		Fetch an article's full text to the local cache (consent required first time; --yes skips consent/confirmation, --json structured); --purge-fulltext [id] clears the cache
--feed-info <id>		Feed identity & health: name/type/author/website/last update/latest article/status (--json structured)
--export-opml [file]		Export all feeds as OPML (default feeds.opml)
--import-opml <file>		Bulk import feeds from OPML (skips existing by FeedUrl)
--like <id> [--ai [reason]]		Mark an article: user like (♥) or AI judgment (🤖); --likes [--json] views all marks
--today [--json] [--refresh] [--quick N]		Today's reading list (rule-based selection, cap = target 5 articles; includes estimated time and reasons). One fixed bowl per day (cached for the day; new articles don't auto-enter the list that day); --refresh regenerates explicitly; for same-day new content use --grep/--show directly; enable Sumenia to track completion progress
--sync [--feed N] [--json]		Update only feeds that are "due" (optional --feed <id> to limit to one; --json structured)
--update-all		Force-update all feeds (equivalent to TUI F6)
--schedule <id> <expression>		Set a feed's update schedule (see Update Scheduling): 30m / 1h / 7d / daily@10:00 / weekly@Mon 08:00 / manual
--purge-fulltext [id]		Clear the full-text cache (no ID = clear all; see Full-Text Fetch)
telemetry status|show|enable|disable|clear|export		View/toggle/delete/export the local reading telemetry Sumenia (off by default; see Telemetry & Privacy)
--init / --config / --index / --reindex / --search / --grep / --summary		AI-related commands, see AI Commands
-h	--help	Show help

Global parameters: --ignoresafeannouncement (skip extraneous output like the safety banner, for scripts/AI), --lang <code> (switch language, e.g. --lang en-US). Output is always UTF-8.

Exit Codes (for scripts / AI to judge success)

CLI commands exit 0 on success, and return a non-zero code by category on failure:

Exit code	Meaning
0	Success (including normal cancellation, e.g. answering n to the -r confirmation)
1	General error (argument/usage error, unknown command, DB error, partial update failure)
2	Network / service unreachable (NETWORK_ERROR, MODEL_UNAVAILABLE, download timeout)
3	Resource not ready (AI not configured, API key missing/invalid, NO_INDEX, feed/article not found, empty query)

In --json mode, errors still output a structured {"success": false, "error": {...}} first, then exit with the corresponding non-zero code.