Skip to main content

Signals and Autopublish

Overview​

Autopublish is the shared control center for author series and news rules, including automations created through MCP or REST. Overview shows enabled/paused state, real next run and timezone, channels and last results. Series and rules lists both types; Queue shows their own preparation/review/delivery states; History provides per-channel results and published-message links.

TypeMaterialsSending
Author seriesYour API or manual catalogChoose automatic sending or review for every edition
News ruleCollected and processed newsReviewer approval through Telegram, the site, or both

Start with First author series, MCP-created automations, or recovery. Catalog browsing is on the separate Materials screen. The news workflow below retains its signal and schedule modes and requires approval; that requirement does not apply to author series explicitly enabled for automatic sending.

Signals notify you about hot events.

Signals​

How signals work​

Processed news -> event clustering -> hotness evaluation -> threshold check -> Telegram alert
  1. Clustering groups related news into events.
  2. Hotness evaluates each event using factors such as source count, velocity, freshness, and AI relevance.
  3. Triggers compare the event against each subscription rule.
  4. Delivery sends the alert to Telegram when the rule is satisfied.

Signal subscription​

A user creates a subscription for a theme with conditions such as:

  • Theme: which topic of interest the subscription watches.
  • Minimum hotness: the threshold required for an alert.
  • Minimum sources: how many distinct sources must confirm the event.
  • Cooldown: the pause between alerts.
  • Telegram chat: where alerts should be delivered.

Follow-up alerts​

Signailist can issue a follow-up alert if the same event grows meaningfully after the first alert, for example through a higher score or more confirming sources.

DND (Do Not Disturb)​

  • Configurable quiet hours with start, end, and time zone.
  • Alerts that fall inside the DND window can be deferred until the quiet period ends.
  • Deferred alerts can be sent later or marked as read without sending.

News autopublish​

Principle​

In Signal mode, the rule subscribes to an existing signal and uses its thresholds. In Schedule mode, it selects processed news by topic and minimum score at the configured time; a signal subscription is not required.

Flow​

Signal fires
-> publication draft is created
-> text and or image generation runs
-> moderation card is sent
-> moderator approves or rejects
-> approved content is published to the target channel

Autopublish rule​

An autopublish rule defines:

  • Which signal or schedule triggers it.
  • Which template is used.
  • Which target channel receives the publication.
  • Which moderation path is enabled: Telegram, web UI, or both.
  • Which outputs are generated: text, image, or both.
  • Whether the rule is currently enabled.

Moderation​

  • Moderation cards are sent to assigned reviewers.
  • Reviewers can approve or reject through Telegram or the web UI.
  • All moderation actions are tracked in the product workflow.
  • Web moderation uses the same editor as digests, including synced Telegram custom emoji. See Premium emoji.

Weekly digest​

Before setting up a schedule, create and review a manual digest. Your workspace needs enabled sources linked to an active topic (and a collection account for Telegram sources), processed news, an enabled template, and sufficient balance. The example channel used to extract a writing style does not automatically become a news source.

  1. In Autopublish, create a rule and choose Trigger mode → Schedule.
  2. Enter a Rule name, select the News topic and Minimum news score.
  3. Set Cadence → Weekdays, choose a weekday, Time, and Timezone. For example, Monday at 09:00 in Asia/Tbilisi. The timezone in the form controls the schedule; check it even if your browser prefilled it.
  4. Choose the enabled Generation template and publication destination in Telegram or Discord. Configure it in Delivery channels first if the list is empty; Telegram also needs a publishing account.
  5. Choose Approval on the site, in Telegram, or both. For Telegram approval, select a connected moderation chat. Discord rules use site approval. Enable text generation; add image generation only if you need it.
  6. Review Enable immediately before saving. Leave it off while preparing the configuration; enable the rule when ready. New Discord rules are saved paused and can be enabled after saving. Check the next scheduled run on the saved rule.
  7. After a run, open the review candidate, check its text and sources, edit if needed, and approve it for publication.

The weekly rule initially looks back seven days. Subsequent selections use the previous processed run as their starting point and the scheduled run time as their end. It selects up to 30 matching items for a digest. An empty selection or one consisting entirely of previously published duplicates produces no candidate, but advances the selection boundary. This is different from the manual digest period.

A scheduled draft without a publication channel is not available yet. For a first draft without delivery setup, use the manual digest workflow. A schedule starts preparation, not immediate unreviewed publication.

If no candidate appears​

Check that the rule is enabled, the time and timezone are correct, the workspace has balance, and sources are working (including the collection account for Telegram). Check the topic-to-source links, processed news dates, minimum score, and published-duplicate filter. Review Tasks queue for pending or failed generation before retrying. Do not assume a paused or failed run will produce a separate issue for every missed period.

Publication modes​

A single-news rule prepares one matching item. A digest rule prepares a collection. Both use the review and approval pipeline before publishing to their configured Telegram or Discord destination. A rule selects one destination; manual publication in the digest editor can select several.