–posted today
–posted first try
–median time in queue
–
Select all
0 in view
0 selected
Nothing here.
Page 1 of 1
Select an item to review it.
Compose
add a link to the queue by hand
lands in pending
Reddit links pull the title, image and stats automatically. Article links use the page title unless you override it below.
preview
·
—
queued for review
Post text will appear here.
where it goes
statuspending
categorynews
accountcategory default
sourcemanual
Accounts
each account posts its own categories
0 accounts
Add an account
posts drafted for its categories go out as this account
gettr.com as that account → DevTools → any API request → the x-app-auth header's token value.
Sources discover; Queue decides and delivers.
GDELT records first appear in the raw inventory below. Only records with readable original-publisher text enter Queue, where the existing account screening, scoring, drafting, moderation, canonical-event dedupe and GETTR delivery continue unchanged. Scheduled discovery checks one due source every two minutes to avoid GDELT rate-limit bursts.
0Configured sources
0Enabled
0Provider cooldown
0Degraded
0Raw routes
0Evidence pending
0Sent to Queue
0Quality filtered
0Evidence blocked
Pre-policy GDELT backlog
0 unscreened records can be re-checked against the current publisher and topic policy without calling AI.
Shared trusted-publisher catalog · 0 enabled Tier A/B domains
Trusted-only sources reject unlisted publishers before article download or AI. Per-source overrides are available under Edit quality.
| Publisher | Tier | Eligible categories | Notes |
|---|
| Source | Destination | Query | Quality policy | Cadence | Health | Raw items |
|---|
No regular-account sources are configured.
Raw discoveries
Records collected before screening and editorial analysis
| Detected | Article | Publisher | Destination | Evidence | Route |
|---|
No raw discoveries match these filters.
Page 1 of 1
Submit a YouTube podcast
Vizard Model v2 generates clips up to three minutes. Its temporary video and full transcript are evaluated by the configured AI; qualifying clips enter the shared Queue as native videos.
0Total clips
0Pending analysis
0Sent to Queue
0Did not qualify
0Failed
Vizard projects
No podcast projects yet.
Generated clips
AI decisions are based on each clip’s full transcript.
No clips match these filters.
Alerts
Official sources, monitored in real time
Official sources
Media events
X
Editorial Scope
Feed items
Sources
Shared First Notice delivery controls
Simplified Chinese mirror
Loading Chinese mirror status…
Advanced processing settings
AI output tokens and attempts are shared with media-event analysis. Changes take effect on the next run.
Government scoring policy
These values are stored in D1 and apply to future government First Notice analyses. Mandatory categories still bypass the overall-score threshold, but not the two confidence gates.
First Notice pipeline capacity
Loading
Operational pressure across Official, Media and X analysis. Review and held records are excluded.
Loading current pipeline metrics…
0Ready to publish
0Needs review
0Pending AI
0Posted
0Held
0Failed
Workflow status
Each alert appears in one operational state. Posted, skipped, and failed delivery records are removed from Needs review.
Advanced audit filters
Use the preserved First Notice and delivery states for historical audits. Selecting one resets the primary workflow filter to All.
First Notice classification
Delivery record
Source inventory
View sources
Review every monitored endpoint, its priority level, current cadence,
adapter, destination account, health, and alert history.
Media events
Clustered reporting scored with the shared First Notice policy
Official sources
Media events
X
Editorial Scope
Feed items
Sources
Shared First Notice delivery controls
Analysis runs every five minutes. A single publisher can only reach review. Automatic eligibility requires an official-source match or at least two independent publishers.
Simplified Chinese mirror
Loading Chinese mirror status…
Advanced processing settings
Clustering groups raw feed items that describe the same real-world event before AI analysis. AI output tokens and attempts are shared with government analysis.
Media scoring policy
These values are stored in D1 and apply to future media-event analyses. Mandatory categories still bypass the overall-score threshold, but not the two confidence gates. Government scoring is configured separately.
First Notice pipeline capacity
Loading
Operational pressure across Official, Media and X analysis. Review and held records are excluded.
Loading current pipeline metrics…
0Ready to publish
0Needs review
0Pending AI
0Posted
0Held
0Failed
Workflow status
Each event appears in one operational state. Posted, skipped, and failed delivery records are removed from Needs review.
Advanced audit filters
Use the preserved AI and delivery states for historical audits. Selecting one resets the primary workflow filter to All.
AI analysis
Delivery record
Media feed items
Raw records collected before event analysis
Official sources
Media events
X
Editorial Scope
Feed items
Sources
Ingestion staging records
Seeded items are the initial feed backlog and are intentionally excluded from event analysis. New items wait to be clustered; clustered items link to a media event.
0Total items
0New
0Clustered
0Seeded
0Failed
| Article | Source | Status | Issued | Detected | Event |
|---|
Feed items are not posts; only verified, analyzed events can reach delivery.
X monitoring
Selected accounts gated by deterministic keywords
Official sources
Media events
X
Editorial Scope
Sources
X delivery controls
Credentials, cadence and the per-poll cap are managed in Settings. Full media details are requested only after the keyword gate passes.
0Total detected
0Filtered
0Pending AI
0Ready
0Posted
0Needs review
0Failed
Monitored accounts
Enabling an account starts at that exact moment. Earlier posts are never imported.
| ACCOUNT | DESTINATION | ACTIVATED | HEALTH | POSTS |
|---|
Keyword policy
Normal keywords require two distinct, non-overlapping matches. Any priority keyword passes alone. Matching is case-insensitive and whole-word/phrase aware.
Detected X posts
Only original posts are eligible. Replies, reposts and quote posts are excluded.
No X posts match this filter.
Editorial Scope
Shared canonical-event selection and per-language delivery audit
Official sources
Media events
X
Editorial Scope
Sources
Policy activation
Loading policy modes…
0Evaluated
0Selected
0Excluded
0Selected by both
0Scope only
0Current scoring only
0Posted
0Analysis failed
0Delivery failed
0Duplicate blocked
Canonical events
Shared event identity across Official, Media, X, and regular accounts
Official sources
Media events
X
Sources
Shared event resolver
Deterministic identity checks run first. Ambiguous wording is retrieved with BGE-M3, reranked with Workers AI, and sent to MiniMax only when the cheaper stages cannot decide safely.
Resolver settings
0Total events
0Active
0Published
0Cross-registry
0Vectors pending
0Resolver errors (24h)
Resolver health
Quality review
Automatic matches, resolver errors, and newly created events with a close rejected candidate. Confirming a result records calibration feedback; it does not delete or publish anything.
| WHEN | CANDIDATE EVENT | COMPARED EVENT | DECISION | SCORES | FEEDBACK |
|---|
No matches, errors, or borderline separations need review yet.
Resolved events
Members show every source record assigned to the event. Regular-account publication claims are independent per destination account.
No canonical events match this filter.
Alert sources
Official and media/public monitoring
Official sources
Media events
X
Editorial Scope
Feed items
Sources
0Total sources
0Enabled
0Level 1
0Media sources
0Terms blocked
0Degraded
| Source | Priority | Adapter | Destination | Activity | Health | Feed items |
|---|
Official Level 1 polls every 5 minutes; Level 2 polls every 30 minutes. Both levels can auto-post Ready reports.
Media Level 1 polls every 30 minutes; Level 2 polls every 60 minutes. Both levels can auto-post verified Ready events.
Poll now Checks the visible registry one priority lane and one source shard per Worker request.
Event lane Media articles are normalized, clustered by event, verified, scored, and drafted before delivery.
Terms A source must record operator acceptance before it can be enabled.
These are the deployed cadences today. Conditional requests reuse ETag and
Last-Modified validators when an upstream source supports them.
Analytics
GETTR and syndicated-platform performance, last 14 days
–posts (14d)
–likes
–reposts
–comments
Publishing lanes
| Account | Posts | Last 7d | Likes | Reposts | Comments | Best post | Avg eng/post |
|---|
Nothing posted (for real) in the last 14 days yet.
| Top posts | Account | Lane | Likes | Reposts | Comments | Views |
|---|
No measured posts yet — stats refresh hourly after posting.
Stats refresh hourly for Queue, Official, Media, and X posts from the last 14 days (dry-run posts excluded). Engagement = likes + reposts.
Facebook, Instagram, Threads and TikTok
Cached Facebook, Instagram, Threads and TikTok insights from GenQueue. Platform credentials remain in GenQueue; metrics are not combined with GETTR totals.
| Top syndicated posts | Platform | Views | Reach | Interactions | Likes | Comments / replies | Shares / reposts |
|---|
No cached GenQueue social insights yet. Use Refresh stats now, or wait for the hourly refresh.
Facebook interactions = reactions + comments + shares. Instagram uses Meta's total interactions when available. Threads = likes + replies + reposts + quotes + shares. TikTok = likes + comments + shares.
Log
every ingestion run lands here
–runs today
–new items
–duplicates skipped
–filtered out
–failed runs
–estimated X cost today
| When | Source | Trigger | New / requests | Duplicates / resources | Filtered | Est. cost | Notes |
|---|
No runs logged yet.
0AI calls today
$0.0000calculated cost today
0input tokens
0cached input
0output tokens
0failed / unpriced
| When | Workflow / operation | Model | Record | Input | Cached | Output | Reasoning | Cost | Status |
|---|
No AI calls logged yet.
Calculated cost uses the effective-dated model rates stored in D1. MiniMax billing remains authoritative after vouchers, credits, subscription allowances or provider adjustments.
manual = the Pull button ·
cron = this worker on its schedule ·
push = the Reddit-hosted app or the Mac pusher delivering items.
X API cost is estimated from resources returned before X's same-day billing deduplication. Confirm actual charges in the X Developer Console.
Settings
worker configuration
unsaved changes
Devvit ingestion
The Devvit app (devvit-app/ in the repo) pushes Reddit candidates here.
Paste this secret into the app's subreddit settings as ingestSecret.
…
Ingestion source
Where the hourly cron (and the "Pull from Reddit now" button) gets its
candidates. Exactly one source is active at a time — switch whenever you like.
Supabase dashboard → Project Settings → API Keys → create a
secret key (legacy service_role JWTs also still work; publishable keys
won't — they can't pass RLS). Also run supabase-schema.sql (repo root)
once in the SQL editor.
Apify console → Settings → API & Integrations → create token.
Ingestion
Cost lever for Apify (pays per result).
Doesn't affect Direct mode.
Categories and their subreddits are managed per account on the
Accounts page.
AI drafting
Any OpenAI-compatible provider works — Kimi (Moonshot), GLM (Zhipu),
MiniMax, DeepSeek, OpenRouter, OpenAI. Paste the provider's base URL (ending in
/v1), a model name, and an API key.Reasoning models think inside this budget —
use 2000+ for MiniMax/DeepSeek.
Leave empty to use the built-in prompt. Each account can override this
on the Accounts page.
One person per line. Separate aliases with
|.
Existing AI calls avoid adverse framing, and the final posting boundary holds drafts containing nearby hostile language.AI screening & auto-posting
Every hour the AI scores pending items 0–100 (relevance, trend,
freshness, news value, sourcing, engagement) and tags a decision + risk on each. Uses the
same provider configured under AI drafting.
Leave empty to use the built-in instruction
(tiered sourcing, risk rules, diversity limits). The JSON output contract at the end must
stay intact if you customize.
X API monitoring
Credentials and cost controls for selected-account monitoring at Alerts → X. Use an app Bearer Token from the X Developer Console. The token is masked after saving and is never returned to the browser.
Text and attachment keys are retrieved first. Full Media resources are requested only for posts that pass the keyword gate.
Vizard podcast clipping
API access requires a paid Vizard plan. YouTube projects use Model v2, vertical output, subtitles and clips up to three minutes. Qualifying transcript analyses enter the shared Queue.
Reddit RSS feed token
Compose looks reddit links up over RSS, which Reddit throttles hard for
anonymous callers (~1 request/minute, shared across all Cloudflare apps). Personal feed
tokens make those lookups authenticated and un-throttled — copy them from any feed URL on
reddit.com/prefs/feeds (the
user= and feed= values).Native video uploader
The Vercel function merges Reddit's separate tracks and remuxes direct
X MP4 variants before GETTR transcodes them. X video delivery fails closed when this is not configured.
Must match the MUX_SECRET environment variable in the Vercel project.
Reddit Data API fallback (optional)
Only needed if you want the worker's hourly cron / "Pull now" to fetch
Reddit directly instead of relying on the Devvit app.
Access
Leave empty to keep the current PIN.
applies from the next ingestion run