System Intelligence

System Intelligence Overview

What OF HQ tracks, how it learns, and what's next.

01

Live System Stats

Real-time counts from the database

Total Fans
3,486
Transactions
2,261
Creators
9
AI Classifications
0
Intent Events
0
Fan Facts
0
02

What We Track Per Fan

55+ data points across 7 categories

Money & Purchases

  • -Lifetime spend
  • -Last purchase date / time / type / amount
  • -Average order value
  • -Biggest single purchase
  • -First purchase date
  • -Price range (low / mid / high / whale)
  • -Buyer type (ppv_buyer, tipper, custom_buyer, etc.)
  • -Conversion rate (PPV sent vs bought)
  • -Discount sensitivity (never / sometimes / always)
  • -Chargeback risk flag

Fan Personality (AI-Detected)

  • -Fan type (submissive / dominant / romantic / transactional / lonely)
  • -Tone preference (playful / assertive / romantic / direct)
  • -Emotional drivers (validation, companionship, escapism...)
  • -Emotional needs
  • -Content format preference (photo / video / audio / text / bundle)
  • -Length preference (short / medium / long)
  • -Narrative summary (rolling AI portrait)

Personal Facts

  • -Hobbies
  • -Pet names
  • -Food & drink preferences
  • -Work schedule / occupation category
  • -Location / timezone / language
  • -Relationship status
  • -Body preferences
  • -Sports teams
  • -Birthday / sub anniversary

Buying Intent (14 Signals)

  • -ready_to_buy
  • -wants_custom
  • -price_question
  • -discount_request
  • -wants_more
  • -high_intent
  • -churn_risk
  • -escalation_intent
  • -trust_intent
  • -attention_intent
  • -status_intent
  • -emotional_support
  • -entertainment
  • -boundary_testing
  • -Each scored with confidence 0-100%

Relationship Stage

  • -new -> warming -> active_buyer
  • -active_buyer -> cooling_off -> at_risk -> churned
  • -reactivated (loops back to active_buyer)
  • -Stage updated timestamp
  • -Stage auto-decay via cron job
  • -Last objection / top objection tracking

Engagement Signals

  • -Temperature ring (HOT / WARM / COOLING / COLD / ICE COLD)
  • -Spend tier dot
  • -Last message time
  • -Active hours (JSON array of peak hours)
  • -Avg reply time (seconds)
  • -Time waster score (0-100)
  • -Follow-up due date
  • -Next best action + reason
  • -Deprioritize cooldown timer
03

How Data Flows In

Three ingestion methods keep intelligence fresh

Real-Time Webhooks

Instant - the second they buy

Events
subscriptions.newtips.receivedmessages.ppv.unlockedtransactions.new
Updates
Fan recordTransaction recordTelegram alertTemperature ring

Periodic Sync

Round-robin: 1 creator every 5 min ~ each synced every 45 min

Events
fans (24h lookback)transactions (24h lookback)
Updates
lifetimeSpendavgOrderValuebiggestPurchasebuyerTypepriceRange

AI Classifier

3-window analysis: first 100 msgs + around purchases + last 400 msgs

Events
fan typeintent signalstone detectionemotional driverspersonal facts
Updates
GPT-4o-mini$0.0005 per classificationConfidence scoresFan facts DB
04

Database Tables

Live row counts from every table in the system

TablePurposeLive Count
CreatorConnected OF accounts9
FanIndividual fan profiles + 55 intelligence fields3,486
TransactionTips, PPV, subs, messages2,261
FanFactStructured long-term memory per fan0
FanIntentEventBuying intent signals from messages0
FanPreferenceContent/format preference tags0
FanLifecycleEventStage changes, milestones, follow-ups3,443
ChatQAReviewChatter quality-assurance scores0
MediaAssetVault content metadata11
UserAuth accounts (agency/cfo/employee)0
CreatorAssignmentTeam-member to creator scoping0
SessionActive login sessions0
AccountOAuth provider links0
VerificationTokenEmail verification tokens0
05

What's NOT Built Yet

Roadmap items in priority order

  1. 1.
    Auto-classify on every message (real-time AI pipeline)
  2. 2.
    Vault content tagging (vision AI on images & videos)
  3. 3.
    "Has he bought this?" indicator per fan per asset
  4. 4.
    Matching layer (fan preferences <-> vault content)
  5. 5.
    Segmented mass messages (audience builder + blast)
  6. 6.
    Live revenue on dashboard (real-time websocket feed)
  7. 7.
    Team / CFO / Settings pages (role-based views)
06

Security

How every entry point is protected

Webhooks
HMAC-SHA256 signature verification on every payload
Sync Endpoint
x-sync-key header protection on all sync routes
Cron Jobs
CRON_SECRET bearer token validation
Role-Based Access
AGENCY / CFO / EMPLOYEE permission tiers
Creator Assignments
Team-member scoping per creator account
Auth Layer
NextAuth session + JWT with secure cookie handling