patrick/sojorn
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity
sojorn/sojorn_docs/reference/PROJECT_STATUS.md

9.6 KiB
Raw Blame History

Sojorn - Project Status

Last Updated: January 6, 2026

Overview

The Supabase backend foundation for Sojorn is complete. All core database schema, Row Level Security policies, Edge Functions, and moderation systems have been implemented.

✅ Completed

Database Schema (5 Migrations)

  1. Core Identity and Boundaries (20260106000001)

    • ✅ profiles
    • ✅ categories
    • ✅ user_category_settings
    • ✅ follows (with mutual follow checking)
    • ✅ blocks (bidirectional, complete separation)
    • ✅ Helper functions: has_block_between(), is_mutual_follow()

  2. Content and Engagement (20260106000002)

    • ✅ posts (with tone labels and CIS scores)
    • ✅ post_metrics (likes, saves, views)
    • ✅ post_likes (boost-only, no downvotes)
    • ✅ post_saves (private bookmarks)
    • ✅ comments (mutual-follow-only)
    • ✅ comment_votes (helpful/unhelpful)
    • ✅ Automatic metric triggers

  3. Moderation and Trust (20260106000003)

    • ✅ reports (strict reasons, immutable)
    • ✅ trust_state (harmony score, tier, counters)
    • ✅ audit_log (complete transparency trail)
    • ✅ Rate limiting functions: can_post(), get_post_rate_limit()
    • ✅ Trust adjustment: adjust_harmony_score()
    • ✅ Audit logging: log_audit_event()

  4. Row Level Security (20260106000004)

    • ✅ RLS enabled on all tables
    • ✅ Policies enforce:
      • Block-based invisibility
      • Category opt-in filtering
      • Mutual-follow conversation gating
      • Private saves and blocks
      • Trust state privacy

  5. Trending System (20260106000005)

    • ✅ trending_overrides (editorial picks with expiration)
    • ✅ RLS policies for override visibility

Seed Data

  • ✅ Default categories (12 categories, all opt-in except "general")

Edge Functions (13 Functions)

  1. publish-post ✅

    • Validates auth, length, category settings
    • Runs tone detection
    • Rejects profanity/hostility immediately
    • Assigns CIS score
    • Enforces rate limits via can_post()
    • Logs audit events

  2. publish-comment ✅

    • Validates auth and mutual follow
    • Runs tone detection
    • Rejects hostile comments
    • Stores with tone metadata

  3. block ✅

    • One-tap blocking (POST) and unblocking (DELETE)
    • Removes follows automatically
    • Silent, complete separation
    • Logs audit events

  4. report ✅

    • Validates target existence
    • Prevents self-reporting
    • Prevents duplicate reports
    • Tracks reporter accuracy in trust counters

  5. feed-personal ✅

    • Chronological feed from followed accounts
    • RLS automatically filters blocked users and disabled categories
    • Returns posts with user-specific like/save flags

  6. feed-sojorn ✅

    • Algorithmic "For You" feed
    • Fetches 500 candidate posts (last 7 days)
    • Enriches with author trust and safety metrics
    • Ranks by authentic engagement algorithm
    • Returns paginated, ranked results
    • Includes ranking explanation

  7. trending ✅

    • Category-scoped trending
    • Merges editorial overrides + algorithmic picks
    • Eligibility: Positive/Neutral tone, CIS >= 0.8, no safety issues
    • Ranks by authentic engagement
    • Limited to last 48 hours

  8. calculate-harmony ✅

    • Cron job for daily harmony score recalculation
    • Gathers behavior metrics for all users
    • Calculates adjustments based on:
      • Blocks received (pattern-based)
      • Trusted reports
      • Content rejection rate
      • False reports filed
      • Validated reports filed
      • Natural decay over time
    • Updates trust_state
    • Logs adjustments to audit_log

  9. signup ✅

    • Handles user registration and profile creation
    • Creates profile and initializes trust_state via trigger

  10. profile ✅

    • GET/PATCH user profiles
    • View other user profiles (public data only)
    • View and update your own profile

  11. follow ✅

    • POST to follow a user
    • DELETE to unfollow a user
    • Enforces block constraints

  12. appreciate ✅

    • POST to appreciate (like) a post
    • DELETE to remove appreciation

  13. save ✅

    • POST to save a post (private bookmark)
    • DELETE to unsave a post

Shared Utilities

  • ✅ supabase-client.ts – Client creation helpers
  • ✅ tone-detection.ts – Pattern-based tone classifier
  • ✅ validation.ts – Input validation with custom errors
  • ✅ ranking.ts – Authentic engagement ranking algorithm
  • ✅ harmony.ts – Harmony score calculation and effects

Documentation

  • ✅ ARCHITECTURE.md – How boundaries are enforced
  • ✅ HOW_SHARP_SPEECH_STOPS.md – Deep dive on tone gating
  • ✅ README.md – Complete project overview

🚧 In Progress / Not Started

Frontend (Flutter Client)

  • ❌ Project setup
  • ❌ Onboarding flow
  • ❌ Category selection screen
  • ❌ Personal feed
  • ❌ Sojorn feed
  • ❌ Trending view
  • ❌ Post composer with tone nudges
  • ❌ Post detail with comments
  • ❌ Profile view with block/filter controls
  • ❌ Reporting flow
  • ❌ Settings (data export, delete account)

Admin Tooling

  • ❌ Report review dashboard
  • ❌ Content moderation interface
  • ❌ Trending override management
  • ❌ Brigading pattern detection
  • ❌ Role-based access control

Additional Edge Functions

  • ❌ Post view tracking (dwell time)
  • ❌ Category management endpoints
  • ❌ Data export endpoint
  • ❌ Account deletion endpoint

Transparency Pages

  • ❌ "How Reach Works" (user-facing)
  • ❌ "Community Rules" (tone and consent guidelines)
  • ❌ "Privacy Policy"
  • ❌ "Terms of Service"

Infrastructure

  • ❌ Cloudflare configuration
  • ❌ Rate limiting at edge
  • ❌ Monitoring and alerting
  • ❌ Backup and recovery procedures

Testing

  • ❌ Unit tests for tone detection
  • ❌ Unit tests for harmony calculation
  • ❌ Unit tests for ranking algorithm
  • ❌ Integration tests for Edge Functions
  • ❌ RLS policy tests
  • ❌ Load testing

Enhancements

  • ❌ Replace pattern-based tone detection with ML model
  • ❌ Read completion tracking
  • ❌ Post view logging with minimum dwell time
  • ❌ Analytics dashboard for harmony trends
  • ❌ A/B testing framework for ranking algorithms

Decision Log

Completed Decisions

  1. Text-only at MVP – No images, video, or links
  2. Supabase Edge Functions – No long-running servers
  3. Pattern-based tone detection – Simple, replaceable, transparent
  4. Harmony score is private – Users see tier effects, not numbers
  5. All categories opt-in – Except "general"
  6. Boost-only engagement – No downvotes or burying
  7. Mutual follow for comments – Consent is structural
  8. Trending is category-scoped – No global trending wars
  9. Editorial overrides expire – Nothing trends forever
  10. Blocks are complete – Bidirectional invisibility

Open Decisions

  • ML model for tone detection – Which provider? Cost? Latency?
  • Flutter state management – Riverpod? Bloc? Provider?
  • Admin authentication – Separate admin portal or in-app roles?
  • Data export format – JSON? CSV? Both?
  • Monitoring stack – Sentry? Custom logging?

Deployment Checklist

Before public beta:

  • All Edge Functions deployed to production
  • Database migrations applied
  • Categories seeded
  • RLS policies verified with security audit
  • Rate limiting enabled at Cloudflare
  • Harmony calculation cron job scheduled
  • Transparency pages published
  • Rules and guidelines finalized
  • Data export functional
  • Account deletion functional
  • Flutter app submitted to app stores (if mobile)
  • Beta signup flow ready
  • Monitoring and alerting configured
  • Backup procedures tested

Performance Considerations

Current State

  • RLS policies may be slow on large datasets (needs indexing review)
  • Feed ranking is CPU-intensive (candidate pool of 500 posts)
  • Harmony recalculation is O(n users) (may need batching)

Optimization Opportunities

  • Add materialized views for feed candidate queries
  • Cache trending results per category (15-min TTL)
  • Batch harmony calculations (process 100 users at a time)
  • Add Redis for session and feed caching
  • Pre-compute engagement scores on post creation

Security Audit Needed

  • Review all RLS policies for bypass vulnerabilities
  • Test block enforcement across all endpoints
  • Verify mutual-follow checking cannot be gamed
  • Audit SQL injection risks in dynamic queries
  • Test rate limiting under load
  • Review audit log for PII leaks
  • Verify trust score cannot be manipulated directly

Next Immediate Steps

  1. Build Flutter client MVP (onboarding, feeds, posting)
  2. Implement user signup flow (Edge Function + profile creation)
  3. Deploy Edge Functions to Supabase production
  4. Write transparency pages ("How Reach Works", "Rules")
  5. Add basic admin tooling (report review, trending overrides)
  6. Security audit of RLS policies
  7. Load testing of feed endpoints
  8. Schedule harmony calculation cron job

Success Metrics (Future)

  • Friendly retention: Users return daily for genuine connection
  • Low block rate: < 1% of relationships result in blocks
  • High save-to-like ratio: Thoughtful curation > quick reactions
  • Diverse trending: No single author/topic dominates
  • Trust growth: Average harmony score increases over time
  • Low report rate: < 0.1% of posts reported
  • High report accuracy: > 80% of reports validated

Contact

For implementation questions or contributions, see README.md.