A production-grade backend service that intelligently scores and validates user-generated content using AI, with category-wise qualification system for flexible user rewards.
Instead of requiring users to complete ALL daily activities, users can now qualify for rewards in individual categories:
- 📝 Posts: Create 2 quality posts → qualify for post rewards
- 👍 Likes: Give 5 likes → qualify for like rewards
- 💬 Comments: Make 5 comments → qualify for comment rewards
- 🪙 Crypto: Complete 3 crypto transactions → qualify for crypto rewards
- 💰 Tipping: Make 1 tip → qualify for tipping rewards
- 🤝 Referrals: Refer 1 person → qualify for referral rewards
- Users who don't qualify in a category can still receive empathy rewards
- Top 10% of non-qualified users per category get empathy rewards based on historical engagement
- Fair distribution ensures loyal users are recognized
- Gibberish Detection: Multi-layered approach (rule-based, statistical, ML)
- Duplicate Detection: Vector-based similarity using Weaviate database
- Quality Scoring: Local LLM (qwen2.5vl) rates content quality 0-10
- Originality Bonus: More unique content gets higher points
┌─────────────────┐ ┌──────────────────┐ ┌─────────────────┐
│ Client App │───▶│ FastAPI Server │───▶│ Celery Worker │
│ (Blockchain) │ │ │ │ (AI Engine) │
└─────────────────┘ └──────────────────┘ └─────────────────┘
│ │
▼ ▼
┌─────────────────┐ ┌─────────────────┐
│ PostgreSQL │ │ Weaviate │
│ (User Scores) │ │ (Embeddings) │
└─────────────────┘ └─────────────────┘
│ │
▼ ▼
┌─────────────────┐ ┌─────────────────┐
│ Redis │ │ Ollama │
│ (Task Queue) │ │ (Local LLM) │
└─────────────────┘ └─────────────────┘
| Category | Max Points | Percentage | Daily Requirement |
|---|---|---|---|
| Posts | 30 | 27.3% | 2 posts/day |
| Crypto | 20 | 18.2% | 3 transactions/day |
| Tipping | 20 | 18.2% | 1 tip/day |
| Likes | 15 | 13.6% | 5 likes/day |
| Comments | 15 | 13.6% | 5 comments/day |
| Referrals | 10 | 9.1% | 1 referral/day |
Final Score = (User's Total Points / 110) × 100
Post Points = Base (0.5) + Quality Bonus (0-1.0) + Originality Bonus (0-0.25)
- Docker & Docker Compose
- Python 3.11+
- Ollama with qwen2.5vl model
-
Clone and Setup
git clone <repository> cd User_Validation_Scoring python -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate pip install -r requirements.txt
-
Start Infrastructure
docker-compose up -d postgres redis weaviate multi2vec-clip
-
Install Ollama Model
ollama pull qwen2.5vl ollama run qwen2.5vl # Keep running in separate terminal -
Start Services
# Terminal 1: Start Celery Worker celery -A celery_worker worker --loglevel=info # Terminal 2: Start API Server uvicorn api.main:app --reload --host 0.0.0.0 --port 8000 # Terminal 3: Start Scheduler (Optional) celery -A celery_worker beat --loglevel=info
-
Verify Setup
curl http://localhost:8000/health curl http://localhost:8000/admin/category-summary
POST /v1/submit_action
Content-Type: application/json
{
"creatorAddress": "0x...",
"interactorAddress": "0x...",
"Interaction": {
"interactionType": "like|crypto|tipping|referral",
"data": "optional_data"
}
}POST /v1/submit_post
Content-Type: multipart/form-data
creatorAddress: 0x...
interactorAddress: 0x...
interactionType: post
data: "Post content here"
webhookUrl: https://your-webhook.com/callback
image: [optional file upload]GET /admin/category-summaryGET /admin/daily-summaryGET /admin/user-activity/{wallet_address}POST /admin/run-daily-analysis# Test crypto interaction
curl -X POST http://localhost:8000/v1/submit_action \
-H "Content-Type: application/json" \
-d '{
"creatorAddress": "0x1234567890abcdef1234567890abcdef12345678",
"interactorAddress": "0x1234567890abcdef1234567890abcdef12345678",
"Interaction": {
"interactionType": "crypto",
"data": "BTC_TRADE_12345"
}
}'python testing/category_test_script.py-
Individual Category Check: For each category (posts, likes, comments, crypto, tipping, referrals):
- Check which users met the daily requirement
- Mark them as "qualified" for that specific category
-
Empathy Selection: For each category:
- Calculate empathy scores for non-qualified users
- Select top 10% as empathy recipients
-
Reward Distribution:
- Send category-wise API calls with qualified and empathy user lists
- Users can qualify for multiple categories simultaneously
{
"posts": {
"qualified": ["0x1111...", "0x4444..."],
"empathy": ["0x5555..."]
},
"likes": {
"qualified": ["0x2222...", "0x4444..."],
"empathy": ["0x1111..."]
},
"crypto": {
"qualified": ["0x3333...", "0x4444..."],
"empathy": []
}
}# Daily qualification requirements
POST_LIMIT_DAY = 2
LIKE_LIMIT_DAY = 5
COMMENT_LIMIT_DAY = 5
CRYPTO_LIMIT_DAY = 3
TIPPING_LIMIT_DAY = 1
REFERRAL_LIMIT_DAY = 1
# Point values
POINTS_PER_POST = 0.5
POINTS_FOR_CRYPTO = 0.5
POINTS_FOR_TIPPING = 0.5
# ... etc
# Empathy reward percentage
REWARD_PERCENTAGE_OF_INACTIVE = 0.10 # Top 10%CREATE TABLE user_scores (
user_id VARCHAR(255) PRIMARY KEY,
points_from_posts REAL DEFAULT 0.0,
points_from_likes REAL DEFAULT 0.0,
points_from_comments REAL DEFAULT 0.0,
points_from_referrals REAL DEFAULT 0.0,
points_from_tipping REAL DEFAULT 0.0,
points_from_crypto REAL DEFAULT 0.0,
daily_posts_timestamps TIMESTAMPTZ[],
daily_likes_timestamps TIMESTAMPTZ[],
daily_comments_timestamps TIMESTAMPTZ[],
daily_referrals_timestamps TIMESTAMPTZ[],
daily_tipping_timestamps TIMESTAMPTZ[],
daily_crypto_timestamps TIMESTAMPTZ[],
last_active_date DATE,
consecutive_activity_days INTEGER DEFAULT 0,
historical_engagement_score REAL DEFAULT 0.0
);python clean.py # Resets all data# API logs
tail -f api_logs.log
# Celery worker logs
celery -A celery_worker worker --loglevel=debug
# Docker logs
docker-compose logs -f api worker- Users needed to complete ALL daily activities to qualify
- Lower participation rate
- Users forced into activities they didn't prefer
- Users qualify independently for each category they're active in
- Higher participation and engagement
- Users can focus on preferred activities
- More flexible and user-friendly
- Better retention and satisfaction
# Update connection strings in code to use service names
# postgres -> postgres:5432
# redis -> redis:6379
# weaviate -> weaviate:8080
docker-compose up --buildPOSTGRES_HOST=postgres
CELERY_BROKER_URL=redis://redis:6379/0
CELERY_RESULT_BACKEND=redis://redis:6379/0
WEAVIATE_HOST=weaviate
OLLAMA_HOST_URL=http://host.docker.internal:11434- API Documentation: http://localhost:8000/docs
- Health Check: http://localhost:8000/health
- Category Status: http://localhost:8000/admin/category-summary
This system revolutionizes user engagement by allowing independent qualification per category, making it much more user-friendly and increasing overall participation while maintaining quality standards through AI validation.