Views
No views yet
├── xgboost_v2_models/ # XGBoost V2 prediction models (31 tabular features)
│ ├── engagement_predictor_likes.pkl
│ ├── engagement_predictor_comments.pkl
│ ├── engagement_predictor_shares.pkl
│ ├── engagement_predictor_engagement_level.pkl
│ ├── engagement_predictor_label_encoder.pkl
│ └── feature_names.pkl
│
├── chroma_db/ # ChromaDB vector store (15,024 Instagram posts, 384-dim embeddings)
│ └── (binary files)
│
├── raw_data/ # Raw Instagram analytics data (Kaggle source)
│ └── all_posts.json # 15,024 posts with engagement metrics
│
└── scripts/
└── reindex_chromadb.py # Re-populate ChromaDB from all_posts.json (if needed)1git clone -b smartengage-ak https://github.com/AHMAD-ALYY/SmartEngage.git
2cd SmartEngage1pip install huggingface_hub
2huggingface-cli download rafau123/smartengage-models --local-dir ./smartengage-assets1# XGBoost models → models/
2cp smartengage-assets/xgboost_v2_models/*.pkl models/
3
4# ChromaDB → chroma_db/
5cp -r smartengage-assets/chroma_db/ chroma_db/
6
7# Raw data (optional, only needed for reindexing)
8mkdir -p data/instagram/raw/
9cp smartengage-assets/raw_data/all_posts.json data/instagram/raw/1cp .env.example .env
2# Edit .env and add:
3# GROQ_API_KEY=your_key
4# GEMINI_API_KEY=your_key (optional)
5# HUGGINGFACE_API_TOKEN=your_token (optional)1pip install -r requirements_backend.txt
2TOKENIZERS_PARALLELISM=false OMP_NUM_THREADS=1 python3 -m app.main
3# → http://localhost:8000| Model | Metric | Value |
|---|---|---|
| XGBoost Likes | R² | 0.974 |
| XGBoost Comments | R² | 0.821 |
| XGBoost Shares | R² | 0.883 |
| XGBoost Classifier | Accuracy | 90.7% |
| ChromaDB | Posts indexed | 15,024 |
| Embeddings | Dimension | 384 (all-MiniLM-L6-v2) |