Blogify

Where research meets the world.

An AI-powered research blogging platform connecting researchers, students, and industry professionals — with smart recommendations, hybrid semantic search, content moderation, and an AI writing assistant, all driven by NLP, computer vision, and vector embeddings.

---

Features

Core Platform

• Rich-text blog editor — Tiptap-based editor with images, code blocks, LaTeX math, and link embeds
• User profiles — follow authors, view writing history, customizable bios and avatars
• Social interactions — like, dislike, bookmark, repost, and comment on posts
• Real-time notifications — in-app notification feed for follows, likes, comments, and reposts
• My blogs dashboard — manage all your drafts and published posts with inline delete

AI & ML Features

• Hybrid semantic search — combines dense vector search (MiniLM) with BM25 lexical search via Reciprocal Rank Fusion, with optional cross-encoder reranking
• Personalized "For You" feed — multi-signal ranking using semantic similarity, freshness decay, engagement quality, and author diversity (MMR)
• Similar posts discovery — vector-based nearest-neighbor recommendations
• AI writing assistant — local LLM (Qwen 2.5 via Ollama) suggests titles, excerpts, and tags from your draft
• Ask a post — readers can ask questions answered exclusively from the post's content (grounded LLM Q&A)
• Auto-summarization — embedding-based extractive 3-sentence summaries
• Auto-tagging — KeyBERT-lite keyword extraction with Maximal Marginal Relevance for diversity
• Content moderation — toxicity scoring via toxic-bert with editorial soft-flagging
• Language detection — automatic language identification (10 languages) via lingua
• Duplicate / plagiarism detection — cosine similarity flagging across authors (>0.92 threshold)
• CLIP image search — text-to-image and image-to-image search using OpenCLIP embeddings
• Contextual ad matching — privacy-first, content-based ad ranking (no user tracking)
• Explainable recommendations — UI chips showing why each post was recommended (semantic match, freshness, etc.)

---

Tech Stack

Layer	Technology
Frontend	Next.js 14 (App Router), React 18, Tailwind CSS, Tiptap Editor
Backend	FastAPI, Python 3.11+, SQLAlchemy 2 (async), Pydantic v2
Database	PostgreSQL 16
Cache / Broker	Redis 7
Vector DB	Qdrant
Object Storage	MinIO
Task Queue	Celery + Redis
LLM	Ollama (Qwen 2.5 1.5B Instruct, configurable)
ML Models	MiniLM-L6-v2 (embeddings), OpenCLIP ViT-B-32 (images), toxic-bert (moderation), ms-marco cross-encoder (reranking), lingua (language)
Reverse Proxy	Nginx
Containerization	Docker Compose

---

Project Structure

Blogify/
├── backend/              # FastAPI Python backend
│   ├── app/
│   │   ├── auth/         # JWT authentication
│   │   ├── models/       # SQLAlchemy models
│   │   ├── routers/      # API endpoints
│   │   ├── schemas/      # Pydantic request/response schemas
│   │   ├── services/     # AI/ML services (embeddings, search, LLM, etc.)
│   │   └── tasks/        # Celery background tasks (ML pipeline)
│   ├── alembic/          # Database migrations
│   └── scripts/          # Backfill & evaluation scripts
├── frontend/             # Next.js React frontend
│   └── src/
│       ├── app/          # Pages (App Router)
│       ├── components/   # Reusable UI components
│       └── lib/          # API client, auth, types, utilities
├── nginx/                # Reverse proxy configuration
├── docs/                 # Documentation
├── docker-compose.yml    # Full-stack orchestration
└── prd.md                # Product Requirements Document

---

Quick Start

Prerequisites

• Docker & Docker Compose
• ~8 GB RAM recommended (ML models load in memory)

1. Clone & start

git clone https://github.com/<your-username>/Blogify.git
cd Blogify
docker compose up --build

The first run will download ML models and the Ollama LLM (~2-3 GB total). Subsequent starts are instant.

2. Run database migrations

docker compose exec backend alembic upgrade head

3. (Optional) Seed demo data

docker compose exec backend python scripts/seed_demo.py

4. Access the app

Service	URL
Frontend	http://localhost:3000
Backend API docs	http://localhost:8000/docs
MinIO Console	http://localhost:9001
Qdrant Dashboard	http://localhost:6333/dashboard

---

Environment Variables

Create a .env file in the project root to override defaults. Key variables:

Variable	Default	Description
JWT_SECRET_KEY	change-me-in-...	Change this in production
OLLAMA_MODEL	qwen2.5:1.5b-instruct	LLM model (e.g. qwen2.5:7b-instruct)
HYBRID_SEARCH_ENABLED	true	Enable RRF fusion of dense + BM25
RERANKER_ENABLED	false	Enable cross-encoder reranking
IMAGE_SEARCH_ENABLED	true	Enable CLIP image search
MODERATION_THRESHOLD	0.75	Toxicity score threshold for flagging

See backend/app/config.py for all available settings.

---

API Highlights

Method	Endpoint	Description
POST	/api/auth/register	Register a new user
POST	/api/auth/login	Login (returns JWT)
GET	/api/posts	List published posts (paginated)
GET	/api/posts/mine	List your own posts (drafts + published)
GET	/api/posts/search/semantic	Hybrid semantic search
GET	/api/recommendations/for-you	Personalized feed
GET	/api/recommendations/similar/{id}	Similar posts
POST	/api/ai/draft	AI writing suggestions
POST	/api/ai/ask/{post_id}	Ask a question about a post
GET	/api/search/images	CLIP text-to-image search

Full interactive docs at /docs (Swagger UI) when the backend is running.

---

Architecture

Client ──► Nginx (:80) ──┬──► Next.js Frontend (:3000)
                         └──► FastAPI Backend (:8000)
                                  │
                    ┌─────────────┼─────────────────┐
                    ▼             ▼                  ▼
               PostgreSQL    Qdrant (vectors)    MinIO (files)
                    ▲             ▲
                    │             │
               Celery Workers ───┘
                    │
               Redis (broker + cache)
                    
               Ollama (local LLM)

---

License

This project is for educational and portfolio purposes.