Tech Stack

Technologies powering the Sports Index platform

Backend

Technology	Purpose
Python 3.13	Primary language
FastAPI	Web framework and API routing
SQLAlchemy	ORM and database models
PostgreSQL	Primary database
pgvector	Semantic embeddings and similarity search
SendGrid	Transactional email delivery
OpenAI	Text embeddings (text-embedding-3-small)
S3 / Cloudflare R2	Media file storage
Pillow	Image optimization and processing
FFmpeg	GIF-to-video conversion

Frontend

Technology	Purpose
Vanilla JavaScript	No framework — lightweight, fast loading
Jinja2	Server-side HTML templating
CSS Custom Properties	Theming and dark mode
ES6 Modules	Component architecture for editors

Infrastructure

Technology	Purpose
Railway	Hosting and deployment
Docker	Containerization
Mintlify	Documentation site

Architecture Overview

Sports Index follows a layered architecture:

Frontend (Vanilla JS + Jinja2 Templates)
    ↓ HTTP/JSON
Routes / API Layer (FastAPI, ~120 endpoints)
    ↓ Function calls
Service Layer (8 services + 7 post type handlers)
    ↓ Function calls
Data Access Layer (28 repository singletons)
    ↓ ORM/SQL
PostgreSQL + pgvector

Key Patterns

• Post Type Handlers: Abstract base class pattern for 7 content types (external_media, chart, data_table, uploaded_media, timeline, page, collection)
• Canonical Deduplication: External URLs stored once globally, users link via relationships
• Dual Tagging: Owned posts use content_post_tags, canonical posts use topic_post_index
• Denormalized Caches: TopicPostIndex, TopicTimelineCache, UserLibraryCache for fast reads
• Feature-Flagged Embeddings: Semantic search gated behind EMBEDDING_ENABLED