System Architecture

Overview

TradeEntry is a full-stack web application built with a decoupled frontend/backend architecture.

┌───────────────────────────────────────────┐
│              Browser (User)                │
│         React 18 + TypeScript             │
│         Vite + Tailwind + ShadCN          │
└──────────────────┬────────────────────────┘
                   │ HTTP / REST
┌──────────────────▼────────────────────────┐
│         FastAPI (Python 3.11+)            │
│         Uvicorn ASGI server               │
│         SQLAlchemy (async)                │
│         Pydantic v2 models                │
└──────────────────┬────────────────────────┘
                   │ psycopg2 / asyncpg
┌──────────────────▼────────────────────────┐
│         PostgreSQL 14+                    │
│         NSE EOD tables (EQ + FO)          │
│         Trade journal tables              │
└───────────────────────────────────────────┘

Core Components

Frontend (/frontend)

Component	Technology
Framework	React 18 + TypeScript
Build Tool	Vite
UI Library	ShadCN / Radix UI + Tailwind CSS
State	React Hooks (useState, useCallback, useRef)
HTTP Client	Axios (src/services/api.js)
Charting	Lightweight Charts

Backend (/backend)

Component	Technology
Framework	FastAPI
ASGI Server	Uvicorn
ORM	SQLAlchemy (async) + psycopg2 (sync tasks)
Task Queue	FastAPI BackgroundTasks
Data Processing	pandas, numpy
NSE Data	Custom nseeod package

Data Pipeline (/backend/nseeod)

The nseeod package handles all NSE data ingestion:

NSE Website
    │
    ▼
BhavProcessor.run_update()
    ├── NseEodEqUpdater.update()   → scr_nseeq_eod table
    ├── NseEodFoUpdater.update()   → scr_nsefo_eod table
    └── bhav.generateamiformat()   → .txt AMI export files

Key Design Decisions

1. Background Tasks over Celery — Simple BackgroundTasks used for bhav downloads to avoid Redis/Celery infrastructure complexity.
2. File-based Status (stats.json) — Progress and logs written to stats.json for polling by the frontend, since WebSocket was not required.
3. Separate AMI Path — AmiBroker export files stored separately from CSV/bhav copies to allow independent cleanup.
4. Direct psycopg2 for Bulk Queries — Used alongside SQLAlchemy async for time-series queries that benefit from raw cursor performance.

Security Notes

• All AMI filenames validated with re.compile(r"^\d{8}\.txt$") before serving.
• Path traversal prevented via os.path.basename() normalization.
• No authentication implemented yet (planned for v2.0).

---

See Deployment Notes for production setup.