# Architecture Overview


The project is built with **FastAPI** as the main backend framework, connected to several supporting services.


# Core Components


# FastAPI – Backend Framework


- Python-based, **asynchronous**, designed for high performance.
- Project follows a **feature-based module structure** (e.g., `user`, `auth`, `search`).
- 📌 **Why FastAPI?**: API-first approach, strong community support, scalability.


---


# PostgreSQL – Relational Database


- Stores core **business and transactional data**.
- 📌 **Why PostgreSQL?**: ACID compliance, strong relational support, BI/reporting friendly.


---


# Milvus – Vector Database


- Manages **vector embeddings** from AI/ML models.
- Used for **semantic search, recommendation, and similarity search**.
- 📌 **Why Milvus?**: Open-source, scalable, optimized for AI-driven workloads.


---


# Redis – Cache & Session Management


- Provides **caching** to speed up responses and reduce database load.
- Manages **user sessions** and temporary tokens.
- 📌 **Why Redis?**: In-memory database with very high performance.


---


# Pydantic Settings – Environment Configuration


- Centralized management of configuration (DB URLs, API keys, secrets).
- Supports multiple environments: **development, testing, production**.
- 📌 **Why Pydantic?**: Reduces risk of misconfiguration, easy validation.


---


# Docker Compose – Development Environment


- Bundles services into a single development environment:FastAPI
- PostgreSQL
- Milvus
- Redis


Developers only need to run **`docker-compose up`** to start working.

📌 **Why Docker Compose?**: Ensures consistent environments across the entire team.

---


# Onboarding Guidelines


| Area | Guideline |
|---|---|
| **Code** | Follow the **feature-based module structure**. |
| **Database** | Use **PostgreSQL** for relational data, **Milvus** for vector data only. |
| **Cache** | Use **Redis** for session and caching. |
| **Config** | All settings managed via **Pydantic Settings** (no hard-coded values). |
| **Environment** | Always use **Docker Compose** instead of manual local installations. |