Purpose
This repository contains the server-side components of PlayWithLLM, a platform for working with large language models. The architecture follows domain-driven design principles and is built as a multi-service application focused on scalability and maintainability.
NOTE: I will be separating the Inference service and Business service into two separate repositories in the future.
Key Principles
- Domain-Driven Design: Organized around business domains with clear separation of concerns.
- Microservices Architecture: Multiple specialized services (business, inference) working together.
- API-First Design: Well-structured RESTful APIs with proper validation and error handling.
- Comprehensive Logging: Structured logging system for monitoring and debugging.
TODO: * Containerization: Docker-based deployment for consistent environments.
Project Structure Overview
- root directory
- docker/ - Docker configurations for containerizing the application services.
- docs/ - Project documentation and sample data.
- scripts/ - Development, deployment, and utility scripts.
- src/ - Source code organized by service and domain.
services/- Service-specific implementations:business/- Core business logic service.inference/- LLM inference and processing service.
shared/- Common utilities and libraries:configs/- Configuration management.libraries/- Shared functionality like db setup, email, logging, email, etc.middlewares/- Express middlewares like auth, error handling, rate limiting, etc.migrations/- Database migrations when we need to run some database commands when the application starts or updates.
- test/ - Unit and integration tests.
Service Architecture
The system consists of three main services:
- Business Service (Port 4000): Manages core business logic and data operations including authentication, user management, etc.
- Inference Service (Queue listener): Interfaces with LLM models and manages inference operations.