Flowly

🇰🇷 한국어

Smart Workflow Platform for Seamless Flow of Work

Flowly is an integrated collaboration tool that connects task management and real-time chat, where conversation seamlessly flows into work and records.

---

1. Project Overview

• One-line Summary: A smart workflow project that experimented with technical expansion in large-scale environments through simulated traffic scenarios.
• Core Values:
  • Integration of Information: Organically connects chat and tasks to preserve collaboration context.
  • Technical Challenge: Established a system architecture to handle simulated traffic targeting 150k DAU.
  • Intelligent Support: Provides knowledge-based answers through an AI Agent layer.

---

2. Infrastructure Evolution

Flowly records the process of identifying and overcoming technical limits in simulated traffic environments.

(1) Background & Situation

This project assumed a challenging environment with a target DAU of 150k. We continuously identified system limits through traffic-generator and evolved the infrastructure. The core goal was to achieve both high availability and cost efficiency within limited resources.

(2) Problem Definition & Constraints

• CI/CD Scalability Limits: The initial Terraform user-data.sh approach caused bottlenecks in configuration management and automated deployment as infrastructure complexity increased.
• Cost Efficiency Issues: Accumulated AWS cloud costs for simulation environments necessitated a transition to a more controllable home server-based On-Premise setup.
• Operational Complexity: Nginx configurations directly on the host system lacked isolation, requiring manual work for every change—a major constraint during rapid development.

(3) Resolution Process

We went through 4 stages of optimization based on limits identified by traffic-generator.

1. [Step 1] Initial Build (Terraform usage): Started with a single shell script build but identified its limitations in configuration management.
2. [Step 2] Hybrid Automation (Terraform + Ansible): Separated infrastructure provisioning and server configuration to build a flexible cluster expansion and automated deployment system in AWS.
3. [Step 3] Cost-Effective Infrastructure (On-Premise conversion): Migrated to a home server for cost optimization. Thanks to the pre-built Ansible scripts, we could migrate relatively easily while maintaining consistent infrastructure specifications across environments.
4. [Step 4] Infrastructure Agility (Containerized Nginx): Switched to Docker container-based Nginx to solve configuration management inconveniences, securing version control for settings and deployment flexibility.

(4) Result

• Infrastructure Flexibility: Built a high-availability architecture independent of the environment through abstraction with Ansible and Docker.
• Cost Optimization: Secured a sustainable experimental environment by eliminating fixed infrastructure costs.

---

3. Product Description & Technical Details

Tech Stack

• Backend:
  • Spring Boot (Kotlin): Secure user authentication and authorization utilizing Spring Security.
  • NestJS (Node.js/TS): High-concurrency core domain management (Todo, Chat, Friends) leveraging NodeJS's asynchronous non-blocking I/O.
  • AI Agent (Spring AI): RAG-based knowledge search and async worker layer.
• Data Layer:
  • PostgreSQL (pgvector): Primary and vector data storage.
  • Redis: Session caching, AI Agent task queue, and real-time message broker.
• Observability & Logging:
  • LogPilot, LogPilot-Lite: Self-built open-source based event streaming and centralized logging engine.
  • Prometheus & Grafana: Real-time system metric monitoring.

Architecture

graph TD
    User[User] -->|HTTPS| Nginx[Nginx Proxy]
    Nginx -->|/api/auth| Spring["Spring Boot<br/>(Auth/Security)"]
    Nginx -->|/api/todos, /api/chat| Nest["NestJS<br/>(Core Domain)"]
    Nginx -->|/| Next["Next.js<br/>(Web Client)"]

    subgraph "Data Layer"
        Nest --> Redis[(Redis)]
        Spring --> Redis
        Nest --> DB[(PostgreSQL)]
        Spring --> DB
    end

    subgraph "AI Agent Layer"
        Nest -->|Trigger| AI_API[AI API]
        AI_API -->|Job Queue| Redis
        Redis -->|Consume| AI_Worker[AI Worker]
        AI_Worker -->|Knowledge Search| DB
    end

Loading

---

4. Documentation

For detailed technical specifications and guides, please refer to the documents below.

• System Architecture: Detailed system structure and data flow
• Troubleshooting: Records of technical issues resolved