Architecture

PiePLM follows a modular, layered architecture designed for extensibility and maintainability.

System Overview

┌─────────────────────────────────────────────────────────────┐
│                      Web UI (React)                          │
├─────────────────────────────────────────────────────────────┤
│                      REST API / gRPC                         │
├─────────────────────────────────────────────────────────────┤
│  ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌────────────────┐  │
│  │Requirements│ │  Design  │ │   BOM    │ │   Changes      │  │
│  │  Module   │ │ Module   │ │ Module   │ │  Module        │  │
│  └──────────┘ └──────────┘ └──────────┘ └────────────────┘  │
├─────────────────────────────────────────────────────────────┤
│                    Core Services Layer                        │
│  ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌────────────────┐  │
│  │  Search  │ │  Auth    │ │  Events  │ │  Notifications │  │
│  │  Engine  │ │  Service │ │  Bus     │ │  Service       │  │
│  └──────────┘ └──────────┘ └──────────┘ └────────────────┘  │
├─────────────────────────────────────────────────────────────┤
│                    Data Layer (PostgreSQL)                     │
└─────────────────────────────────────────────────────────────┘

Core Principles

1. Single Binary

• Pure Go implementation (no CGO dependencies)
• PostgreSQL for reliable data storage
• Easy deployment and distribution

2. Modular Design

Each domain (requirements, designs, BOM, changes) is a self-contained module with:

• Independent data models
• REST/gRPC interfaces
• Event publishing
• Extension points

3. AI-Native Integration

• Built-in vector search + embeddings
• MCP (Model Context Protocol) server
• LLM-ready APIs for agent integration

4. Local-First

• Works offline by default
• Optional cloud sync
• Data ownership stays with you

Technology Stack

Layer	Technology
Language	Go 1.25+
Database	PostgreSQL
Search	Full-text search + vector embeddings
API	REST + gRPC + MCP
Web UI	React + Vite
Build	Go modules + esbuild

Data Flow

1. Input - CLI, Web UI, API, MCP tools, file watchers
2. Processing - Validation, enrichment, embedding generation
3. Storage - PostgreSQL (relational + full-text) + vectors
4. Indexing - Real-time search index updates
5. Output - Search results, notifications, webhooks, AI agents

Extension Points

• Plugins - Custom modules via Go plugins
• Webhooks - Event-driven integrations
• MCP Tools - AI agent capabilities
• Templates - Document & workflow templates

Next Steps

• Core Modules - Domain module details
• Search Engine - How search works
• MCP Integration - AI agent setup