@QingQ77: Open-source architecture knowledge base featuring 25 real-world system architecture templates and a systematic architecture thinking tutorial. https://github.com/study8677/awesome-architecture… This repo focuses on architecture, not code. It includes 25 architecture templates for popular systems.

Open-Source Architecture Knowledge Base: 25 Real-World System Architecture Templates + A Systematic Architecture Thinking Tutorial. https://github.com/study8677/awesome-architecture… This repo doesn’t talk about code, only architecture. It includes architecture templates for 25 popular systems, covering e-commerce, social feeds, payments, AI gateways, RAG, AI Agents, and AI coding products like Claude Code and Codex. Each template breaks down key decisions and common pitfalls. — # study8677/awesome-architecture Source: https://github.com/study8677/awesome-architecture # Awesome Architecture · Architecture Atlas > An open-source knowledge base focused on architecture, not code. > Collects real-world system architecture templates and a tutorial to help you become a better architect. English · 简体中文 🌐 Read Online (Interactive · Bilingual, HTTPS): 🧭 Companion Skill: architecture-copilot (https://github.com/study8677/architecture-copilot) — turn this knowledge into an interactive skill that guides you step by step through architecture design in Claude Code / Cursor / Codex. — ## Why This Repo Exists In the past two decades, a programmer’s core strength was “writing code correctly and quickly.” But something is happening right before our eyes: “writing code” is disappearing. Not becoming harder or less frequent, but ending as a “scarce craft done by humans.” At frontier labs like OpenAI and Anthropic, code is almost entirely written by AI. Human engineers no longer type out implementations — they do only two things: tell the AI what to build, and judge whether it built it right. When a machine can spit out working code in seconds, once-valued skills like “should I use for or map? Do I know this API? Am I fluent in this syntax?” become worthless overnight. What truly won’t depreciate — and will only become more valuable — is another ability: > Before writing the first line of code, clearly envision what the system should look like. > > Where does data come from? Where does it go? Which parts must be strongly consistent, and which can be eventually consistent? > What will break first? When users grow from 10,000 to 100 million, where is the first bottleneck? > What am I willing to give up (B) in order to get A? That is architecture thinking. It has nothing to do with a specific language, framework, or even what’s trending this year. It’s a judgment of “look at the map before hitting the road.” This repo’s belief: The best developers of the future will first be people who make sound architectural judgments, and only second be people who write code. You should first have a clear architectural understanding of what you’re building; code is just one means to implement that understanding. — ## What’s in This Repo awesome-architecture/ ├── tutorial/ 📚 Tutorial — systematically teaches you "how to think like an architect" └── templates/ 🗺️ Templates — architecture maps of real popular systems, covering architecture only, not syntax ### 📚 tutorial/ — Becoming a Better Architect Tutorial Not about “how to use a certain framework,” but about a transferable method of thinking: how to break vague requirements into constraints, make decisions among trade-offs, draw architecture diagrams that communicate clearly, and design a new system from scratch. | Chapter | Topic | What You’ll Learn | |—|—|—| | 01 | Why Architecture Thinking First | Why “architecture first” is a core skill in this era | | 02 | The Architect’s Thinking Framework | Generic process: Requirements → Constraints → Quality Attributes → Trade-offs | | 03 | Reading & Drawing Architecture Diagrams | Use the C4 model to visualize and communicate your system | | 04 | Ten Core Architectural Patterns | Layered, Microservices, Event-Driven, CQRS… what problems each solves | | 05 | Data & State | Why “data” is the real hard part of any system | | 06 | Quality Attributes & Trade-offs | How to balance performance/availability/consistency/cost | | 07 | Designing a System from Scratch | A practical methodology you can follow | | 08 | Architecture Decision Records & Evolution | Use ADRs to record decisions, let architecture grow with the business | | 09 | Architecture Taste | What sets you apart beyond frameworks; build judgment from real cases (microservices back to monolith, design aesthetics of major companies) | > 👉 Newcomers start with tutorial/README.md — there’s a complete learning path. ### 🗺️ templates/ — Architecture Templates of Real Systems Each template is an “architecture map.” We deliberately avoid discussing what language or framework is used. We only discuss: what problem this kind of system solves, what components it has, how data flows, what the key decisions and trade-offs are, and where it will fail when scaled. > Currently 25 templates (16 classic/general + 5 AI-native + 4 AI coding/autonomous agents). Each includes links to real open-source projects / engineering documentation at the end. Follow them to read the source code. Classic/General Systems: | Template | Representative Product | Key Architecture Points | |—|—|—| | AI Chat Product | Claude, ChatGPT | LLM inference, streaming, context management, RAG, cost control | | Browser Extension | Honey, Grammarly | Content script/background separation, page injection, privacy boundaries, monetization | | Standard Web App | Corporate website, blog, SaaS backend | Classic three-tier, caching, read-write separation — “good enough” | | Mobile App | Most iOS/Android apps | Offline-first, data sync, client state, push notifications | | E-commerce Platform | Amazon, Shopify, Taobao | Inventory, orders, payments, overselling, flash sale spikes | | Social Feed | Twitter/X, Instagram | Feed pull/push, follow relationships, trending topics | | Video Streaming | Netflix, YouTube | Transcoding, CDN, adaptive bitrate, recommendations | | Real-time Chat | WhatsApp, Slack, WeChat | Long connections, message ordering, offline delivery, group fan-out | | URL Shortener | Bitly, TinyURL, t.co | Read-heavy, cache, 301/302, distributed unique IDs | | Payment System | Stripe, Alipay, PayPal | Idempotency, double-entry bookkeeping, reconciliation, state machine | | Search Engine | Google, Elasticsearch | Inverted index, relevance ranking, recall+ranking, sharding | | Ride-hailing | Uber, DiDi | Geospatial indexing, real-time location, supply-demand matching, dynamic pricing | | Real-time Collaborative Doc | Google Docs, Figma | OT/CRDT, single writer serialization, operation log, offline sync | | Cloud Storage | Dropbox, iCloud | File chunking, content-addressed dedup, incremental sync, resumable upload | | Notification/Push System | Novu, FCM/APNs | Multi-channel fan-out, dedup rate limiting, async retry, priority | | Online Ticketing | Ticketmaster, Damai, 12306 | Virtual waiting room, atomic deduction to prevent overselling, seat lock timeout | 🤖 AI-Native Systems (New in the LLM era): | Template | Representative Product/Prototype | Key Architecture Points | |—|—|—| | AI Gateway/Proxy | One API, LiteLLM, Portkey | Unified API, billing/rate limiting, load balancing, failover, caching | | RAG Knowledge Base | RAGFlow, LlamaIndex, Dify | Chunking, vector search, hybrid search + re-ranking, citation provenance | | AI Agent / Workflow | Dify, Coze, LangGraph | Action loop, tool sandbox, memory, controllable fallback | | Model Inference Serving | vLLM, SGLang, Triton | Continuous batching, paged KV cache, quantization, multi-replica | | Vector Database | Milvus, Qdrant, pgvector | ANN approximate nearest neighbor, HNSW/IVF, recall-latency trade-off | 🦾 AI Coding / Autonomous Agents (New 2026, real agent product architectures): | Template | Representative Product/Prototype | Key Architecture Points | |—|—|—| | Claude Code | Claude Code (Anthropic) | Local-first coding agent, sub-agents/hooks/skills/MCP, dual-layer permission + OS sandbox, context compression | | OpenAI Codex | Codex CLI + Cloud | Dual-mode: local CLI + cloud async sandbox, sandbox × approval axes, default no-network for injection prevention, auto-open PR | | OpenClaw (Lobster 🦞) | OpenClaw (formerly Clawdbot) | Self-hosted Gateway, chat software as UI, heartbeat/cron, pluggable harness, memory = plain text | | Hermes | Hermes (Nous Research) | Resident self-growth, FTS5 persistent memory, automatic skill distillation, cron, multi-channel / multi-provider | > 👉 Want to add your own template? Use the unified format in templates/_TEMPLATE.md. — ## How to Use This Repo If you’re a beginner / want to shift to architecture thinking: Read through tutorial/ in order. After each chapter, pick a system from templates/ that interests you, and try to “read” it using what you just learned. If you’re about to design a new system: Start with tutorial/07 for methodology, then find the template closest to your scenario in templates/. Use it as a starting point, not an answer — go through each “Key Decision” and “Common Pitfall” and ask yourself the same questions. If you’re preparing for a system design interview: Each template in templates/ covers high-frequency topics (overselling, feed fan-out, message ordering, streaming output…), organized in a consistent format for systematic review. If you’re a senior engineer / architect: Go directly to “Key Decisions & Trade-offs” and “Evolution Path” in each template — that’s the most concentrated part. Contributions of your own hard-learned lessons are welcome. — ## Three Reading Principles 1. Ask “why” first, then “how.” Every architectural choice comes from a constraint or a trade-off. If you don’t see the trade-off, you haven’t understood it. 2. There is no best architecture, only the most suitable one. A “chat” for an internal tool vs. for WeChat yields wildly different answers. Scale, team, cost, and compliance determine everything. 3. Architecture grows. Don’t apply a mature-stage architecture to an MVP. Each template includes an “evolution path” telling you when to upgrade and when you’re over-engineering. — ## TL;DR > Code tells the computer what to do; architecture decides whether it’s worth doing, can be done, and can withstand the load. > This repo helps you train that second kind of judgment. — ## ⭐ Star History > If this helped you, a star is the best encouragement. — ## 🔗 Friends - LINUX DO (https://linux.do/) — A new ideal community where developers passionate about tech come together to share.