Explore Verified Production Tech Stacks
13+ battle-tested open-source software architectures with architectural rationale, reliability metrics, and instant AI scaffolding prompts.
Connects enterprise databases directly to AI and machine learning models using standard SQL statements. Enables developers to query, train, and orchestrate generative AI agents and predictive models directly over PostgreSQL, Snowflake, and MongoDB data sources.
The gold standard for production LLM inference; invented PagedAttention to eliminate memory fragmentation in GPU KV caches, delivering 2x-4x throughput over standard Hugging Face runtimes.
Industry standard for context-aware reasoning; modular chains and LangGraph cyclic graphs allow orchestrating stateful agents with deterministic fallbacks and tool execution.
Self-hosted AI conversation and autonomous agent workspace supporting all major LLMs (OpenAI, Anthropic, Ollama, HuggingFace). Built with Node.js Express, MongoDB for flexible conversation trees, Redis for rate limiting and streaming buffers, and a responsive React UI.
Vector database built from scratch in Rust for low-latency similarity search and payload filtering. Utilizes memory-mapped (mmap) disk storage for billion-scale embeddings, custom HNSW graph indexing with hardware SIMD acceleration (AVX-512/NEON), and dual gRPC/REST APIs.
React framework for in-app AI copilots; bridges frontend UI state directly into LangGraph agent loops with bidirectional tool execution.
Extensible OpenAI/Ollama interface; SvelteKit frontend delivers instant responsive streaming while Python FastAPI handles multi-modal RAG and tool calls.
Universal memory layer for AI agents; extracts and persists structured personal facts and episodic memories across user chat sessions with low latency.
Visual canvas for building and debugging multi-agent LLM workflows; TypeORM handles dynamic agent graph configurations across SQLite or Postgres.
Full-stack data science & AI framework; Python business logic binds directly to reactive React frontend components without frontend coding.
Frequently Asked Questions
Everything you need to know about how this directory works.
What counts as a "verified" production stack?
A verified badge means the architecture is either a well-known curated open-source project we benchmarked directly from its public repository, or a community submission whose maintainer claimed the listing and confirmed the stack, scale, and reliability details are accurate.
Can I submit my own project to this directory?
Yes. Click "Share Your Stack" to submit your production architecture — name, tech stack, team scale, and why you chose it. Submissions are reviewed and go live as community listings; you can claim and verify them later from the project page.
What is the difference between AI-Agent, Hybrid, and Classic modes?
AI-Agent means the majority of the codebase was generated by an AI coding agent with light human review. Hybrid means agent-generated code paired with active human pair programming. Classic means the codebase was hand-written with AI used only for assistance, not generation.
How is the "Most Upvoted" sort order calculated?
Upvotes are community signals from engineers confirming a stack worked well for a production use case similar to theirs. The default sort ranks by net upvotes, then by most recently verified, so newer confirmations surface architectures that are still actively maintained.
Stack information is sourced from each project's public repository (package.json, go.mod, requirements.txt, or equivalent) and is not affiliated with or endorsed by the listed projects.