Generative AI Full-Stack Learning Guide: A Complete Resource Library from Beginner to Practical Application

This open-source project provides a systematic generative AI learning roadmap covering core technologies such as Python basics, Large Language Models (LLMs), prompt engineering, RAG architecture, AI agents, and vector databases, suitable for both beginners and advanced developers. Adopting a progressive learning philosophy, the project organizes content according to technical dependencies, helping users master the full set of skills from basic concepts to practical applications and avoid fragmented learning.

Amid the rapid iteration of generative AI technology, developers face the challenge of systematically mastering skills. This project was created to address this pain point, providing a complete A-to-Z learning path. The core concept is "progressive learning", organizing content based on technical dependencies and learning curves: starting from Python programming basics, gradually diving into Transformer architecture, LLM principles, then moving to cutting-edge fields like prompt engineering, RAG, and AI agent development—users of different levels can find their entry point.

Large Language Models (LLMs) Basics

Gain an in-depth understanding of components like Transformer's self-attention mechanism and positional encoding, implement a simplified Transformer from scratch, master the differences between models like GPT and BERT, and understand the relationship between model scale and emergent capabilities, as well as the characteristics of mainstream open-source models (Llama, Mistral, etc.).

Prompt Engineering

Covers zero-shot/few-shot prompting, chain-of-thought, self-consistency, and the ReAct framework; emphasizes structured prompts (role setting, output specifications); introduces automatic prompt optimization techniques (APE, OPRO).

RAG Architecture

Breaks down RAG components: document splitting, embedding model selection, vector database selection, retrieval optimization, and re-ranking techniques; provides code examples to convert unstructured documents into knowledge bases and improve the accuracy of generated content.

AI Agents

Explains core concepts like tool use, planning, memory, and multi-agent collaboration; through cases, master building systems for autonomous task decomposition, API calling, and long-term memory maintenance, covering frameworks like LangChain Agents and AutoGPT.

Vector Databases

Compares the characteristics of mainstream databases like Milvus and Pinecone; explains embedding model selection and fine-tuning, vector indexing algorithms (HNSW, IVF), and hybrid retrieval strategies.

Provides end-to-end practical projects such as chatbots, document question-answering systems, and code generation assistants, including data flow design, model selection, performance optimization, and deployment solutions (local/cloud/edge). Focuses on production-ready details: handling model hallucinations, implementing output traceability and interpretability, designing human-machine collaborative feedback mechanisms, and building a closed loop for continuous evaluation and iteration.

Beginners are advised to solidify Python and machine learning basics in the recommended order before diving into specialized skills; advanced developers can directly jump to modules of interest and use code templates to quickly build prototypes. The project has an active discussion area where learners can exchange experiences, share troubleshooting records, and get technical updates, forming a community-driven dynamic knowledge network.

This project is a knowledge map in the field of generative AI, helping users avoid getting lost in fragmented information and move from basics to practice. As new directions like multimodal models, world models, and embodied intelligence develop, systematic learning resources become even more valuable. For developers who want to build solid generative AI capabilities, this is a high-quality starting point worth investing time in.