LLM.

Agentic RAG Explained: Turning Retrieval into a Planning and Reflection Loop

A practical guide to Agentic RAG: how it differs from traditional RAG, how the control loop works, common architectures, implementation patterns, …

FastGPT in Depth: RAG, Workflows, and Self-Hosting

A practical guide to FastGPT architecture, RAG retrieval, visual workflows, self-hosting, APIs, production tuning, and project fit.

KV Cache 深入理解：为什么只存 K/V，不存 Q

从 Transformer 自注意力、Prefill/Decode、缓存张量形状、显存公式到工程优化，系统讲清楚 KV Cache 的工作方式，以及为什么旧 Query 没有缓存价值。

How Token Counting Works: How AI Providers and API Proxies Track Usage

A practical explanation of tokenizers, context windows, input and output tokens, streaming usage, tool calls, multimodal billing, and how API proxy …

About the timing of fine-tuning

Classic interview question:How exactly should you set RAG retrieval top-k?

RAG Fully Explained:From "Closed-Book" to "Open-Book" LLMs