LLM Tutorials

Learn how to run DeepSeek OCR-2 locally with vLLM and build an incremental PDF Q&A system that performs mini-batch OCR, FAISS-based retrieval, and real-time querying with a Gradio interface.

Run GLM-5, the best open-weight AI model, on a single GPU with llama.cpp, and connect it to Aider to turn it into a powerful local coding agent.

A complete guide to Claude Code plugins. Discover how to install extensions, choose between Skills and MCPs, and build a custom session logger from scratch.

Run Qwen3-Coder-Next locally on an RTX 3090 with llama.cpp, then vibe code a complete analytics dashboard in minutes using Qwen Code CLI.

Learn how to run a top open-source model locally with llama.cpp, connect it to the Kimi CLI, and one-shot an interactive game using vibe coding.

Learn how to combine local FLUX.2 Klein 4B generation with API-based image editing, multi-reference conditioning, and session history to create an image editor with Gradio.

Run GLM 4.7 Flash locally (RTX 3090) with Claude Code and Ollama in minutes, no cloud, no lock-in, just pure speed and control.

Learn what Moonshot’s Kimi K2.5 is, how Agent Swarm works, and see it in action through four hands-on, real-world experiments.

Set up OpenClaw(formerly Clawdbot/Moltbot), a self-hosted agent connecting Claude to your Mac via WhatsApp. Search files and run shell commands from your phone.

Upgrade to Transformers v5. A practical guide to the unified Rust backend, API changes, and side-by-side v4 vs v5 migration patterns for encoding and chat.

Explore the architecture of Google’s managed MCP servers and learn how to turn LLMs into proactive operators for BigQuery, Maps, GCE, and Kubernetes.

Setting up llama.cpp to run the GLM-4.7 model on a single NVIDIA H100 80GB GPU, achieving up to 20 tokens per second using GPU offloading, Flash Attention, optimized context size, efficient batching, and tuned CPU threading.