Practical AI Engineering

Kunpeng AI Lab

Notes on getting AI coding tools and Agent workflows running in real environments.

We start from setup, troubleshooting, verification, and review notes, then organize posts into topic paths. You can begin with one concrete issue or follow the Claude Code, Agent workflow, and OpenClaw tracks.

Live Hot Topic / Current Topic

DeepSeek V4 Practical Hub

A practical DeepSeek V4 hub for API migration, Pro / Flash routing, and 1M context workflows, organized as an execution path for developers.

Open Hub

Practical Routes

Find practical content by use case

AI Tools And Environments

Setup, configuration, networking, and model access notes from real tool use, with practical paths through common failures.

Agent Workflow Cases

Reusable AI Agent workflows for coding, publishing, debugging, and content operations, grounded in real constraints.

Open Projects And Reviews

Tools, skills, upstream contributions, and postmortems that turn one-off experiments into reusable public assets.

Main Paths

Three main reading paths

Recent Field Notes

Recently updated practical notes

View all posts ->

Tools And Entries

Tool entries

View all ->

Projects

Project notes

View project list ->

JSON / XML / YAML Utility Kit

Beginner

A lightweight browser-based toolkit for common JSON, XML, and YAML formatting and conversion tasks.

Claude Code Windows Setup Lab

Intermediate

A reusable setup lab for Claude Code on Windows, covering install, shell choice, proxy troubleshooting, and verification.

ESP32 Voice Assistant

Intermediate

Smart voice assistant project based on ESP32 with offline voice recognition and smart home control.

Editorial Method

How we write a technical page

A page should make the scenario, environment, reproduction path, conclusion boundary, and next step explicit. When something can be verified, we verify it; when it cannot, we label the source and reasoning.

Read methodology ->

About this site

What does Kunpeng AI Lab focus on?

The site focuses on AI coding tools, Agent workflows, Windows / CLI troubleshooting, open-source tool practice, and real project notes rather than broad AI news coverage.

Who is this site for?

It is built for developers and small teams configuring Claude Code, OpenClaw, Codex, Cursor, and related tools, or turning Agent workflows into project work.

How is it different from a general AI news site?

The emphasis is on whether something can be reproduced, where it fails, how to verify it, and what to read next. Hot topics matter only when they connect to practical work.