HarmonyOS Sudoku Case and AI DevSpace: Today's Hands-On Notes
Today, Thursday 2026-11-12, I spent my Huawei Cloud developer session on two hands-on things: reading a HarmonyOS Sudoku case in the Harmony Development Practice Center, and trying a real cloud query inside AI DevSpace. This note records what I actually did and what I actually saw, including the one result that was simply empty.
The Sudoku case I read
The case is “Sudoku: HarmonyOS + ArkUI pure client-side development” in the Harmony Development Practice Center. I opened its detail page and stayed on it for more than 40 seconds. The case page describes a Sudoku game built with ArkTS and ArkUI that runs fully offline. Its stated learning goals are declarative UI rendering of a 9x9 board, deterministic puzzle generation, and local file persistence. The page lists 54 puzzles, three pages, and an estimated 60 minutes, and it names the @ohos.file.fs file API for clearing progress. Its repository is harmony-practice-center/Harmony-APP-sudoku.git.
What I took from reading it, rather than from building it, is that a small game is a good fit for learning ArkUI: the board is a fixed grid, the state is small and easy to reason about, and saving progress locally is a concrete persistence exercise. I have not built or run this Sudoku app today, so I am not claiming a finished app or any gameplay result. The value for me today was the case design: a deterministic puzzle set, a clear board model, and a file-based save that a beginner can follow without a backend.
The AI DevSpace query I actually ran
After the case, I opened AI DevSpace from the AI Agent Development Practice Center and entered the trial version. The environment was created with a three-hour trial window. Inside AI DevSpace I opened AI Shell and used its preset prompt: “Query running ECS instances.”
AI Shell loaded the huawei-cloud-computing-query skill, ran its environment check, and reported that the environment check passed and the credentials were valid. The check found Python 3.11.16 and the Huawei Cloud SDKs, then read the ECS query guide and ran the list script with the ACTIVE status filter.
The result was straightforward and worth writing down exactly: in region cn-north-4, it found no ECS instances in the running (ACTIVE) state. AI Shell explained the likely reasons plainly: the account may have no running instance in that region, or instances may be in another region, since the default region comes from the HW_REGION_NAME environment variable. It offered to check another region, all statuses, or associated resources such as internal IPs, elastic IPs, security groups, and disks.
That empty result is still a useful outcome. It shows the skill works end to end: the query ran against real credentials, used the correct status filter, and returned a clear answer instead of an error. Before running a heavier cloud query, a read-only check like this is a sensible first step, because it confirms the environment, the skill, and the region before I spend time on anything else.
What I learned today
First, start with a small, self-contained case. The Sudoku case keeps everything on the device, so the learning is about UI and state, not about setting up a server. Second, let the agent show its work. In AI Shell I could see each step: loading the skill, checking the environment, reading the guide, and finally running the query. When a step failed, I would have known where it failed. Third, treat an empty result as data. “No running ECS instances in cn-north-4” is not a failure; it is an accurate answer for that region and that status at 14:32 on 2026-11-12.
Next time I would follow up in two ways: try the same ECS query without the status filter, to see stopped or other states, and then return to the Sudoku case to build it in the cloud practice environment instead of only reading it. For today, the reading and the query were the real work, and both are recorded here as they happened.
- 点赞
- 收藏
- 关注作者
评论(0)