CmdWise Tech Stack Analysis & Choices
Purpose
This article captures the thinking behind CmdWise’s technology choices for frontend – docs – backend – deployment, making future maintenance and expansion easier.
1. Frontend Website
1.1 Stack
| Goal | Solution |
|---|---|
| Page rendering | Pure HTML + JavaScript + Tailwind CSS |
| Icons | Lucide |
| Light/Dark theme | Tailwind Dark Mode |
Why not Vue / React?
- Startup speed: Static HTML previews instantly without a build step—great for rapid iteration.
- Learning cost: I hadn’t used Vue/React for a while; getting back up to speed would take time.
- AI-friendly: LLMs generate plain HTML snippets more reliably, which can be saved and used right away.
1.2 Workflow
- Design styles based on
ui-style-guide. - Use
index.htmlas a reference and let Cursor generate each page. - Manually tweak details to ensure consistency.
1.3 Deployment (Cloudflare Pages)
bash
wrangler deploy --minify- After deploy, bind a custom domain to obtain HTTPS automatically, no manual certificates.
- Static site → zero server cost.
2. Docs
2.1 Tooling
- VitePress: supports Chinese/English switching.
- Follow SEO & editorial style rules, with AI helping layout and polishing.
2.2 Flow
- Decide directory structure → configure VitePress.
- Ask AI to review for rationality.
- Write Chinese draft → polish for SEO → manual review → AI generates English version.
3. Backend Server
3.1 Framework
- Hono — natively integrates with Cloudflare Workers and provides LLM helper docs, so AI tools can read API specs and generate code easily.
3.2 Cloud Services
| Feature | Service |
|---|---|
| Runtime | Cloudflare Workers |
| Database | D1 (SQLite) |
| Object Storage | R2 |
Choosing the Cloudflare family—Workers + D1 + R2—means one-line configuration, almost zero ops.
3.3 Local Development
wrangler devspins up local D1 & R2 under.wrangler/stateand supports breakpoint debugging, almost like a local Node server.- You can open D1 with any SQLite GUI; the Cloudflare console also shows data, but the browser plugin Drizzle Studio feels smoother.
3.4 Deployment Notes
bash
wrangler deploy --minify- Just add a
wrangler.jsoncorwrangler.tomlat root, then run the command above. - Add
keep_vars = trueinwrangler.tomlto avoid env vars being reset on deploy. - For secrets,
wrangler secretis faster than pasting in the Dashboard (though the Dashboard now supports multi-line paste too).
4. Desktop App
Options: Electron vs Tauri
| Metric | Electron | Tauri |
|---|---|---|
| Community & ecosystem | ★★★★★ | ★★★★☆ |
| Cross-platform | macOS / Windows / Linux | Same |
| Debug experience | Mature | Newer |
After weighing needs, AI (Cursor) chose Electron + React.
Why:
- Electron has a mature ecosystem with plenty of examples; its footprint in AI training data is large, so code generation is easier.
- React + Electron is a proven combo with lots of community cases, reducing the risk of pitfalls.
5. Conclusion
Using the Cloudflare stack plus a simple frontend keeps deployment cheap and development fast. With AI, we can validate ideas quickly and iterate continuously.