v0 Alternatives — FREE Unlimited UI Generation (Arsenal Skill)

Goal: Replace Vercel v0 ($20/mo) with a 100% free, unlimited UI-generation stack — on the server AND on Dr. Wael's laptop — using OUR own keys, no extra subscriptions.

Built 2026-06-24. Live-tested ✅ (generated a full Next.js dashboard via Gemini 3.1 Pro, free).

---

Route 1 — Server-side generator (PRIMARY) ✅

skills/v0-alternatives/ui_gen.py — one-line prompt → full React/Next.js + Tailwind + shadcn.

Engines (--engine) — free by default, best on demand

| --engine | Chain | Cost | Speed | Use | |---|---|---|---|---| | free (default) | Gemini 3.1 Pro → 2.5 Pro → Flash → Opus | $0 | ~20-50s | daily / unlimited | | best / glm | GLM-5.2 (evolink → openrouter z-ai) → Gemini → Opus | ~$0.01-0.05/gen | ~90-160s | top-tier frontend (#2 Code Arena) | | gemini | Gemini → Opus → GLM | $0 | fast | force free | | anthropic | Opus → Gemini → GLM | paid | med | force Opus |

All engines are never-stop: if the first channel fails, it falls through the rest (verified live — GLM Evolink timed out, auto-fell to OpenRouter, still succeeded).

bash
source /data/.openclaw/secrets/env.sh
cd /data/.openclaw/workspace/skills/v0-alternatives
FREE default ($0) — fast, unlimited
python3 ui_gen.py "investment dashboard, portfolio cards, TRY/USD chart, dark mode" \
        --name invest --kind page
BEST quality (GLM-5.2, cheap paid) — top frontend, slower
python3 ui_gen.py "pricing section, 3 tiers, RTL Arabic" --kind component --engine best

Why GLM-5.2 for best (value verdict): Z.ai GLM-5.2 = #1 open-weight (AA Index 51) + #2 Code Arena frontends (1,595 Elo). Vs our existing top frontend engine Opus 4.8 it gives equal/higher frontend quality at ~1/5 the cost ($1.40/$4.40 vs $5/$25) — a clean best-value win, so it is the RECOMMENDED engine for any UI deliverable that matters (not a free-only afterthought). Cost is ~1-5 cents/generation.

• Keep free (Gemini, $0) for fast iteration / bulk experiments (GLM is a reasoning model,

slower ~2-3 min).

• Keep Opus 4.8 for hardest marathon/agentic + general reasoning, where GLM is weaker

(cheaper-but-less-accurate is rejected by our accuracy-first rule).

• qwen-codingplan/qwen3.7-max = flat-rate prepaid alt for general coding (zero marginal

cost) but below GLM on frontend Code Arena.

Doctrine: zero-subscription is NOT mandatory — a paid option that is cheaper than the existing one AND matches/exceeds its value is adopted (Best-Value Hierarchy). Output → output/<name>/page.tsx (or component.tsx) + a README (how to use + deploy). SSL gotcha handled (brew py3.14 → forces /etc/ssl/certs/ca-certificates.crt).

Deploy the result (free edge)

bash
python3 ../vercel/vercel_tool.py deploy output/<name> --prod   # Vercel Hobby (personal/demo)
or copy into a Next.js app on Hostinger for commercial/KCPC

Route 2 — Our coding agents (deepest, free) ✅

For a FULL app (not just a screen), drive Claude Code + Superpowers / Codex (GPT-5.6 Sol) / Kilo: they scaffold + iterate a complete Next.js + shadcn project. This is strictly more powerful than v0 (v0 = single screens; agents = whole product + tests + deploy).

Route 3 — Dyad on the laptop (v0-like visual app, free, BYOK) ✅

Dyad = open-source local AI app builder ("like v0/Lovable/Bolt, on your machine"). Free, bring-your-own-key — runs on our free Gemini / OpenRouter keys. Desktop GUI → best on the GPD Win Max 2 laptop (not the headless server).

Install on the laptop (Windows): 1. Download: https://www.dyad.sh/ (or GitHub: https://github.com/dyad-sh/dyad/releases) 2. Install + open the app. 3. Settings → AI → add a provider key: - Google Gemini (free tier) — paste GEMINI_API_KEY, model gemini-2.5-pro / gemini-3.1-pro-preview - or OpenRouter — paste OPENROUTER_API_KEY (free models available) 4. New App → describe in plain language → live preview + editable code. Export / git push. Result: full visual v0 experience, $0, unlimited (bounded only by the free key quotas).

Route 4 — Google Stitch (free, hosted, quick mockups) ✅

https://stitch.withgoogle.com (Google Labs) — the only fully-free hosted option with no meaningful limits. Browser, prompt → UI design + frontend code. Good for fast visual drafts; then refine/deploy via Route 1/2.

---

Decision guide

| Need | Use | |---|---| | Quick screen, on server, scripted | Route 1 (ui_gen.py) | | Full app / product | Route 2 (coding agents) | | Visual v0-like editing on laptop | Route 3 (Dyad) | | Fast free mockup in browser | Route 4 (Google Stitch) |

No v0 subscription. No AI Gateway card. 100% free, our keys, unlimited.

Gotchas

• ui_gen.py needs GEMINI_API_KEY (or GOOGLE_API_KEY) sourced from env.sh.
• Vercel Hobby = personal/demo deploys only; commercial/KCPC → Hostinger.
• Dyad is a desktop GUI app → laptop, not the headless server.