Initial commit: AI-driven visual novel scaffold

- Monorepo (pnpm workspace): apps/web + packages/{types,ai-client,engine}
- Next.js 16 web app with three-stage AI orchestration
- Three independently configurable providers: text LLM, image generator, vision model
- Warm minimalist editorial UI design
- One-click Vercel deploy ready

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

.env.example

@@ -0,0 +1,21 @@
+# =============================================================
+# Dada — AI Visual Novel
+# Three independently configurable AI providers
+# Any OpenAI-compatible endpoint works (OpenAI, Anthropic, Gemini,
+# OpenRouter, DeepSeek, Ollama, ...).
+# =============================================================
+
+# ---- 1. Text LLM (story director) -----------------------------
+TEXT_BASE_URL=https://api.anthropic.com/v1
+TEXT_API_KEY=sk-ant-xxx
+TEXT_MODEL=claude-opus-4-7
+
+# ---- 2. Image generator (renders the whole UI screen) ---------
+IMAGE_BASE_URL=https://api.openai.com/v1
+IMAGE_API_KEY=sk-xxx
+IMAGE_MODEL=gpt-image-2
+
+# ---- 3. Vision model (interprets where the user clicked) ------
+VISION_BASE_URL=https://generativelanguage.googleapis.com/v1beta/openai
+VISION_API_KEY=xxx
+VISION_MODEL=gemini-3-flash