chore: seal Design Vision v1 and monorepo scaffold

Archive the differentiated YuXinGu product docs, AI engineering system,
design contract, and Go/Vue scaffold. Next execution prioritizes Cece-parity
over early innovation (see .ai/product/STRATEGY.md).

Co-authored-by: Cursor <cursoragent@cursor.com>

.ai/adr/0001-use-golang.md

@@ -0,0 +1,20 @@
+# ADR-0001 Use Golang for API
+
+## Status
+
+Accepted
+
+## Decision
+
+Backend is Go only (`apps/api`).
+
+## Reason
+
+1. Fits small team / AI-assisted CRUD + rule engines
+2. Simple deploy (single binary)
+3. Strong concurrency for future AI/proxy workloads
+
+## Never
+
+Never replace the API language with Node/Java/Python without a new ADR and explicit approval.  
+Legacy `server.py` is temporary and must not grow new product APIs.

.ai/adr/0002-use-gin.md

@@ -0,0 +1,19 @@
+# ADR-0002 Use gin HTTP framework
+
+## Status
+
+Accepted
+
+## Decision
+
+Use `github.com/gin-gonic/gin` for HTTP routing and middleware.
+
+## Reason
+
+1. Widely known; good AI completion quality
+2. Enough features for REST MVP
+3. Low ceremony
+
+## Never
+
+Never switch to echo/fiber/chi “for fun” mid-feature. Framework change needs a new ADR.

.ai/adr/0003-rest-api.md

@@ -0,0 +1,20 @@
+# ADR-0003 REST API (not GraphQL)
+
+## Status
+
+Accepted
+
+## Decision
+
+Public HTTP API is REST under `/api/v1` with envelope `{code,message,data}`.
+
+## Reason
+
+1. Simpler mental model for H5 + mini-program
+2. Better Cursor/Claude reliability than GraphQL schemas mid-flight
+3. Easier debugging with curl
+
+## Never
+
+Never replace with GraphQL without approval + new ADR.  
+Internal gRPC later is allowed only behind a new ADR; public clients stay REST.

.ai/adr/0004-use-jwt.md

@@ -0,0 +1,19 @@
+# ADR-0004 JWT (or opaque Bearer) for auth
+
+## Status
+
+Accepted (direction)
+
+## Decision
+
+Clients send `Authorization: Bearer <token>`. Prefer JWT for stateless MVP; server may later move to opaque tokens + session store via new ADR.
+
+## Reason
+
+1. Works for H5 and future mini-program
+2. Simple middleware story
+
+## Never
+
+Never invent parallel auth headers (`X-User-Id` as sole auth).  
+Never trust client-only “isVip” flags.

.ai/adr/0005-use-redis.md

@@ -0,0 +1,22 @@
+# ADR-0005 Redis — deferred
+
+## Status
+
+Deferred (not accepted for MVP)
+
+## Decision
+
+MVP does **not** require Redis. Use PostgreSQL + process memory only when safe.
+
+## Reason
+
+1. Reduce moving parts for first ship
+2. Membership/session can start in Postgres
+
+## When to revisit
+
+Rate limiting, hot session cache, or distributed locks → write ADR-0005b Accepted before adding Redis.
+
+## Never
+
+Never add Redis “just in case” without an Accepted ADR update.

.ai/adr/0006-monorepo-vue-h5.md

@@ -0,0 +1,21 @@
+# ADR-0006 Monorepo + Vue3 user-h5
+
+## Status
+
+Accepted
+
+## Decision
+
+- npm workspaces Monorepo: `apps/*` + `packages/*`
+- Primary client: Vue 3 + TypeScript + Vite (`apps/user-h5`)
+- Shared network: `packages/sdk`
+
+## Reason
+
+1. Multi-platform path to mini-program via SDK adapters
+2. AI-friendly explicit boundaries
+3. Retire root Legacy HTML gradually
+
+## Never
+
+Never start a second unrelated frontend framework for user-h5 without ADR.