chore(spec): 归档 add-blog-infra——Phase 0 七任务收敛交付

- T1-T7 全部勾选，spec_verify --all 七项 exit=0 入 runs.jsonl
- check 门禁收敛；交付：备份三件套/gitea 28.2.0/Artalk 2.10.0/
  git.wait4x.online + /artalk 公网接入/发布链地基与回滚预案
- 生产 Halo 全程未动（healthy + 200 复查通过）
- 含 spec pre-commit/pre-push 门禁钩子入库

Spec: 2026-10-24_add-blog-infra

Co-authored-by: GLM-5.3

.spec/specs/archive/2026-10-24_add-blog-infra/evidence/new-api-check.json

@@ -0,0 +1,11 @@
+{
+ "gateway": "http://127.0.0.1:3000 (new-api, docker: new-api)",
+ "checked_at": "2026-10-24T14:05+08:00",
+ "result": "PARTIAL",
+ "ok": ["鉴权可用（Bearer sk-，token: legion/lan-share, group=default）",
+        "/v1/models 与 /api/status 端点正常", "sqlite 直读 tokens/channels 表正常"],
+ "blocked": ["chat/completions 无可用通用模型：GLM 渠道(type14) status=2 禁用中；OpenRouter 渠道仅挂 typesafe/jev-router 与 stealth/space-bunny-alpha 两个特殊用途别名（决策协议/stealth），不供普通对话",
+             "/v1/models 偶发 404（重试可过），疑与渠道探测缓存有关"],
+ "user_action": "在 new-api 控制台启用 GLM 渠道或为 default 组补一个通用对话模型渠道；完成后 Phase 3 翻译管线即可用",
+ "phase3_impact": "不阻塞 Phase 0；Phase 3 前置依赖已记录"
+}

.spec/specs/archive/2026-10-24_add-blog-infra/evidence/verify_backup.sh

@@ -0,0 +1,15 @@
+#!/bin/sh
+# Phase 0 T1 备份验收：dump 与目录归档三件套存在、非空、结构可读
+# 路径约定见 tasks.md T1；pg_restore 宿主缺失时经 postgres 容器兜底
+set -e
+B="$HOME/backups/blog/halo-20261024"
+test -s "$B/halo_db_20261024.dump"
+test -s "$B/halo2-dir-20261024.tar.gz"
+if command -v pg_restore >/dev/null 2>&1; then
+  pg_restore --list "$B/halo_db_20261024.dump" >/dev/null
+else
+  docker run --rm -v "$B:/b:ro" postgres:15.4 pg_restore --list /b/halo_db_20261024.dump >/dev/null
+fi
+N=$(tar tzf "$B/halo2-dir-20261024.tar.gz" | wc -l)
+[ "$N" -gt 3000 ]
+echo "backup verified: dump readable, tar entries=$N"