feat: M4 文风 + M5 生成/多provider/Skill + Kimi Code 订阅接入 + 本地联调修复

M4（文风）: style-auditor 双轨（提取指纹/漂移第四审）+ jobs 长任务框架(zombie reaper) + 回炉 refine + GET /style read-back。
M5（生成+扩展）: worldbuilder/character-gen（入库 continuity 409 gate + partition_writes 白名单 + schema→JSONB 形变）；
  网关多 provider 回退链/熔断/能力降级（Anthropic/Gemini 适配器）；Skill registry + 表权限沙箱 + 规则；
  前端 角色生成器/世界观/Codex/规则页/技能库/⌘K 命令面板。
K1（Kimi Code 订阅接入）: OAuth device-flow（kimi-code）+ 静态 Console key（kimi-code-key）两路径；
  coding 端点 KimiCLI 伪造头（实测 UA allow-list 门禁，缺则 403）+ JSON 模式结构化（thinking ⊥ tool_choice）。
本地联调修复: CORS 中间件；assemble 注入 premise+「写第N章」指令（修空 prompt 400）；
  GET /outline·/draft read-back + 大纲/工作台/审稿页重载；写页 client/server 常量边界 + notFound 健壮化；
  字数 toLocaleString locale 水合；审稿页终稿从已存草稿 seed（修 accept 422）。
门禁: backend ruff/mypy(157)/alembic 无漂移/pytest 451 · frontend lint/tsc/vitest/build。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

packages/llm_gateway/tests/test_multi_provider_accounting.py

@@ -0,0 +1,51 @@
+"""T5.4 多 provider 记账维度（ARCH §4.8）。
+
+记账行须带**实际服务方** provider/model（回退后不能记成主模型），让 usage_ledger
+的多 provider 维度可查询/聚合。
+"""
+
+from __future__ import annotations
+
+from fakes_resilience import FakeLedger, ScriptedAdapter, chain, chain_resolver, transient
+from ww_llm_gateway.gateway import Gateway
+from ww_llm_gateway.types import LlmRequest
+
+
+async def test_ledger_records_actual_serving_provider_after_fallback(req: LlmRequest) -> None:
+    primary = ScriptedAdapter("deepseek", failures=[transient() for _ in range(10)])
+    backup = ScriptedAdapter("openai", text="备用", input_tokens=222, output_tokens=33)
+    ledger = FakeLedger()
+    gw = Gateway(
+        {"deepseek": primary, "openai": backup},
+        ledger,
+        chain_resolver=chain_resolver(chain(("deepseek", "deepseek-chat"), ("openai", "gpt-4o"))),
+        max_retries=1,
+    )
+
+    await gw.run(req)
+
+    assert len(ledger.records) == 1
+    rec = ledger.records[0]
+    assert rec.provider == "openai"
+    assert rec.model == "gpt-4o"
+    assert rec.input_tokens == 222
+    assert rec.output_tokens == 33
+
+
+async def test_stream_ledger_records_actual_provider_after_fallback(req: LlmRequest) -> None:
+    primary = ScriptedAdapter("deepseek", failures=[transient() for _ in range(10)])
+    backup = ScriptedAdapter("openai", text="流式备用", input_tokens=10, output_tokens=4)
+    ledger = FakeLedger()
+    gw = Gateway(
+        {"deepseek": primary, "openai": backup},
+        ledger,
+        chain_resolver=chain_resolver(chain(("deepseek", "deepseek-chat"), ("openai", "gpt-4o"))),
+        max_retries=1,
+    )
+
+    collected = [d.text async for d in gw.stream(req)]
+
+    assert "".join(collected) == "流式备用"
+    assert len(ledger.records) == 1
+    assert ledger.records[0].provider == "openai"
+    assert ledger.records[0].model == "gpt-4o"