@modusensus/dsh-mneme 0.8.14 → 0.8.16

package/src/api.js

@@ -1,1433 +1,1605 @@
-import { URL } from "node:url";
-import { timingSafeEqual, randomBytes } from "node:crypto";
-import { FEATURE_FLAG_SPEC } from "./settings.js";
-import { TYPE_FILE, renderMirrorText, renderFileHeader, parseHumanEdits, MIRROR_READONLY_TYPES } from "./mirror.js";
-import { langOf } from "./lang.js";
-import { computeHeat } from "./heat.js";
-import { describeStreamFailure, resolveRoute } from "./dream.js";
-import { getInjectionSnapshot } from "./inject.js";
-import { describeLocalRuntime, publicRuntimeStatus } from "./runtime/loader.js";
-import { hostModulesDir, provisionRuntime } from "./runtime/provision.js";
-import { classify, fetchLatestVersion, PACKAGE_VERSION } from "./version-check.js";
-
-// headers：少数端点（/export 附件下载）需要追加 Content-Disposition 等响应头。
-function sendJson(res, status, payload, headers = {}) {
-  res.writeHead(status, { "Content-Type": "application/json; charset=utf-8", ...headers });
-  res.end(JSON.stringify(payload));
-}
-
-// 附件下载响应（导出端点）：Content-Type 与文件名（含日期后缀）由调用方给定。
-function sendAttachment(res, status, contentType, filename, body) {
-  res.writeHead(status, {
-    "Content-Type": contentType,
-    "Content-Disposition": `attachment; filename="${filename}"`
-  });
-  res.end(body);
-}
-
-// 版本号统一从 version-check.js 取（同一份读 package.json 逻辑，读取失败
-// 同样降级 "unknown"）——/info 与 /version-check 两个路由共用，不再各写一份。
-
-// Defaults for the standalone external API — keep in step with the schema
-// defaults in config.js (externalApiPort / externalApiHost).
-const EXTERNAL_API_DEFAULTS = { enabled: false, port: 8790, host: "127.0.0.1" };
-
-/** Merge persisted external-api kv over the defaults for client display. */
-function fullExternalConfig(kv = {}) {
-  return {
-    enabled: kv.enabled === true,
-    port: Number.isInteger(kv.port) && kv.port > 0 ? kv.port : EXTERNAL_API_DEFAULTS.port,
-    host: kv.host || EXTERNAL_API_DEFAULTS.host,
-    token: kv.token || ""
-  };
-}
-
-/**
- * Mask an API key for client display: keep a recognizable prefix and suffix,
- * hide the middle. Empty keys stay empty; short keys are fully hidden.
- * The mask only exists in the API layer — storage keeps the real key.
- */
-function maskApiKey(key) {
-  if (!key) return "";
-  if (key.length <= 8) return "***";
-  return `${key.slice(0, 3)}***${key.slice(-4)}`;
-}
-
-/** True when the request carries the configured apiToken (or no token is set). */
-function isAuthorized(req, apiToken) {
-  if (!apiToken) return true;
-  const raw = req.headers?.authorization ?? req.headers?.["x-dsh-mneme-token"] ?? "";
-  const token = raw.startsWith("Bearer ") ? raw.slice(7).trim() : raw.trim();
-  if (token === "" || token.length !== apiToken.length) return false;
-  // Constant-time comparison: avoid leaking the token via timing when the API
-  // is exposed beyond loopback.
-  return timingSafeEqual(Buffer.from(token), Buffer.from(apiToken));
-}
-
-/**
- * Reject a request with 401 when auth is enabled and the token is missing or
- * wrong. Returns true when the request may proceed.
- */
-function requireAuth(req, res, apiToken) {
-  if (isAuthorized(req, apiToken)) return true;
-  sendJson(res, 401, { error: "unauthorized" });
-  return false;
-}
-
-/** Collect the request body as text (tolerant of empty/invalid bodies). */
-function readBody(req) {
-  return new Promise((resolve) => {
-    let body = "";
-    req.on("data", (chunk) => { body += chunk; });
-    req.on("end", () => resolve(body));
-    req.on("error", () => resolve(""));
-  });
-}
-
-function parseBody(text) {
-  try {
-    return JSON.parse(text || "{}");
-  } catch {
-    return {};
-  }
-}
-
-export function createApi(ctx, service, settings, commands, embedder, semantic = null, apiToken = "", config = null) {
-  const disposers = [];
-
-  // #182 注入状态探测：面板状态页的「极简模式下注入按宿主设计关闭」提示需要
-  // 知道当前会话的 agentPreset。注入回调在 minimal 下被宿主整体压制（#175 的
-  // 定论），不能作为检测源；session/event 是 preset 无关钩子（蒸馏同款），任
-  // 何事件都带完整 session 对象。只记最近一次会话头——这是面板展示用的「当前
-  // 会话」最佳近似，不是逐会话账本。读法与 scope.js 同源（session.header 冻结
-  // 的 SessionHeader，降级 requestHeader().config；agentPreset 取不到留 null，
-  // 面板对 null 静默）。任何失败静默：纯可观测性，不新增故障面。
-  let lastAgentPreset = null;
-  try {
-    const unsubscribe = ctx.on?.("session/event", (session) => {
-      try {
-        const header = session?.header ?? session?.requestHeader?.()?.config ?? null;
-        const preset = header?.agentPreset;
-        lastAgentPreset = typeof preset === "string" && preset.trim() ? preset.trim() : null;
-      } catch { /* 观测失败不影响会话流 */ }
-    });
-    if (typeof unsubscribe === "function") disposers.push(unsubscribe);
-  } catch { /* 宿主无 session/event 时面板提示退化为「未知」，不报错 */ }
-
-  // webServer 在 index.js 的 inject 声明中（cordis 等宿主服务就绪后 apply），
-  // 这里做防御性读取：cordis ctx 的 Proxy 不允许直接访问未 inject 的属性
-  // （会抛 "cannot get property without inject"），用 ctx.reflect.get 免
-  // inject 读取（未提供返回 undefined）；对象字面量 mock ctx（测试）没有
-  // reflect，退回直接属性访问。createApi 只在有 webServer 时被调用（index.js
-  // 守卫 + 测试 mock），故下方 register 用之非空。
-  const webServer = typeof ctx.reflect?.get === "function" ? ctx.reflect.get("webServer") : ctx.webServer;
-
-  // feature flags 快照（GET/PUT 共用）：overrides 是持久化的用户显式覆盖；
-  // effective 是启动配置在白名单键上被 overrides 覆盖后的最终值。config 缺席
-  // （旧调用方直连、未传 cfg）时只报被覆盖的键，不把不存在的默认值编造给前端。
-  const flagKeys = [
-    ...FEATURE_FLAG_SPEC.booleans,
-    ...Object.keys(FEATURE_FLAG_SPEC.ints),
-    ...Object.keys(FEATURE_FLAG_SPEC.numbers),
-    ...FEATURE_FLAG_SPEC.strings,
-    ...FEATURE_FLAG_SPEC.urls,
-    ...Object.keys(FEATURE_FLAG_SPEC.enums)
-  ];
-  // 嵌套键在 kv 里按点号平铺（"memoryQualityFilter.enabled"），运行时 cfg 里
-  // 是嵌套对象，effective 从对象子字段取值；其余键照旧从 cfg 顶层取。
-  // 这张表必须与 settings.js 白名单里的点号键一一对应：#254 加键时漏了这里，
-  // effective 会静默少两个键（configFlagValue 取不到 undefined 就不写进去），
-  // 面板上看不出来、PUT 也存不进去——test/api.test.js 的计数锁就是钉这件事的。
-  const NESTED_FLAG_PATHS = {
-    "memoryQualityFilter.enabled": ["memoryQualityFilter", "enabled"],
-    "llmAudit.enabled": ["llmAudit", "enabled"],
-    "writeAdmission.enabled": ["writeAdmission", "enabled"],
-    "writeAdmission.enforce": ["writeAdmission", "enforce"]
-    // sensitiveScanEnabled（#164 A2）是顶层扁平键，不需要进这张表——走 configFlagValue
-    // 的默认分支即可。
-  };
-  function configFlagValue(key) {
-    const path = NESTED_FLAG_PATHS[key];
-    if (path) return config?.[path[0]]?.[path[1]];
-    return config?.[key];
-  }
-  function featureSnapshot() {
-    const overrides = settings.getFeatureFlags();
-    const effective = {};
-    for (const key of flagKeys) {
-      if (overrides[key] !== undefined) effective[key] = overrides[key];
-      else {
-        const value = configFlagValue(key);
-        if (value !== undefined) effective[key] = value;
-      }
-    }
-    return { overrides, effective };
-  }
-
-  // Ensure the service has an embedder when the API layer was handed one
-  // (tests wire the embedder through the API instead of index.js). Without
-  // this, /api/dsh-mneme/search would silently degrade to keyword-only.
-  if (embedder && typeof service.setEmbedder === "function") {
-    service.setEmbedder(embedder);
-  }
-
-  const register = (route) => {
-    disposers.push(webServer.register(route));
-  };
-
-  // /api/dsh-mneme prefix fallback → 404 JSON for unknown sub-paths
-  register({
-    kind: "prefix",
-    path: "/api/dsh-mneme",
-    handler(req, res) {
-      sendJson(res, 404, { error: "not-found" });
-    }
-  });
-
-  // 反馈入口预填用的插件版本（面板「帮助与反馈」卡片拉取）。读取失败（打包
-  // 环境）返回 "unknown"，链接照常可用，纯展示信息不拦截。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/info",
-    handler(req, res) {
-      sendJson(res, 200, { version: PACKAGE_VERSION });
-    }
-  });
-
-  // 版本自检（#174 后续）：运行版本 vs npm registry latest。只读，宿主围栏
-  // 统一鉴权（与 dream-status 同款，面板 apiFetch 直接可用）。registry 查询
-  // 全失败时 latest=null、status=unknown——前端对 unknown 完全静默，版本提示
-  // 是锦上添花，绝不因它新增故障面。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/version-check",
-    handler(req, res) {
-      if (req.method !== "GET") {
-        sendJson(res, 404, { error: "not-found" });
-        return;
-      }
-      fetchLatestVersion()
-        .then((latest) => sendJson(res, 200, {
-          version: PACKAGE_VERSION,
-          latest,
-          status: classify(PACKAGE_VERSION, latest),
-          checkedAt: new Date().toISOString()
-        }))
-        .catch(() => sendJson(res, 200, { version: PACKAGE_VERSION, latest: null, status: "unknown" }));
-    }
-  });
-
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/list",
-    handler(req, res) {
-      try {
-        const url = new URL(req.url, "http://localhost");
-        const type = url.searchParams.get("type") ?? undefined;
-        const limit = Number(url.searchParams.get("limit") ?? 50);
-        const offset = Number(url.searchParams.get("offset") ?? 0);
-        // order=chrono: pure newest-first for paged browsing; default keeps
-        // the importance-ranked order other callers rely on.
-        const order = url.searchParams.get("order") ?? undefined;
-        // minImportance: numeric lower bound on importance (absent/NaN → no
-        // floor); source: exact match on the source column (empty → no filter).
-        const minRaw = url.searchParams.get("minImportance");
-        const minImportance = minRaw !== null && minRaw !== "" && !Number.isNaN(Number(minRaw))
-          ? Number(minRaw)
-          : undefined;
-        const source = url.searchParams.get("source") || undefined;
-        // updatedFrom/updatedTo：updated_at 闭区间过滤（ISO 日期或完整时间戳）。
-        // 边界归一化在 store 的 updatedAtBounds（list/count 共用同一纯函数）
-        // 完成，非法值在那里被忽略——这里原样透传即可。
-        const updatedFrom = url.searchParams.get("updatedFrom") ?? undefined;
-        const updatedTo = url.searchParams.get("updatedTo") ?? undefined;
-        // archived=only：只看归档（状态页的归档列表用）。归档行不进默认列表，
-        // 所以这是独立的视图开关，而不是 includeArchived 的混看模式。
-        const onlyArchived = url.searchParams.get("archived") === "only";
-        // deposited=only：只看 autoDream 巩固过的记忆（receipt_chain 的
-        // merge/update verdict ∪ source=dream 直写）——状态页「查看全部」
-        // 与记忆库的「沉淀」筛选 chip 共用这个视图。
-        const depositedOnly = url.searchParams.get("deposited") === "only";
-        const rows = service.list({ type, limit, offset, order, minImportance, source, updatedFrom, updatedTo, onlyArchived, depositedOnly });
-        // 面板行在 wire DTO 之上补 archived/quality_score——模型工具的输出
-        // schema 严格复用 toApiList，扩展只发生在 HTTP 层。
-        // heat 投影（阶段二前端数据源）：仅 heatEnabled=true 时下发逐条热度
-        // （heat.js 纯函数，λ=0 免疫类型恒 1.0）；字段缺省时前端徽章自动隐藏。
-        const heatOn = config?.heatEnabled === true;
-        const items = service.toApiList(rows).map((m, i) => ({
-          ...m,
-          archived: rows[i].archived === true || rows[i].archived === 1,
-          quality_score: rows[i].quality_score ?? null,
-          ...(heatOn ? { heat: computeHeat(rows[i], Date.now(), config ?? {}) } : {})
-        }));
-        // Total honors the same filters as the rows, or the pager's
-        // has-more math breaks whenever minImportance/source/updated-at
-        // bounds are active.
-        sendJson(res, 200, { items, total: service.count(type, { minImportance, source, updatedFrom, updatedTo, onlyArchived, depositedOnly }) });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // 记忆复用统计（#217，只读聚合）：Top-N 召回 + 僵尸率 + 覆盖度标注。窗口
-  // 参数钳制 1-365 天（缺省 30），聚合口径见 service.recallStats；注入命中
-  // 率待注入留痕口径拍板后接入。失败走既有 500 JSON 惯例，前端整卡不渲染。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/recall-stats",
-    handler(req, res) {
-      if (req.method !== "GET") {
-        sendJson(res, 404, { error: "not-found" });
-        return;
-      }
-      try {
-        const url = new URL(req.url, "http://localhost");
-        const raw = Number(url.searchParams.get("window"));
-        const windowDays = Number.isInteger(raw) && raw >= 1 && raw <= 365 ? raw : 30;
-        sendJson(res, 200, service.recallStats({ windowDays }));
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/search",
-    handler(req, res) {
-      try {
-        const url = new URL(req.url, "http://localhost");
-        const q = url.searchParams.get("q") ?? "";
-        const limit = Number(url.searchParams.get("topK") ?? url.searchParams.get("limit") ?? 20);
-        // mode selects the recall strategy (defaults to auto):
-        //   auto    (default) keyword first, vector fills remaining slots
-        //   hybrid  vector first, keyword fills remaining slots; scores of
-        //           memories hit by both sides are weight-blended
-        //   vector  vector only, falls back to keyword when the vector path
-        //           is unavailable (no embedder or a throwing one)
-        //   keyword literal text only; never queries the embedder
-        // rerank=false disables the cross-encoder reorder for this request;
-        // the response `mode` field reports which path actually produced rows.
-        const mode = url.searchParams.get("mode") ?? "auto";
-        const rerank = url.searchParams.get("rerank") !== "false";
-        const query = q.trim();
-        if (!query) {
-          sendJson(res, 200, { items: [], mode: "keyword" });
-          return;
-        }
-        // Route through the unified semantic pipeline; any vector/rerank
-        // failure degrades to keyword results inside searchMemories. The
-        // returned promise lets the test double await the async search.
-        return Promise.resolve(
-          service.searchMemories(query, { mode, topK: limit, useRerank: rerank })
-        ).then((rows) => {
-          // mode reflects what actually happened: rows marked `vector` came
-          // through the semantic path, everything else is keyword fallback.
-          const used = rows.some((m) => m.vector === true) ? "vector" : "keyword";
-          sendJson(res, 200, { items: service.toApiList(rows), mode: used });
-        }).catch(() => {
-          sendJson(res, 200, { items: service.toApiList(service.search(query, { limit })), mode: "keyword" });
-        });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- user profile ---
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/profile",
-    handler(req, res) {
-      try {
-        if (req.method === "PUT" || req.method === "POST") {
-          if (!requireAuth(req, res, apiToken)) return;
-          return readBody(req).then((text) => {
-            const body = parseBody(text);
-            settings.setProfile(typeof body.profile === "string" ? body.profile : "");
-            sendJson(res, 200, { profile: settings.getProfile() });
-          });
-        }
-        sendJson(res, 200, { profile: settings.getProfile() });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- dream/sleep 模型连通性 + 模型发现（settings 面板）----------------------
-  // Connectivity probe for a dream/sleep model route: runs one minimal LLM call
-  // through the same harness path the consolidation uses and reports success /
-  // failure + latency + error. Reuses describeStreamFailure so the panel shows
-  // the same error wording the audit rows carry. Optional reasoningEffort
-  // mirrors what the consolidation would send — the panel can verify a model
-  // actually accepts the configured effort without tripping a full run.
-  async function testLlmConnectivity(llm, provider, model, reasoningEffort) {
-    const startedAt = Date.now();
-    const elapsed = () => Date.now() - startedAt;
-    const modelId = `${provider}:${model}`;
-    try {
-      let reply = "";
-      let error = "";
-      for await (const chunk of llm.stream({
-        provider,
-        model,
-        purpose: "dsh-mneme-connectivity-test",
-        // 给思考模型留出「思考完还有正文」的空间：16 会被 reasoning 全部吃掉，
-        // reply 恒为空，「真的答了 ok」对思考模型不成立。按实际用量计费，手动
-        // 按钮无放大风险。
-        maxTokens: 1024,
-        ...(reasoningEffort && reasoningEffort !== "none" ? { reasoningEffort } : {}),
-        // role 用 user 而非 system（#189 后续）：0.1.6-alpha.1 起适配器把 system
-        // 抽成顶层参数，system-only 对话会被官方 API 以
-        // 「messages: at least one message is required」整单拒绝——探测按钮 502。
-        // user 消息在任何适配器下都是一条合法对话，探测从此不怕 host 演进。
-        messages: [
-          {
-            role: "user",
-            content: [{ type: "text", text: "Reply with exactly one word: ok" }],
-            source: { kind: "plugin:dsh-mneme", plugin: "dsh-mneme" }
-          }
-        ]
-      })) {
-        if (chunk.type === "text-delta" && typeof chunk.text === "string") reply += chunk.text;
-        if (chunk.type === "finish" && (chunk.reason?.kind === "error" || chunk.reason?.kind === "aborted")) {
-          error = describeStreamFailure(chunk.reason);
-          break;
-        }
-      }
-      if (error) return { ok: false, durationMs: elapsed(), error, modelId };
-      return { ok: true, durationMs: elapsed(), modelId, reply: reply.trim().slice(0, 100) };
-    } catch (error) {
-      return { ok: false, durationMs: elapsed(), error: String(error?.message ?? error), modelId };
-    }
-  }
-
-  // List every provider + model the host has registered, so the panel can offer
-  // a dropdown over REAL models instead of free-text guesses. Read-only, no auth
-  // (same as list/search). Best-effort per provider: one whose model list can't
-  // be resolved yields an empty models array instead of failing the whole call.
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/llm-providers",
-    handler(req, res) {
-      try {
-        const llm = ctx.llm;
-        if (typeof llm?.listProviders !== "function" || typeof llm?.listModels !== "function") {
-          return sendJson(res, 501, { error: "llm-unavailable" });
-        }
-        let providers;
-        try { providers = llm.listProviders() ?? []; } catch { providers = []; }
-        return Promise.all(providers.map(async (p) => {
-          let models = [];
-          try { models = (await llm.listModels(p.id)) ?? []; } catch { models = []; }
-          return {
-            provider: p.id,
-            models: models.map((m) => ({ id: m.id, name: m.name ?? m.id }))
-          };
-        })).then((rows) => sendJson(res, 200, { providers: rows }));
-      } catch {
-        return sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // Connectivity test for a dream/sleep model route. Runs one tiny LLM call and
-  // reports ok / failure + latency + error. Auth-gated: it spends the user's
-  // API quota, so a stray local page must not be able to drain it via loopback.
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/test-model",
-    handler(req, res) {
-      try {
-        if (req.method !== "POST") return sendJson(res, 404, { error: "not-found" });
-        if (!requireAuth(req, res, apiToken)) return;
-        return readBody(req).then(async (text) => {
-          const body = parseBody(text);
-          let provider = typeof body.provider === "string" ? body.provider.trim() : "";
-          let model = typeof body.model === "string" ? body.model.trim() : "";
-          if ((!provider && model) || (provider && !model)) return sendJson(res, 400, { error: "missing-provider-or-model" });
-          if (!provider && !model) {
-            // 空 = 按巩固路由解析（dreamProvider/dreamModel > agent 默认模型），
-            // 让面板一键测"当前巩固模型"而无需手填。
-            const route = resolveRoute(ctx, config ?? {}, ctx.logger);
-            if (!route) return sendJson(res, 400, { error: "no-route" });
-            provider = route.provider;
-            model = route.model;
-          }
-          const llm = ctx.llm;
-          if (typeof llm?.stream !== "function") return sendJson(res, 501, { error: "llm-unavailable" });
-          const result = await testLlmConnectivity(llm, provider, model, body.reasoningEffort);
-          return sendJson(res, result.ok ? 200 : 502, result);
-        });
-      } catch {
-        return sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- delete one memory by id ---
-  // Mutation route: apiToken-gated like the profile/rules writes above. POST
-  // body is JSON { id }. store.remove deletes silently, so existence is checked
-  // up front to give clients a distinguishable 404 instead of a fake success.
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/delete",
-    handler(req, res) {
-      try {
-        if (req.method !== "POST") {
-          sendJson(res, 404, { error: "not-found" });
-          return;
-        }
-        if (!requireAuth(req, res, apiToken)) return;
-        return readBody(req).then((text) => {
-          const body = parseBody(text);
-          const id = typeof body.id === "string" ? body.id.trim() : "";
-          if (!id) {
-            sendJson(res, 400, { error: "missing-id" });
-            return;
-          }
-          if (!service.getById(id)) {
-            sendJson(res, 404, { error: "not-found" });
-            return;
-          }
-          service.remove(id);
-          sendJson(res, 200, { ok: true });
-        });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- 编辑/归档一条记忆（面板写路径，与 /delete 对称命名）-------------------
-  // POST body {id, title?, content?, importance?, tags?, archived?}。字段校验
-  // 只做类型/范围检查：字段出现（!== undefined）就必须合法，宁 400 不静默纠正
-  // ——静默丢字段会让面板误以为保存成功。写入走 service 的正规更新路径：
-  // updated_at 由 store 的单调时钟推进；content 被改写时旧版本按 human_override
-  // 入档（与镜像人工编辑回灌 mergeHumanEdits 的语义对齐，FIFO 上限 20）；
-  // archived 走 setArchived。两条路径都会触发镜像重渲染（afterSync）。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/update",
-    handler(req, res) {
-      try {
-        if (req.method !== "POST") {
-          sendJson(res, 404, { error: "not-found" });
-          return;
-        }
-        if (!requireAuth(req, res, apiToken)) return;
-        return readBody(req).then((text) => {
-          const body = parseBody(text);
-          const id = typeof body.id === "string" ? body.id.trim() : "";
-          if (!id) {
-            sendJson(res, 400, { error: "missing-id" });
-            return;
-          }
-          const patch = {};
-          if (body.title !== undefined) {
-            if (typeof body.title !== "string" || !body.title.trim()) {
-              sendJson(res, 400, { error: "invalid-title" });
-              return;
-            }
-            patch.title = body.title.trim();
-          }
-          if (body.content !== undefined) {
-            if (typeof body.content !== "string" || !body.content.trim()) {
-              sendJson(res, 400, { error: "invalid-content" });
-              return;
-            }
-            patch.content = body.content.trim();
-          }
-          if (body.importance !== undefined) {
-            if (!Number.isInteger(body.importance) || body.importance < 1 || body.importance > 5) {
-              sendJson(res, 400, { error: "invalid-importance" });
-              return;
-            }
-            patch.importance = body.importance;
-          }
-          if (body.tags !== undefined) {
-            if (!Array.isArray(body.tags) || !body.tags.every((t) => typeof t === "string")) {
-              sendJson(res, 400, { error: "invalid-tags" });
-              return;
-            }
-            patch.tags = body.tags;
-          }
-          // v0.8.1 底座（issue #170）：显式 scope 修正。null=放宽到全局，
-          // 字符串=收窄/改标；归一化（"global"/"*"/空白→NULL + explicit 章 +
-          // 审计行）在 service.updateMemory 内做，路由只做形状校验。
-          if (body.agent_scope !== undefined && body.agent_scope !== null && typeof body.agent_scope !== "string") {
-            sendJson(res, 400, { error: "invalid-agent-scope" });
-            return;
-          }
-          if (body.workspace_scope !== undefined && body.workspace_scope !== null && typeof body.workspace_scope !== "string") {
-            sendJson(res, 400, { error: "invalid-workspace-scope" });
-            return;
-          }
-          if (body.agent_scope !== undefined) patch.agent_scope = body.agent_scope;
-          if (body.workspace_scope !== undefined) patch.workspace_scope = body.workspace_scope;
-          if (body.archived !== undefined && typeof body.archived !== "boolean") {
-            sendJson(res, 400, { error: "invalid-archived" });
-            return;
-          }
-          if (Object.keys(patch).length === 0 && body.archived === undefined) {
-            sendJson(res, 400, { error: "no-fields" });
-            return;
-          }
-          const existing = service.getById(id);
-          if (!existing) {
-            sendJson(res, 404, { error: "not-found" });
-            return;
-          }
-          // 内容被改写时旧版本先入档（human_override），人工修正不静默销毁旧值。
-          if (patch.content !== undefined && patch.content !== existing.content) {
-            const history = Array.isArray(existing.content_history) ? existing.content_history : [];
-            patch.content_history = [
-              { content: existing.content ?? "", source: "human_override", updated_at: new Date().toISOString() },
-              ...history
-            ].slice(0, 20);
-          }
-          if (Object.keys(patch).length) service.update(id, patch, { actor: "panel" });
-          if (body.archived !== undefined) service.setArchived(id, body.archived);
-          sendJson(res, 200, { memory: service.getById(id) });
-        });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- rules ---
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/rules",
-    handler(req, res) {
-      try {
-        if (req.method === "PUT" || req.method === "POST") {
-          if (!requireAuth(req, res, apiToken)) return;
-          return readBody(req).then((text) => {
-            const body = parseBody(text);
-            settings.setRules(Array.isArray(body.rules) ? body.rules : []);
-            sendJson(res, 200, { rules: settings.getRules() });
-          });
-        }
-        sendJson(res, 200, { rules: settings.getRules() });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- vector search config ---
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/vector-config",
-    handler(req, res) {
-      try {
-        // Secret-bearing endpoint: fully protected when apiToken is set.
-        if (!requireAuth(req, res, apiToken)) return;
-        if (req.method === "PUT" || req.method === "POST") {
-          return readBody(req).then((text) => {
-            const body = parseBody(text);
-            // An empty apiKey, or one that already looks masked (round-trips
-            // through maskApiKey unchanged), means "keep the existing key".
-            // Only a fresh, unmasked key is treated as a real replacement.
-            const prev = settings.getVectorConfig();
-            const incoming = typeof body.apiKey === "string" ? body.apiKey.trim() : "";
-            const isMaskedOrEmpty = incoming === "" || maskApiKey(incoming) === incoming;
-            const key = isMaskedOrEmpty
-              ? (prev?.apiKey ?? "")
-              : incoming;
-            const cfg = settings.setVectorConfig({
-              enabled: body.enabled,
-              baseUrl: body.baseUrl,
-              apiKey: key,
-              model: body.model
-            });
-            sendJson(res, 200, { config: { ...cfg, apiKey: maskApiKey(cfg.apiKey) } });
-          });
-        }
-        const cfg = settings.getVectorConfig() ?? { enabled: false, baseUrl: "", apiKey: "", model: "" };
-        sendJson(res, 200, { config: { ...cfg, apiKey: maskApiKey(cfg.apiKey) } });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- vector re-index (backfill embeddings for rows missing them) ---
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/vector-reindex",
-    handler(req, res) {
-      try {
-        if (!requireAuth(req, res, apiToken)) return;
-        if (!embedder) {
-          sendJson(res, 200, { indexed: 0, skipped: 0, error: "vector-unavailable" });
-          return;
-        }
-        const url = new URL(req.url, "http://localhost");
-        const limit = Number(url.searchParams.get("limit") ?? 100);
-        // Unified re-index entry: works for both the legacy OpenAI embedder and
-        // the new local/ollama backends (which have no reindexMissing method).
-        const viaIndex = semantic?.vectorIndex && semantic?.vectorIndex.rebuildIndex;
-        const task = viaIndex
-          ? semantic.vectorIndex.rebuildIndex(embedder, { limit })
-          : embedder.reindexMissing ? embedder.reindexMissing(limit) : Promise.resolve({ indexed: 0, skipped: 0, error: "vector-unavailable" });
-        // Return the chain so awaiting callers (tests/health checks) observe the
-        // finished response rather than racing the async backfill.
-        return task.then((result) => {
-          sendJson(res, 200, result);
-        }).catch(() => {
-          sendJson(res, 200, { indexed: 0, skipped: 0, error: "vector-failed" });
-        });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- semantic pipeline status (model, index, reranker) ---
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/semantic",
-    handler(req, res) {
-      try {
-        const stats = semantic?.vectorIndex?.getStats?.() ?? null;
-        // #118: expose readiness so the status card can tell "initializing /
-        // unreachable" from "disabled". An embedder carrying no `ready` prop at
-        // all (test doubles, third-party adapters) is assumed usable.
-        // issue #135: 但「可用」和「配了」是两件事 —— legacy OpenAI embedder
-        // 此前恒报 ready，于是 baseUrl/apiKey/model 全空时状态卡依然是绿的
-        // （"绿色假阳性"）。下面把两者分开，并附上索引覆盖度。
-        const ready = embedder ? ("ready" in embedder ? embedder.ready === true : true) : null;
-        // local/ollama 的配置来自 settings 兜底，没有「未配置」态：它们的
-        // ready=false 表示「还在初始化」，不是「没配」。
-        const configured = embedder ? (typeof embedder.configured === "boolean" ? embedder.configured : true) : null;
-        const embedded = Number(stats?.embeddedCount ?? 0);
-        const total = Number(stats?.totalCount ?? 0);
-        sendJson(res, 200, {
-          ready,
-          configured,
-          // 机器可读的原因；null = 一切正常。面板按它选文案，不用猜布尔组合。
-          reason: !embedder ? "no-embedder"
-            : configured === false ? "not-configured"
-              : ready !== true ? "initializing"
-                : null,
-          // 空库时 ratio 必须是 null 而非 0 —— 0 会被读成「一条都没嵌上」。
-          coverage: { embedded, total, ratio: total > 0 ? embedded / total : null },
-          // 配了、库里有东西可嵌、却一条都没嵌上 = 真的降级（本机此前长期
-          // 处于 ready:true + embeddedCount:0，面板却显示正常）。
-          degraded: configured === true && total > 0 && embedded === 0,
-          // Explicit display name first: the legacy OpenAI-compatible embedder
-          // is an object literal, so constructor.name would be "Object".
-          embedProvider: embedder ? (embedder.name ?? embedder.constructor?.name ?? "unknown") : null,
-          modelHash: embedder?.modelHash ?? null,
-          dimension: embedder?.dimension ?? null,
-          reranker: semantic?.reranker ? "ready" : null,
-          index: stats,
-          // issue #131 / PR-A：自管运行时的状态。无论当前 provider 是不是 local
-          // 都报 —— 面板需要区分「已收编但没启用」和「要用本地却没收编」。
-          // 这里只做结构检查（不跑推理），失败降级成状态对象，不会让本端点 500。
-          // 必须经 publicRuntimeStatus 投影：本端点是免鉴权的，绝对路径与原始错误
-          // 文本不能出去（CWE-200）。完整诊断走本机 CLI 的 status。
-          localRuntime: publicRuntimeStatus(describeLocalRuntime({ runtimeDir: config?.runtimeDir ?? "" }))
-        });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- 一键取回自管运行时（issue #131 / PR-C：三档来源）-------------------------
-  // 写路径（requireAuth）：它会往 ~/.dsh/mneme/runtime/ 落一份依赖闭包。
-  //
-  // 第一档源固定取本 profile 的 node_modules（由插件自身路径推出），不接受客户端传源目录 ——
-  // 「从任意目录往运行时目录里搬东西」没有必要成为对外能力。
-  //
-  // 一律回 200：请求本身合法，成不成看 body 里的 ok/status/reason。让 HTTP 码只表达
-  // 「有没有鉴权/路径对不对」，结论交给 ok —— 面板要能显示「收编了但闭包缺件」这种
-  // 中间状态，用一个 4xx 概括会把信息压掉。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/runtime/provision",
-    handler(req, res) {
-      try {
-        if (req.method !== "POST") {
-          sendJson(res, 405, { error: "method-not-allowed" });
-          return;
-        }
-        if (!requireAuth(req, res, apiToken)) return;
-        return readBody(req).then(async (text) => {
-          const body = parseBody(text);
-          const runtimeDir = config?.runtimeDir ?? "";
-          // 三档来源依次尝试（收编本机 node_modules → 本地 .tgz 目录 → registry）。
-          // 所以这个端点会**联网**：它可能拉一份几十到上百 MB 的闭包。
-          const result = await provisionRuntime({
-            hostModulesDir: hostModulesDir(import.meta.url),
-            runtimeDir,
-            localTarballDir: config?.runtimeTarballDir ?? "",
-            mirror: config?.runtimeMirror ?? "",
-            overwrite: body?.overwrite === true
-          });
-          // 顺手带回最新状态，面板不必再打一次 /semantic（且它走的是免鉴权投影）。
-          sendJson(res, 200, {
-            ...result,
-            localRuntime: publicRuntimeStatus(describeLocalRuntime({ runtimeDir }))
-          });
-        });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- LLM audit trail (Bug8): paginated read + aggregate stats ---
-  // Read-only endpoints, so like list/search/semantic they stay open even when
-  // apiToken is set. The stats aggregate budget by source over the last N days.
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/semantic/llm-audit",
-    handler(req, res) {
-      try {
-        const url = new URL(req.url, "http://localhost");
-        const page = Math.max(1, Number(url.searchParams.get("page") ?? 1) || 1);
-        const pageSize = Math.min(200, Math.max(1, Number(url.searchParams.get("pageSize") ?? 50) || 50));
-        const source = url.searchParams.get("source") ?? undefined;
-        // 写入准入（#254）的 session_key 是内部计数键——session_id 不是对外的归属
-        // 标识（见 scope.js 文件头），而本端点按注释在 apiToken 之下也保持开放，所以
-        // 不回传它。按会话离线聚合直接读库。
-        const items = (service.listLlmAudits?.({ limit: pageSize, offset: (page - 1) * pageSize, source }) ?? [])
-          .map((row) => (row.session_key === undefined ? row : { ...row, session_key: undefined }));
-        const total = service.countLlmAudits?.({ source }) ?? items.length;
-        sendJson(res, 200, { items, total, page, pageSize });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/semantic/llm-audit/stats",
-    handler(req, res) {
-      try {
-        const url = new URL(req.url, "http://localhost");
-        const days = Math.max(1, Math.min(365, Number(url.searchParams.get("days") ?? 7) || 7));
-        const stats = service.getLlmAuditStats?.({ days }) ?? null;
-        sendJson(res, 200, stats ?? { error: "unavailable" });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- ego graph: 1-2 hop neighborhood of one entity (graph panel P1) ---
-  // Read-only like list/search/semantic, so it stays open when apiToken is set.
-  // BFS from the root entity over entity_relations (both directions; the
-  // idx_relations_from/to indexes keep a 2-hop walk in the tens of ms even
-  // for a few thousand nodes). `distance` on each node is the hop count from
-  // the root so the UI can shade the frontier. The API is graph-traversal
-  // only — nodes carry no attr payload; hover summaries come from
-  // /semantic/graph/entity-attrs.
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/semantic/graph/ego",
-    handler(req, res) {
-      try {
-        const url = new URL(req.url, "http://localhost");
-        const name = (url.searchParams.get("entity") ?? "").trim();
-        if (!name) {
-          sendJson(res, 400, { error: "missing-entity" });
-          return;
-        }
-        const root = service.findEntityByName?.(name);
-        if (!root) {
-          sendJson(res, 404, { error: "entity-not-found" });
-          return;
-        }
-        const depth = Math.max(1, Math.min(2, Number(url.searchParams.get("depth") ?? 1) || 1));
-        const limit = Math.max(1, Math.min(100, Number(url.searchParams.get("limit") ?? 40) || 40));
-
-        const nodes = new Map([[root.id, { ...root, distance: 0 }]]);
-        let frontier = [root.id];
-        for (let d = 1; d <= depth && nodes.size < limit; d++) {
-          const next = [];
-          for (const id of frontier) {
-            for (const rel of service.getRelations?.(id) ?? []) {
-              const other = rel.from_entity === id ? rel.to_entity : rel.from_entity;
-              if (nodes.has(other) || nodes.size >= limit) continue;
-              const entity = service.findEntityById?.(other);
-              if (!entity) continue;
-              nodes.set(other, { ...entity, distance: d });
-              next.push(other);
-            }
-          }
-          frontier = next;
-        }
-
-        // Collect every relation whose endpoints both survived the limit cut;
-        // each edge is visited twice (once per endpoint) so dedupe by id.
-        const edgeMap = new Map();
-        for (const id of nodes.keys()) {
-          for (const rel of service.getRelations?.(id) ?? []) {
-            if (nodes.has(rel.from_entity) && nodes.has(rel.to_entity)) {
-              edgeMap.set(rel.id, rel);
-            }
-          }
-        }
-
-        sendJson(res, 200, {
-          root: { id: root.id, name: root.name, type: root.type ?? null, mention_count: root.mention_count ?? 1 },
-          nodes: [...nodes.values()].map((n) => ({
-            id: n.id,
-            name: n.name,
-            type: n.type ?? null,
-            mention_count: n.mention_count ?? 1,
-            distance: n.distance,
-            // v0.7.0 实体热投影：实体热 = 关联记忆 heat 聚合（max），前端据此
-            // 缩放节点大小/明暗。heatEnabled=false 时 entityHeat 返回 null。
-            heat: service.entityHeat?.(n.id) ?? null
-          })),
-          edges: [...edgeMap.values()].map((e) => ({
-            id: e.id,
-            from: e.from_entity,
-            to: e.to_entity,
-            relation_type: e.relation_type,
-            memory_id: e.memory_id ?? null,
-            created_at: e.created_at
-          }))
-        });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- entity attrs: current valid attrs for one entity (graph hover panel) ---
-  // Read-only; mirrors getCurrentAttrs (valid_until IS NULL). Also used as the
-  // graph panel's fallback list when the ego graph is too sparse to draw.
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/semantic/graph/entity-attrs",
-    handler(req, res) {
-      try {
-        const url = new URL(req.url, "http://localhost");
-        const name = (url.searchParams.get("entity") ?? "").trim();
-        if (!name) {
-          sendJson(res, 400, { error: "missing-entity" });
-          return;
-        }
-        const entity = service.findEntityByName?.(name);
-        if (!entity) {
-          sendJson(res, 404, { error: "entity-not-found" });
-          return;
-        }
-        const attrs = service.getCurrentAttrs?.(entity.id) ?? [];
-        sendJson(res, 200, {
-          entity: { id: entity.id, name: entity.name, type: entity.type ?? null, mention_count: entity.mention_count ?? 1 },
-          attrs: Array.isArray(attrs)
-            ? attrs.map((a) => ({
-                key: a.attr_key,
-                value: a.attr_value,
-                confidence: a.confidence ?? null,
-                valid_from: a.valid_from ?? null
-              }))
-            : []
-        });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- entity directory (entity gene v0.3.0): read-only list for the UI ---
-  // Flat entity list (newest-seen first). Attributes and relations stay on
-  // their own per-entity endpoints, the directory only carries the columns
-  // the rail needs: name, type, mention_count, first/last seen.
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/entities",
-    handler(req, res) {
-      try {
-        const url = new URL(req.url, "http://localhost");
-        const limit = Math.min(1000, Math.max(1, Number(url.searchParams.get("limit") ?? 500) || 500));
-        const entities = service.listEntities?.({ limit }) ?? [];
-        sendJson(res, 200, { entities, total: entities.length });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- 一条记忆的关联实体（记忆详情侧栏）-------------------------------------
-  // 只读：entity_attrs.memory_id 反查实体（store 一条 JOIN 完成，按提及去重）。
-  // 与目录 /entities 不同，这里以记忆为锚点，回答"这条记忆提到了谁"。记忆不
-  // 存在或无关联一律返回空数组——详情侧栏不需要区分这两种情况。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/memories/entities",
-    handler(req, res) {
-      try {
-        const url = new URL(req.url, "http://localhost");
-        const memoryId = (url.searchParams.get("memoryId") ?? "").trim();
-        if (!memoryId) {
-          sendJson(res, 400, { error: "missing-memory-id" });
-          return;
-        }
-        sendJson(res, 200, { entities: service.entitiesForMemory?.(memoryId) ?? [] });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- 导出（面板备份/迁移）---------------------------------------------------
-  // 只读。json：全字段行（含 archived/forgotten，布尔化），updated_at DESC。
-  // markdown：一个文档 = 文档级 frontmatter + 按类型分节。frontmatter 只放文档
-  // 最前一份（分节不再各自带文件头，否则 9 个 `---` 块串在一个 .md 里，外部
-  // frontmatter 解析器只认第一段、其余降级成正文），它的 coverage: all 就是一句
-  // 看得见的声明：这份导出含 archived/forgotten（document 指针行除外——它两个
-  // 落点都不进，见 mirror.js 的 MIRROR_EXCLUDED_TYPES）。条目块与磁盘镜像同构（共用
-  // renderMirrorText 的条目渲染），所以导出文本能被 /import 原样吃回；「同构」
-  // 止于条目——磁盘镜像只写活跃集，导出的行集更宽，两者的覆盖声明因此不同
-  // （active-only vs all）。
-  // 全量一次性取回（store.all 按 updated_at DESC）——导出是一次性备份动作，
-  // 不需要流式。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/export",
-    handler(req, res) {
-      try {
-        const url = new URL(req.url, "http://localhost");
-        const format = url.searchParams.get("format") ?? "json";
-        if (format !== "json" && format !== "markdown") {
-          sendJson(res, 400, { error: "invalid-format" });
-          return;
-        }
-        const rows = service.all?.() ?? [];
-        const stamp = new Date().toISOString().slice(0, 10).replace(/-/g, "");
-        if (format === "json") {
-          sendJson(res, 200, {
-            exported_at: new Date().toISOString(),
-            version: PACKAGE_VERSION,
-            count: rows.length,
-            memories: rows
-          }, { "Content-Disposition": `attachment; filename="dsh-mneme-export-${stamp}.json"` });
-          return;
-        }
-        const byType = {};
-        for (const row of rows) {
-          // 没有 TYPE_FILE 条目的 type 不进 markdown 导出，只读 type（#296 的
-          // document 指针行）同样不进：导出/导入是 round-trip 通道，只读视图没有
-          // 可回填的正文。两个排除都在 mirror.js 那张表里（MIRROR_EXCLUDED_TYPES /
-          // MIRROR_READONLY_TYPES）。json 分支是全字段导出，不受影响。
-          if (TYPE_FILE[row.type] && !MIRROR_READONLY_TYPES.has(row.type)) (byType[row.type] ??= []).push(row);
-        }
-        const sections = [];
-        for (const type of Object.keys(TYPE_FILE)) {
-          if (byType[type]?.length) {
-            sections.push(renderMirrorText(type, byType[type], langOf(config), { fileHeader: false }));
-          }
-        }
-        const included = Object.values(byType).flat();
-        const docHeader = renderFileHeader("memory-export", included, "all");
-        sendAttachment(res, 200, "text/markdown; charset=utf-8", `dsh-mneme-export-${stamp}.md`, docHeader + sections.join("\n"));
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- 导入（Markdown 镜像回填）----------------------------------------------
-  // 写路径（requireAuth）。body {type, markdown}：type 是镜像 TYPE_FILE 键，
-  // markdown 是与镜像文件同构的文本。解析复用 readHumanEdits 的纯函数核心 parseHumanEdits（同一实现，
-  // 行为一致是硬约束），合并走 mergeHumanEdits（只吃 title/content；digest 命
-  // 中或无差异的条目在 service 侧自动跳过）。解析出 0 条不算错误。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/import",
-    handler(req, res) {
-      try {
-        if (req.method !== "POST") {
-          sendJson(res, 404, { error: "not-found" });
-          return;
-        }
-        if (!requireAuth(req, res, apiToken)) return;
-        return readBody(req).then((text) => {
-          const body = parseBody(text);
-          if (typeof body.type !== "string" || !Object.hasOwn(TYPE_FILE, body.type)) {
-            sendJson(res, 400, { error: "invalid-type" });
-            return;
-          }
-          // 只读 type 是合法 type，只是没有可回填的正文（#296）：单独的码，调用方
-          // 不会被误导去改 type 名重试。
-          if (MIRROR_READONLY_TYPES.has(body.type)) {
-            sendJson(res, 400, { error: "readonly-type" });
-            return;
-          }
-          if (typeof body.markdown !== "string" || !body.markdown.trim()) {
-            sendJson(res, 400, { error: "invalid-markdown" });
-            return;
-          }
-          const edits = parseHumanEdits(body.markdown);
-          service.mergeHumanEdits(body.type, edits);
-          sendJson(res, 200, { merged: edits.length, type: body.type });
-        });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- 巩固状态（dream 面板）--------------------------------------------------
-  // 只读：最近巩固运行（最多 5 条，created_at DESC）+ 未解决冲突队列。
-  // pendingMemoryIds 从未解决冲突行提取涉及的记忆 id，去重后上限 200，避免积
-  // 压很大时响应失控。listDreamRuns/listConflictPending 走 service 现成的审计
-  // 只读通道（bookkeeping 语义，不触发写钩子）。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/dream-status",
-    handler(req, res) {
-      try {
-        const runs = (service.listDreamRuns?.({ limit: 5 }) ?? []).map((r) => ({
-          created_at: r.created_at,
-          status: r.status,
-          provider: r.provider ?? null,
-          model: r.model ?? null,
-          error: r.error ?? null,
-          run_type: r.run_type ?? "auto",
-          // sleep 审计：heat/时间分层降级决策计数（工作动态可展示"降级 N 条"）。
-          // 数据来自 runSleep 写入的 decisions.demotion（{demoted,archived} 数组）。
-          demotion: r.decisions?.demotion
-            ? { demoted: (r.decisions.demotion.demoted ?? []).length, archived: (r.decisions.demotion.archived ?? []).length }
-            : null
-        }));
-        const pendingConflicts = service.countConflictPending?.() ?? 0;
-        const ids = new Set();
-        for (const row of service.listConflictPending?.({ limit: 200, includeResolved: false }) ?? []) {
-          if (ids.size >= 200) break;
-          if (row.memory_a) ids.add(row.memory_a);
-          if (ids.size >= 200) break;
-          if (row.memory_b) ids.add(row.memory_b);
-        }
-        sendJson(res, 200, {
-          lastRun: runs[0] ?? null,
-          runs,
-          pendingConflicts,
-          pendingMemoryIds: [...ids].slice(0, 200)
-        });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- 冲突集中处理（v0.8.0）：队列 + 人工确认 --------------------------------
-  // GET /conflicts：未解决冲突 + 双方当前内容（此前面板只有计数与徽章，
-  // resolveConflictPending 在存储层躺着没有任何调用方）。POST /conflicts/resolve：
-  // {id, winner: "a"|"b"|null, apply}——盖章审计，apply 时按 dream 非冻结
-  // conflict 路径同款处置（service.resolveConflictPending 内部复用
-  // applyDecisions：CAS + 事务 + 注记 + 幂等）。认证与其他写路由一致。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/conflicts",
-    handler(req, res) {
-      try {
-        if (req.method !== "GET") {
-          sendJson(res, 404, { error: "not-found" });
-          return;
-        }
-        if (!requireAuth(req, res, apiToken)) return;
-        const url = new URL(req.url, "http://localhost");
-        const limit = Math.min(200, Math.max(1, Number(url.searchParams.get("limit") ?? 50) || 50));
-        const offset = Math.max(0, Number(url.searchParams.get("offset") ?? 0) || 0);
-        const items = service.listConflictQueue({ limit, offset });
-        sendJson(res, 200, { items, total: service.countConflictPending?.() ?? items.length });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/conflicts/resolve",
-    handler(req, res) {
-      try {
-        if (req.method !== "POST") {
-          sendJson(res, 404, { error: "not-found" });
-          return;
-        }
-        if (!requireAuth(req, res, apiToken)) return;
-        return readBody(req).then((text) => {
-          let body = {};
-          try { body = JSON.parse(text || "{}"); } catch { /* 坏 JSON → 走缺参分支 */ }
-          const id = typeof body.id === "string" ? body.id.trim() : "";
-          if (!id) {
-            sendJson(res, 400, { error: "missing-id" });
-            return;
-          }
-          const winner = body.winner === "a" || body.winner === "b" ? body.winner : null;
-          const apply = body.apply !== false;
-          const result = service.resolveConflictPending(id, { winner, apply });
-          if (!result) {
-            sendJson(res, 404, { error: "not-found" });
-            return;
-          }
-          sendJson(res, 200, { ok: true, conflict: result });
-        }).catch(() => sendJson(res, 500, { error: "internal" }));
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- health: mirror sync state (F-NEW-03 / v0.3.6) ---
-  // Auth-gated; only returns a sanitized error code (never raw last_error which
-  // may leak paths/token-like strings/internal hosts). On state read failure it
-  // reports unknown/degraded (fail-closed) instead of a false dirty=false.
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/health",
-    handler(req, res) {
-      if (!requireAuth(req, res, apiToken)) return;
-      let state = null;
-      try {
-        state = service.getMirrorHealth?.() ?? null;
-      } catch {
-        // read failure is itself a health signal: do not report a false clean
-        sendJson(res, 200, { mirror: { dirty: null, status: "unknown", last_error: null, last_attempt: null, success_at: null } });
-        return;
-      }
-      if (!state) {
-        sendJson(res, 200, { mirror: { dirty: null, status: "unknown", last_error: null, last_attempt: null, success_at: null } });
-        return;
-      }
-      // Real read failure surfaces as dirty === null (peer blocker 5): report
-      // unknown explicitly instead of collapsing into a false "ok"/"degraded".
-      if (state.dirty === null) {
-        sendJson(res, 200, {
-          mirror: { dirty: null, status: "unknown", last_error: null, last_attempt: null, success_at: null }
-        });
-        return;
-      }
-      // Sanitized: boolean dirty + coarse status only; error string is mapped to
-      // a bounded code, never echoed verbatim.
-      let code = null;
-      if (state.last_error) {
-        const e = String(state.last_error);
-        code = /enospc|no space/i.test(e) ? "no-space" : /permission|eacces/i.test(e) ? "permission" : "sync-failed";
-      }
-      sendJson(res, 200, {
-        mirror: {
-          dirty: state.dirty === true,
-          status: state.dirty === true ? "degraded" : (code ? "degraded" : "ok"),
-          last_error: code,
-          last_attempt: state.last_attempt ?? null,
-          success_at: state.success_at ?? null
-        }
-      });
-    }
-  });
-
-  // --- panel mode (v0.7.12): light / standard -------------------------------
-  // Persists the Web panel's feature preset into the settings kv
-  // ("panel_mode"). PUT requires auth like every other settings write; the
-  // value is validated against the enum. index.js applies the light preset on
-  // the next boot (persisted mode wins over the bundle config).
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/mode",
-    handler(req, res) {
-      try {
-        if (req.method === "PUT" || req.method === "POST") {
-          if (!requireAuth(req, res, apiToken)) return;
-          return readBody(req).then((text) => {
-            const body = parseBody(text);
-            if (body.mode !== "light" && body.mode !== "standard") {
-              sendJson(res, 400, { error: "invalid-mode" });
-              return;
-            }
-            settings.setPanelMode(body.mode);
-            sendJson(res, 200, { mode: settings.getPanelMode() });
-          });
-        }
-        sendJson(res, 200, { mode: settings.getPanelMode() });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- external API settings (the standalone server's panel-facing config) ---
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/external-api",
-    handler(req, res) {
-      try {
-        if (req.method === "PUT" || req.method === "POST") {
-          if (!requireAuth(req, res, apiToken)) return;
-          return readBody(req).then((text) => {
-            const body = parseBody(text);
-            const patch = {};
-            if (body.enabled !== undefined) patch.enabled = body.enabled === true;
-            if (body.port !== undefined) {
-              const port = Number(body.port);
-              if (!Number.isInteger(port) || port < 1 || port > 65535) {
-                sendJson(res, 400, { error: "invalid-port" });
-                return;
-              }
-              patch.port = port;
-            }
-            if (body.host !== undefined) {
-              const host = String(body.host).trim();
-              if (!host || host.includes("://")) {
-                sendJson(res, 400, { error: "invalid-host" });
-                return;
-              }
-              patch.host = host;
-            }
-            sendJson(res, 200, { config: fullExternalConfig(settings.setExternalApi(patch)) });
-          });
-        }
-        // GET: always hand back a token — the standalone server generates its
-        // own on first enabled boot, but the panel displays it before that,
-        // so materialize and persist one here.
-        const kv = settings.getExternalApi() ?? {};
-        if (!kv.token) {
-          kv.token = randomBytes(24).toString("base64url");
-          settings.setExternalApi({ token: kv.token });
-        }
-        sendJson(res, 200, { config: fullExternalConfig(kv) });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- feature flags（功能开关：面板逐项开关后端能力）-------------------------
-  // 读路由保持开放（与 mode/list 一致），前端无需 token 即可渲染开关状态；PUT
-  // 与其他设置写一致走 requireAuth。空/非对象 patch 直接 400：它不携带任何
-  // 意图，静默成功只会掩盖前端 bug。校验失败（未知键/类型/越界）的 400 把键
-  // 名放进 error，前端能直接定位写坏的开关。生效节奏与 panel_mode 相同：
-  // 持久化后下次启动合并进 cfg，本次启动的运行时行为不变。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/features",
-    handler(req, res) {
-      try {
-        if (req.method === "PUT" || req.method === "POST") {
-          if (!requireAuth(req, res, apiToken)) return;
-          return readBody(req).then((text) => {
-            const body = parseBody(text);
-            if (typeof body !== "object" || body === null || Array.isArray(body) || Object.keys(body).length === 0) {
-              sendJson(res, 400, { error: "invalid-patch" });
-              return;
-            }
-            try {
-              settings.setFeatureFlags(body);
-            } catch (error) {
-              sendJson(res, 400, { error: error.message });
-              return;
-            }
-            sendJson(res, 200, featureSnapshot());
-          });
-        }
-        sendJson(res, 200, featureSnapshot());
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // 注入状态（#182，只读）：面板状态页据此决定是否渲染「极简模式下注入按宿主
-  // 设计关闭」提示卡。suppressed=true 需同时满足：autoInject 生效值开着（否则
-  // 关注入是用户自己的选择，不是黑盒）、会话头 preset 为 minimal、且 preset 确
-  // 实观测到了（null=未知会话/宿主不支持，面板静默不猜）。preset 值按字符串
-  // 精确匹配 "minimal"——host bundle 的 preset id（#175 逐字证据），不做模糊。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/inject-status",
-    handler(req, res) {
-      try {
-        if (req.method !== "GET") {
-          sendJson(res, 404, { error: "not-found" });
-          return;
-        }
-        const effective = featureSnapshot().effective;
-        const autoInject = effective.autoInject === true;
-        const agentPreset = lastAgentPreset;
-        const suppressed = autoInject && agentPreset === "minimal";
-        sendJson(res, 200, { autoInject, agentPreset, suppressed });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // 注入预览（#179，只读）：最近一帧 systemPrompt 组装的构成与体积——快照由
-  // inject.js 在真实渲染路径旁路缓存（同一份 candidates，不二次检索）。null =
-  // 尚未发生过渲染（新宿主/新会话）或 autoInject 关闭（注入器未注册），面板对
-  // 两者都以「暂无预览」呈现，不区分也不猜。
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/inject-preview",
-    handler(req, res) {
-      try {
-        if (req.method !== "GET") {
-          sendJson(res, 404, { error: "not-found" });
-          return;
-        }
-        sendJson(res, 200, { snapshot: getInjectionSnapshot() });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  // --- custom commands ---
-  register({
-    kind: "exact",
-    path: "/api/dsh-mneme/commands",
-    handler(req, res) {
-      try {
-        if (req.method === "POST") {
-          if (!requireAuth(req, res, apiToken)) return;
-          return readBody(req).then((text) => {
-            const body = parseBody(text);
-            try {
-              const command = commands.add({
-                name: body.name,
-                description: body.description,
-                instruction: body.instruction
-              });
-              sendJson(res, 200, { command });
-            } catch (error) {
-              sendJson(res, 400, { error: error.message });
-            }
-          });
-        }
-        if (req.method === "DELETE") {
-          if (!requireAuth(req, res, apiToken)) return;
-          const url = new URL(req.url, "http://localhost");
-          const id = url.searchParams.get("id");
-          const removed = id ? commands.remove(id) : false;
-          sendJson(res, 200, { removed });
-          return;
-        }
-        sendJson(res, 200, { commands: commands.list() });
-      } catch {
-        sendJson(res, 500, { error: "internal" });
-      }
-    }
-  });
-
-  return {
-    routes: 32,
-    dispose: () => {
-      for (const dispose of disposers) dispose();
-    }
-  };
-}
+import { URL } from "node:url";
+import { timingSafeEqual, randomBytes } from "node:crypto";
+import { FEATURE_FLAG_SPEC } from "./settings.js";
+import { TYPES } from "./store.js";
+import { TYPE_FILE, renderMirrorText, renderFileHeader, parseHumanEdits, MIRROR_READONLY_TYPES } from "./mirror.js";
+import { langOf } from "./lang.js";
+import { computeHeat } from "./heat.js";
+import { describeStreamFailure, resolveRoute } from "./dream.js";
+import { getInjectionSnapshot } from "./inject.js";
+import { describeLocalRuntime, publicRuntimeStatus } from "./runtime/loader.js";
+import { hostModulesDir, provisionRuntime } from "./runtime/provision.js";
+import { classify, fetchLatestVersion, PACKAGE_VERSION } from "./version-check.js";
+import { parseReclaimQuery, parseReclaimJson, readReclaimFields, reclaimReply, RECLAIM_UNAVAILABLE } from "./reclaim-http.js";
+
+// headers：少数端点（/export 附件下载）需要追加 Content-Disposition 等响应头。
+function sendJson(res, status, payload, headers = {}) {
+  res.writeHead(status, { "Content-Type": "application/json; charset=utf-8", ...headers });
+  res.end(JSON.stringify(payload));
+}
+
+// 附件下载响应（导出端点）：Content-Type 与文件名（含日期后缀）由调用方给定。
+function sendAttachment(res, status, contentType, filename, body) {
+  res.writeHead(status, {
+    "Content-Type": contentType,
+    "Content-Disposition": `attachment; filename="${filename}"`
+  });
+  res.end(body);
+}
+
+// 版本号统一从 version-check.js 取（同一份读 package.json 逻辑，读取失败
+// 同样降级 "unknown"）——/info 与 /version-check 两个路由共用，不再各写一份。
+
+// Defaults for the standalone external API — keep in step with the schema
+// defaults in config.js (externalApiPort / externalApiHost).
+const EXTERNAL_API_DEFAULTS = { enabled: false, port: 8790, host: "127.0.0.1" };
+
+/** Merge persisted external-api kv over the defaults for client display. */
+function fullExternalConfig(kv = {}) {
+  return {
+    enabled: kv.enabled === true,
+    port: Number.isInteger(kv.port) && kv.port > 0 ? kv.port : EXTERNAL_API_DEFAULTS.port,
+    host: kv.host || EXTERNAL_API_DEFAULTS.host,
+    token: kv.token || ""
+  };
+}
+
+/**
+ * Mask an API key for client display: keep a recognizable prefix and suffix,
+ * hide the middle. Empty keys stay empty; short keys are fully hidden.
+ * The mask only exists in the API layer — storage keeps the real key.
+ */
+function maskApiKey(key) {
+  if (!key) return "";
+  if (key.length <= 8) return "***";
+  return `${key.slice(0, 3)}***${key.slice(-4)}`;
+}
+
+/** True when the request carries the configured apiToken (or no token is set). */
+function isAuthorized(req, apiToken) {
+  if (!apiToken) return true;
+  const raw = req.headers?.authorization ?? req.headers?.["x-dsh-mneme-token"] ?? "";
+  const token = raw.startsWith("Bearer ") ? raw.slice(7).trim() : raw.trim();
+  if (token === "" || token.length !== apiToken.length) return false;
+  // Constant-time comparison: avoid leaking the token via timing when the API
+  // is exposed beyond loopback.
+  return timingSafeEqual(Buffer.from(token), Buffer.from(apiToken));
+}
+
+/**
+ * Reject a request with 401 when auth is enabled and the token is missing or
+ * wrong. Returns true when the request may proceed.
+ */
+function requireAuth(req, res, apiToken) {
+  if (isAuthorized(req, apiToken)) return true;
+  sendJson(res, 401, { error: "unauthorized" });
+  return false;
+}
+
+/** Collect the request body as text (tolerant of empty/invalid bodies). */
+function readBody(req) {
+  return new Promise((resolve) => {
+    let body = "";
+    req.on("data", (chunk) => { body += chunk; });
+    req.on("end", () => resolve(body));
+    req.on("error", () => resolve(""));
+  });
+}
+
+function parseBody(text) {
+  try {
+    return JSON.parse(text || "{}");
+  } catch {
+    return {};
+  }
+}
+
+// maintenance 是第 9 个位置参数，默认 null：既有调用点（测试里的 14 处）全是位置传参、
+// 不传它，必须继续工作——不传时 /maintenance/reclaim 回 503 maintenance-unavailable，
+// 而不是把「没接线」伪装成崩溃（#386 第 4 条）。
+export function createApi(ctx, service, settings, commands, embedder, semantic = null, apiToken = "", config = null, maintenance = null) {
+  const disposers = [];
+
+  // #182 注入状态探测：面板状态页的「极简模式下注入按宿主设计关闭」提示需要
+  // 知道当前会话的 agentPreset。注入回调在 minimal 下被宿主整体压制（#175 的
+  // 定论），不能作为检测源；session/event 是 preset 无关钩子（蒸馏同款），任
+  // 何事件都带完整 session 对象。只记最近一次会话头——这是面板展示用的「当前
+  // 会话」最佳近似，不是逐会话账本。读法与 scope.js 同源（session.header 冻结
+  // 的 SessionHeader，降级 requestHeader().config；agentPreset 取不到留 null，
+  // 面板对 null 静默）。任何失败静默：纯可观测性，不新增故障面。
+  let lastAgentPreset = null;
+  try {
+    const unsubscribe = ctx.on?.("session/event", (session) => {
+      try {
+        const header = session?.header ?? session?.requestHeader?.()?.config ?? null;
+        const preset = header?.agentPreset;
+        lastAgentPreset = typeof preset === "string" && preset.trim() ? preset.trim() : null;
+      } catch { /* 观测失败不影响会话流 */ }
+    });
+    if (typeof unsubscribe === "function") disposers.push(unsubscribe);
+  } catch { /* 宿主无 session/event 时面板提示退化为「未知」，不报错 */ }
+
+  // webServer 在 index.js 的 inject 声明中（cordis 等宿主服务就绪后 apply），
+  // 这里做防御性读取：cordis ctx 的 Proxy 不允许直接访问未 inject 的属性
+  // （会抛 "cannot get property without inject"），用 ctx.reflect.get 免
+  // inject 读取（未提供返回 undefined）；对象字面量 mock ctx（测试）没有
+  // reflect，退回直接属性访问。createApi 只在有 webServer 时被调用（index.js
+  // 守卫 + 测试 mock），故下方 register 用之非空。
+  const webServer = typeof ctx.reflect?.get === "function" ? ctx.reflect.get("webServer") : ctx.webServer;
+
+  // feature flags 快照（GET/PUT 共用）：overrides 是持久化的用户显式覆盖；
+  // effective 是启动配置在白名单键上被 overrides 覆盖后的最终值。config 缺席
+  // （旧调用方直连、未传 cfg）时只报被覆盖的键，不把不存在的默认值编造给前端。
+  const flagKeys = [
+    ...FEATURE_FLAG_SPEC.booleans,
+    ...Object.keys(FEATURE_FLAG_SPEC.ints),
+    ...Object.keys(FEATURE_FLAG_SPEC.numbers),
+    ...FEATURE_FLAG_SPEC.strings,
+    ...FEATURE_FLAG_SPEC.urls,
+    ...Object.keys(FEATURE_FLAG_SPEC.enums)
+  ];
+  // 嵌套键在 kv 里按点号平铺（"memoryQualityFilter.enabled"），运行时 cfg 里
+  // 是嵌套对象，effective 从对象子字段取值；其余键照旧从 cfg 顶层取。
+  // 这张表必须与 settings.js 白名单里的点号键一一对应：#254 加键时漏了这里，
+  // effective 会静默少两个键（configFlagValue 取不到 undefined 就不写进去），
+  // 面板上看不出来、PUT 也存不进去——test/api.test.js 的计数锁就是钉这件事的。
+  const NESTED_FLAG_PATHS = {
+    "memoryQualityFilter.enabled": ["memoryQualityFilter", "enabled"],
+    "llmAudit.enabled": ["llmAudit", "enabled"],
+    "writeAdmission.enabled": ["writeAdmission", "enabled"],
+    "writeAdmission.enforce": ["writeAdmission", "enforce"],
+    // Issue #380：注入前判定。漏登记的后果是 effective 静默缺键、面板盲——
+    // 上面的 round-trip 测试专门锁这张表与白名单成对。
+    "preInjectGate.enabled": ["preInjectGate", "enabled"],
+    "preInjectGate.enforce": ["preInjectGate", "enforce"],
+    // Issue #399：蒸馏原文留存 + 溯源。两个键都是嵌套子字段（enabled 开关 +
+    // retentionDays 整数），漏登记的后果同上——effective 静默缺键。
+    "distillRecords.enabled": ["distillRecords", "enabled"],
+    "distillRecords.retentionDays": ["distillRecords", "retentionDays"]
+    // sensitiveScanEnabled（#164 A2）是顶层扁平键，不需要进这张表——走 configFlagValue
+    // 的默认分支即可。
+  };
+  function configFlagValue(key) {
+    const path = NESTED_FLAG_PATHS[key];
+    if (path) return config?.[path[0]]?.[path[1]];
+    return config?.[key];
+  }
+  function featureSnapshot() {
+    const overrides = settings.getFeatureFlags();
+    const effective = {};
+    for (const key of flagKeys) {
+      if (overrides[key] !== undefined) effective[key] = overrides[key];
+      else {
+        const value = configFlagValue(key);
+        if (value !== undefined) effective[key] = value;
+      }
+    }
+    return { overrides, effective };
+  }
+
+  // Ensure the service has an embedder when the API layer was handed one
+  // (tests wire the embedder through the API instead of index.js). Without
+  // this, /api/dsh-mneme/search would silently degrade to keyword-only.
+  if (embedder && typeof service.setEmbedder === "function") {
+    service.setEmbedder(embedder);
+  }
+
+  const register = (route) => {
+    disposers.push(webServer.register(route));
+  };
+
+  // /api/dsh-mneme prefix fallback → 404 JSON for unknown sub-paths
+  register({
+    kind: "prefix",
+    path: "/api/dsh-mneme",
+    handler(req, res) {
+      sendJson(res, 404, { error: "not-found" });
+    }
+  });
+
+  // 反馈入口预填用的插件版本（面板「帮助与反馈」卡片拉取）。读取失败（打包
+  // 环境）返回 "unknown"，链接照常可用，纯展示信息不拦截。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/info",
+    handler(req, res) {
+      sendJson(res, 200, { version: PACKAGE_VERSION });
+    }
+  });
+
+  // 版本自检（#174 后续）：运行版本 vs npm registry latest。只读，宿主围栏
+  // 统一鉴权（与 dream-status 同款，面板 apiFetch 直接可用）。registry 查询
+  // 全失败时 latest=null、status=unknown——前端对 unknown 完全静默，版本提示
+  // 是锦上添花，绝不因它新增故障面。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/version-check",
+    handler(req, res) {
+      if (req.method !== "GET") {
+        sendJson(res, 404, { error: "not-found" });
+        return;
+      }
+      fetchLatestVersion()
+        .then((latest) => sendJson(res, 200, {
+          version: PACKAGE_VERSION,
+          latest,
+          status: classify(PACKAGE_VERSION, latest),
+          checkedAt: new Date().toISOString()
+        }))
+        .catch(() => sendJson(res, 200, { version: PACKAGE_VERSION, latest: null, status: "unknown" }));
+    }
+  });
+
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/list",
+    handler(req, res) {
+      try {
+        const url = new URL(req.url, "http://localhost");
+        const type = url.searchParams.get("type") ?? undefined;
+        const limit = Number(url.searchParams.get("limit") ?? 50);
+        const offset = Number(url.searchParams.get("offset") ?? 0);
+        // order=chrono: pure newest-first for paged browsing; default keeps
+        // the importance-ranked order other callers rely on.
+        const order = url.searchParams.get("order") ?? undefined;
+        // minImportance: numeric lower bound on importance (absent/NaN → no
+        // floor); source: exact match on the source column (empty → no filter).
+        const minRaw = url.searchParams.get("minImportance");
+        const minImportance = minRaw !== null && minRaw !== "" && !Number.isNaN(Number(minRaw))
+          ? Number(minRaw)
+          : undefined;
+        const source = url.searchParams.get("source") || undefined;
+        // updatedFrom/updatedTo：updated_at 闭区间过滤（ISO 日期或完整时间戳）。
+        // 边界归一化在 store 的 updatedAtBounds（list/count 共用同一纯函数）
+        // 完成，非法值在那里被忽略——这里原样透传即可。
+        const updatedFrom = url.searchParams.get("updatedFrom") ?? undefined;
+        const updatedTo = url.searchParams.get("updatedTo") ?? undefined;
+        // archived=only：只看归档（状态页的归档列表用）。归档行不进默认列表，
+        // 所以这是独立的视图开关，而不是 includeArchived 的混看模式。
+        const onlyArchived = url.searchParams.get("archived") === "only";
+        // deposited=only：只看 autoDream 巩固过的记忆（receipt_chain 的
+        // merge/update verdict ∪ source=dream 直写）——状态页「查看全部」
+        // 与记忆库的「沉淀」筛选 chip 共用这个视图。
+        const depositedOnly = url.searchParams.get("deposited") === "only";
+        const rows = service.list({ type, limit, offset, order, minImportance, source, updatedFrom, updatedTo, onlyArchived, depositedOnly });
+        // 面板行在 wire DTO 之上补 archived/quality_score——模型工具的输出
+        // schema 严格复用 toApiList，扩展只发生在 HTTP 层。
+        // heat 投影（阶段二前端数据源）：仅 heatEnabled=true 时下发逐条热度
+        // （heat.js 纯函数，λ=0 免疫类型恒 1.0）；字段缺省时前端徽章自动隐藏。
+        const heatOn = config?.heatEnabled === true;
+        const items = service.toApiList(rows).map((m, i) => ({
+          ...m,
+          archived: rows[i].archived === true || rows[i].archived === 1,
+          quality_score: rows[i].quality_score ?? null,
+          ...(heatOn ? { heat: computeHeat(rows[i], Date.now(), config ?? {}) } : {})
+        }));
+        // Total honors the same filters as the rows, or the pager's
+        // has-more math breaks whenever minImportance/source/updated-at
+        // bounds are active.
+        sendJson(res, 200, { items, total: service.count(type, { minImportance, source, updatedFrom, updatedTo, onlyArchived, depositedOnly }) });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // 记忆复用统计（#217，只读聚合）：Top-N 召回 + 僵尸率 + 覆盖度标注。窗口
+  // 参数钳制 1-365 天（缺省 30），聚合口径见 service.recallStats；注入命中
+  // 率待注入留痕口径拍板后接入。失败走既有 500 JSON 惯例，前端整卡不渲染。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/recall-stats",
+    handler(req, res) {
+      if (req.method !== "GET") {
+        sendJson(res, 404, { error: "not-found" });
+        return;
+      }
+      try {
+        const url = new URL(req.url, "http://localhost");
+        const raw = Number(url.searchParams.get("window"));
+        const windowDays = Number.isInteger(raw) && raw >= 1 && raw <= 365 ? raw : 30;
+        sendJson(res, 200, service.recallStats({ windowDays }));
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/search",
+    handler(req, res) {
+      try {
+        const url = new URL(req.url, "http://localhost");
+        const q = url.searchParams.get("q") ?? "";
+        const limit = Number(url.searchParams.get("topK") ?? url.searchParams.get("limit") ?? 20);
+        // mode selects the recall strategy (defaults to auto):
+        //   auto    (default) keyword first, vector fills remaining slots
+        //   hybrid  vector first, keyword fills remaining slots; scores of
+        //           memories hit by both sides are weight-blended
+        //   vector  vector only, falls back to keyword when the vector path
+        //           is unavailable (no embedder or a throwing one)
+        //   keyword literal text only; never queries the embedder
+        // rerank=false disables the cross-encoder reorder for this request;
+        // the response `mode` field reports which path actually produced rows.
+        const mode = url.searchParams.get("mode") ?? "auto";
+        const rerank = url.searchParams.get("rerank") !== "false";
+        const query = q.trim();
+        if (!query) {
+          sendJson(res, 200, { items: [], mode: "keyword" });
+          return;
+        }
+        // Route through the unified semantic pipeline; any vector/rerank
+        // failure degrades to keyword results inside searchMemories. The
+        // returned promise lets the test double await the async search.
+        return Promise.resolve(
+          service.searchMemories(query, { mode, topK: limit, useRerank: rerank })
+        ).then((rows) => {
+          // mode reflects what actually happened: rows marked `vector` came
+          // through the semantic path, everything else is keyword fallback.
+          const used = rows.some((m) => m.vector === true) ? "vector" : "keyword";
+          sendJson(res, 200, { items: service.toApiList(rows), mode: used });
+        }).catch(() => {
+          sendJson(res, 200, { items: service.toApiList(service.search(query, { limit })), mode: "keyword" });
+        });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- user profile ---
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/profile",
+    handler(req, res) {
+      try {
+        if (req.method === "PUT" || req.method === "POST") {
+          if (!requireAuth(req, res, apiToken)) return;
+          return readBody(req).then((text) => {
+            const body = parseBody(text);
+            settings.setProfile(typeof body.profile === "string" ? body.profile : "");
+            sendJson(res, 200, { profile: settings.getProfile() });
+          });
+        }
+        sendJson(res, 200, { profile: settings.getProfile() });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- dream/sleep 模型连通性 + 模型发现（settings 面板）----------------------
+  // Connectivity probe for a dream/sleep model route: runs one minimal LLM call
+  // through the same harness path the consolidation uses and reports success /
+  // failure + latency + error. Reuses describeStreamFailure so the panel shows
+  // the same error wording the audit rows carry. Optional reasoningEffort
+  // mirrors what the consolidation would send — the panel can verify a model
+  // actually accepts the configured effort without tripping a full run.
+  async function testLlmConnectivity(llm, provider, model, reasoningEffort) {
+    const startedAt = Date.now();
+    const elapsed = () => Date.now() - startedAt;
+    const modelId = `${provider}:${model}`;
+    try {
+      let reply = "";
+      let error = "";
+      for await (const chunk of llm.stream({
+        provider,
+        model,
+        purpose: "dsh-mneme-connectivity-test",
+        // 给思考模型留出「思考完还有正文」的空间：16 会被 reasoning 全部吃掉，
+        // reply 恒为空，「真的答了 ok」对思考模型不成立。按实际用量计费，手动
+        // 按钮无放大风险。
+        maxTokens: 1024,
+        ...(reasoningEffort && reasoningEffort !== "none" ? { reasoningEffort } : {}),
+        // role 用 user 而非 system（#189 后续）：0.1.6-alpha.1 起适配器把 system
+        // 抽成顶层参数，system-only 对话会被官方 API 以
+        // 「messages: at least one message is required」整单拒绝——探测按钮 502。
+        // user 消息在任何适配器下都是一条合法对话，探测从此不怕 host 演进。
+        messages: [
+          {
+            role: "user",
+            content: [{ type: "text", text: "Reply with exactly one word: ok" }],
+            source: { kind: "plugin:dsh-mneme", plugin: "dsh-mneme" }
+          }
+        ]
+      })) {
+        if (chunk.type === "text-delta" && typeof chunk.text === "string") reply += chunk.text;
+        if (chunk.type === "finish" && (chunk.reason?.kind === "error" || chunk.reason?.kind === "aborted")) {
+          error = describeStreamFailure(chunk.reason);
+          break;
+        }
+      }
+      if (error) return { ok: false, durationMs: elapsed(), error, modelId };
+      return { ok: true, durationMs: elapsed(), modelId, reply: reply.trim().slice(0, 100) };
+    } catch (error) {
+      return { ok: false, durationMs: elapsed(), error: String(error?.message ?? error), modelId };
+    }
+  }
+
+  // List every provider + model the host has registered, so the panel can offer
+  // a dropdown over REAL models instead of free-text guesses. Read-only, no auth
+  // (same as list/search). Best-effort per provider: one whose model list can't
+  // be resolved yields an empty models array instead of failing the whole call.
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/llm-providers",
+    handler(req, res) {
+      try {
+        const llm = ctx.llm;
+        if (typeof llm?.listProviders !== "function" || typeof llm?.listModels !== "function") {
+          return sendJson(res, 501, { error: "llm-unavailable" });
+        }
+        let providers;
+        try { providers = llm.listProviders() ?? []; } catch { providers = []; }
+        return Promise.all(providers.map(async (p) => {
+          let models = [];
+          try { models = (await llm.listModels(p.id)) ?? []; } catch { models = []; }
+          return {
+            provider: p.id,
+            models: models.map((m) => ({ id: m.id, name: m.name ?? m.id }))
+          };
+        })).then((rows) => sendJson(res, 200, { providers: rows }));
+      } catch {
+        return sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // Connectivity test for a dream/sleep model route. Runs one tiny LLM call and
+  // reports ok / failure + latency + error. Auth-gated: it spends the user's
+  // API quota, so a stray local page must not be able to drain it via loopback.
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/test-model",
+    handler(req, res) {
+      try {
+        if (req.method !== "POST") return sendJson(res, 404, { error: "not-found" });
+        if (!requireAuth(req, res, apiToken)) return;
+        return readBody(req).then(async (text) => {
+          const body = parseBody(text);
+          let provider = typeof body.provider === "string" ? body.provider.trim() : "";
+          let model = typeof body.model === "string" ? body.model.trim() : "";
+          if ((!provider && model) || (provider && !model)) return sendJson(res, 400, { error: "missing-provider-or-model" });
+          if (!provider && !model) {
+            // 空 = 按巩固路由解析（dreamProvider/dreamModel > agent 默认模型），
+            // 让面板一键测"当前巩固模型"而无需手填。
+            const route = resolveRoute(ctx, config ?? {}, ctx.logger);
+            if (!route) return sendJson(res, 400, { error: "no-route" });
+            provider = route.provider;
+            model = route.model;
+          }
+          const llm = ctx.llm;
+          if (typeof llm?.stream !== "function") return sendJson(res, 501, { error: "llm-unavailable" });
+          const result = await testLlmConnectivity(llm, provider, model, body.reasoningEffort);
+          return sendJson(res, result.ok ? 200 : 502, result);
+        });
+      } catch {
+        return sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- delete one memory by id ---
+  // Mutation route: apiToken-gated like the profile/rules writes above. POST
+  // body is JSON { id }. store.remove deletes silently, so existence is checked
+  // up front to give clients a distinguishable 404 instead of a fake success.
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/delete",
+    handler(req, res) {
+      try {
+        if (req.method !== "POST") {
+          sendJson(res, 404, { error: "not-found" });
+          return;
+        }
+        if (!requireAuth(req, res, apiToken)) return;
+        return readBody(req).then((text) => {
+          const body = parseBody(text);
+          const id = typeof body.id === "string" ? body.id.trim() : "";
+          if (!id) {
+            sendJson(res, 400, { error: "missing-id" });
+            return;
+          }
+          if (!service.getById(id)) {
+            sendJson(res, 404, { error: "not-found" });
+            return;
+          }
+          service.remove(id);
+          sendJson(res, 200, { ok: true });
+        });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- 编辑/归档一条记忆（面板写路径，与 /delete 对称命名）-------------------
+  // POST body {id, title?, content?, importance?, tags?, archived?}。字段校验
+  // 只做类型/范围检查：字段出现（!== undefined）就必须合法，宁 400 不静默纠正
+  // ——静默丢字段会让面板误以为保存成功。写入走 service 的正规更新路径：
+  // updated_at 由 store 的单调时钟推进；content 被改写时旧版本按 human_override
+  // 入档（与镜像人工编辑回灌 mergeHumanEdits 的语义对齐，FIFO 上限 20）；
+  // archived 走 setArchived。两条路径都会触发镜像重渲染（afterSync）。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/update",
+    handler(req, res) {
+      try {
+        if (req.method !== "POST") {
+          sendJson(res, 404, { error: "not-found" });
+          return;
+        }
+        if (!requireAuth(req, res, apiToken)) return;
+        return readBody(req).then((text) => {
+          const body = parseBody(text);
+          const id = typeof body.id === "string" ? body.id.trim() : "";
+          if (!id) {
+            sendJson(res, 400, { error: "missing-id" });
+            return;
+          }
+          const patch = {};
+          if (body.title !== undefined) {
+            if (typeof body.title !== "string" || !body.title.trim()) {
+              sendJson(res, 400, { error: "invalid-title" });
+              return;
+            }
+            patch.title = body.title.trim();
+          }
+          if (body.content !== undefined) {
+            if (typeof body.content !== "string" || !body.content.trim()) {
+              sendJson(res, 400, { error: "invalid-content" });
+              return;
+            }
+            patch.content = body.content.trim();
+          }
+          if (body.importance !== undefined) {
+            if (!Number.isInteger(body.importance) || body.importance < 1 || body.importance > 5) {
+              sendJson(res, 400, { error: "invalid-importance" });
+              return;
+            }
+            patch.importance = body.importance;
+          }
+          if (body.tags !== undefined) {
+            if (!Array.isArray(body.tags) || !body.tags.every((t) => typeof t === "string")) {
+              sendJson(res, 400, { error: "invalid-tags" });
+              return;
+            }
+            patch.tags = body.tags;
+          }
+          // v0.8.1 底座（issue #170）：显式 scope 修正。null=放宽到全局，
+          // 字符串=收窄/改标；归一化（"global"/"*"/空白→NULL + explicit 章 +
+          // 审计行）在 service.updateMemory 内做，路由只做形状校验。
+          if (body.agent_scope !== undefined && body.agent_scope !== null && typeof body.agent_scope !== "string") {
+            sendJson(res, 400, { error: "invalid-agent-scope" });
+            return;
+          }
+          if (body.workspace_scope !== undefined && body.workspace_scope !== null && typeof body.workspace_scope !== "string") {
+            sendJson(res, 400, { error: "invalid-workspace-scope" });
+            return;
+          }
+          if (body.agent_scope !== undefined) patch.agent_scope = body.agent_scope;
+          if (body.workspace_scope !== undefined) patch.workspace_scope = body.workspace_scope;
+          if (body.archived !== undefined && typeof body.archived !== "boolean") {
+            sendJson(res, 400, { error: "invalid-archived" });
+            return;
+          }
+          if (Object.keys(patch).length === 0 && body.archived === undefined) {
+            sendJson(res, 400, { error: "no-fields" });
+            return;
+          }
+          const existing = service.getById(id);
+          if (!existing) {
+            sendJson(res, 404, { error: "not-found" });
+            return;
+          }
+          // 内容被改写时旧版本先入档（human_override），人工修正不静默销毁旧值。
+          if (patch.content !== undefined && patch.content !== existing.content) {
+            const history = Array.isArray(existing.content_history) ? existing.content_history : [];
+            patch.content_history = [
+              { content: existing.content ?? "", source: "human_override", updated_at: new Date().toISOString() },
+              ...history
+            ].slice(0, 20);
+          }
+          if (Object.keys(patch).length) service.update(id, patch, { actor: "panel" });
+          if (body.archived !== undefined) service.setArchived(id, body.archived);
+          sendJson(res, 200, { memory: service.getById(id) });
+        });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- 新建一条记忆（面板写路径，与 /update、/delete 对称命名）---------------
+  // #386 第 5 条：CLI `add` 与 standalone `POST /memories` 都能建行，面板此前只
+  // 能看/改/删/导入。字段与 memory_save 工具对齐（type/title/content/importance/
+  // tags），走同一条 saveWithDedupe 路径——去重键仍是 (type, title, scope,
+  // sensitivity)，所以「面板建」与「工具建」同标题会并入同一行，不会长出两份。
+  //
+  // 三处刻意的口径差异（都写在代码里，免得下一个人以为是抄漏）：
+  //  1. content 要求 trim 后非空，与 /update 一致而非 standalone 的宽松版：
+  //     面板是人工输入，空正文只可能是误操作；standalone 面对的是程序化调用，
+  //     放宽是为了兼容「先建壳再补内容」的既有客户端。
+  //  2. document 直接 400：行铸造口唯一（registerDocument，#230），saveWithDedupe
+  //     会 throw，这里提前给干净错误码而不是让它落成 500。
+  //  3. 不接受 scope/sensitivity 字段——面板的 scope 编辑在详情抽屉里（带放宽
+  //     确认的安全阀，见 lib/client.js 的 MemoryDrawer），新建时就盖章 explicit
+  //     会绕过那道确认。传了也不报错，只是不生效（与 /update 的「字段出现即校验」
+  //     不同，因为这里根本没有该字段的语义）。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/create",
+    handler(req, res) {
+      try {
+        if (req.method !== "POST") {
+          sendJson(res, 404, { error: "not-found" });
+          return;
+        }
+        if (!requireAuth(req, res, apiToken)) return;
+        return readBody(req).then((text) => {
+          const body = parseBody(text);
+          if (!TYPES.has(body.type)) {
+            sendJson(res, 400, { error: "invalid-type" });
+            return;
+          }
+          if (body.type === "document") {
+            sendJson(res, 400, { error: "document-requires-register" });
+            return;
+          }
+          if (typeof body.title !== "string" || !body.title.trim()) {
+            sendJson(res, 400, { error: "invalid-title" });
+            return;
+          }
+          if (typeof body.content !== "string" || !body.content.trim()) {
+            sendJson(res, 400, { error: "invalid-content" });
+            return;
+          }
+          if (body.importance !== undefined
+            && (!Number.isInteger(body.importance) || body.importance < 1 || body.importance > 5)) {
+            sendJson(res, 400, { error: "invalid-importance" });
+            return;
+          }
+          if (body.tags !== undefined && (!Array.isArray(body.tags) || !body.tags.every((tag) => typeof tag === "string"))) {
+            sendJson(res, 400, { error: "invalid-tags" });
+            return;
+          }
+          try {
+            const result = service.saveWithDedupe({
+              type: body.type,
+              title: body.title.trim(),
+              content: body.content.trim(),
+              importance: body.importance,
+              tags: body.tags ?? [],
+              // 面板不是工具也不是外部程序，来源要能在详情抽屉里一眼认出是谁写的。
+              source: "panel"
+            });
+            // #254 写入准入在无会话身份时放行（write-admission.js 首行早退），面板
+            // 路由恰好不带 _sessionKey，所以今天到不了这个分支。仍然显式写出来：
+            // 将来若把会话身份接进面板，落到 toApiList([null]) 上会变成 500，把
+            // 「被拒」伪装成「服务器出错」——两者的处置完全不同。
+            if (result.action === "denied") {
+              sendJson(res, 422, { error: "write-rejected", reason: result.reason });
+              return;
+            }
+            // action 随行透出（created/merged）：面板据此提示「已并入同标题的既有条目」
+            // 还是「新建成功」——两者对用户是不同的事实，不能都报「已保存」。
+            sendJson(res, result.action === "created" ? 201 : 200, {
+              ...service.toApiList([result.memory])[0],
+              action: result.action
+            });
+          } catch (error) {
+            ctx.logger?.warn?.(`[dsh-mneme] panel create failed: ${String(error)}`);
+            sendJson(res, 500, { error: "internal" });
+          }
+        });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- maintenance/reclaim（#386 第 4 条：面板有了回收入口）------------------
+  // 与 standalone 数据面的 POST /maintenance/reclaim 是同一能力的两张皮：口径（错误码
+  // 字符串、默认值、报告形状）只写在 src/reclaim-http.js 里，这里与数据面都调它。
+  // 差别只有鉴权与路径——两张皮的其余部分必须没有第二份实现（#411 的教训：以前是整段
+  // 复制，于是 #410 修好这边之后那边还停在旧形态）。
+  //
+  // 为什么 GET 也能跑：dry-run 是只读的（src/maintenance.js 里 dry-run 与执行同一条
+  // 代码路径，只是不落 UPDATE/VACUUM），面板 mount 时拿它渲染「现在能回收多少」。
+  //
+  // 为什么两法都要 requireAuth：这是不可逆内容丢弃的入口，不能有例外。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/maintenance/reclaim",
+    handler(req, res) {
+      try {
+        if (req.method === "GET") {
+          if (!requireAuth(req, res, apiToken)) return;
+          if (!maintenance) {
+            sendJson(res, 503, { error: RECLAIM_UNAVAILABLE });
+            return;
+          }
+          const parsed = parseReclaimQuery(new URL(req.url, "http://localhost").searchParams.get("olderThanDays"));
+          if (!parsed.ok) {
+            sendJson(res, 400, { error: parsed.error });
+            return;
+          }
+          const { status, body } = reclaimReply({
+            maintenance,
+            dryRun: true,
+            olderThanDays: parsed.olderThanDays,
+            warn: (msg) => ctx.logger?.warn?.(msg)
+          });
+          sendJson(res, status, body);
+          return;
+        }
+        if (req.method !== "POST") {
+          sendJson(res, 404, { error: "not-found" });
+          return;
+        }
+        if (!requireAuth(req, res, apiToken)) return;
+        return readBody(req).then((text) => {
+          const parsed = parseReclaimJson(text);
+          if (!parsed.ok) {
+            sendJson(res, 400, { error: parsed.error });
+            return;
+          }
+          if (!maintenance) {
+            sendJson(res, 503, { error: RECLAIM_UNAVAILABLE });
+            return;
+          }
+          const fields = readReclaimFields(parsed.body);
+          if (!fields.ok) {
+            sendJson(res, 400, { error: fields.error });
+            return;
+          }
+          const { status, body } = reclaimReply({
+            maintenance,
+            dryRun: fields.dryRun,
+            olderThanDays: fields.olderThanDays,
+            vacuum: fields.vacuum,
+            confirm: fields.confirm,
+            warn: (msg) => ctx.logger?.warn?.(msg)
+          });
+          sendJson(res, status, body);
+        });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- rules ---
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/rules",
+    handler(req, res) {
+      try {
+        if (req.method === "PUT" || req.method === "POST") {
+          if (!requireAuth(req, res, apiToken)) return;
+          return readBody(req).then((text) => {
+            const body = parseBody(text);
+            settings.setRules(Array.isArray(body.rules) ? body.rules : []);
+            sendJson(res, 200, { rules: settings.getRules() });
+          });
+        }
+        sendJson(res, 200, { rules: settings.getRules() });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- vector search config ---
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/vector-config",
+    handler(req, res) {
+      try {
+        // Secret-bearing endpoint: fully protected when apiToken is set.
+        if (!requireAuth(req, res, apiToken)) return;
+        if (req.method === "PUT" || req.method === "POST") {
+          return readBody(req).then((text) => {
+            const body = parseBody(text);
+            // An empty apiKey, or one that already looks masked (round-trips
+            // through maskApiKey unchanged), means "keep the existing key".
+            // Only a fresh, unmasked key is treated as a real replacement.
+            const prev = settings.getVectorConfig();
+            const incoming = typeof body.apiKey === "string" ? body.apiKey.trim() : "";
+            const isMaskedOrEmpty = incoming === "" || maskApiKey(incoming) === incoming;
+            const key = isMaskedOrEmpty
+              ? (prev?.apiKey ?? "")
+              : incoming;
+            const cfg = settings.setVectorConfig({
+              enabled: body.enabled,
+              baseUrl: body.baseUrl,
+              apiKey: key,
+              model: body.model
+            });
+            sendJson(res, 200, { config: { ...cfg, apiKey: maskApiKey(cfg.apiKey) } });
+          });
+        }
+        const cfg = settings.getVectorConfig() ?? { enabled: false, baseUrl: "", apiKey: "", model: "" };
+        sendJson(res, 200, { config: { ...cfg, apiKey: maskApiKey(cfg.apiKey) } });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- vector re-index (backfill embeddings for rows missing them) ---
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/vector-reindex",
+    handler(req, res) {
+      try {
+        if (!requireAuth(req, res, apiToken)) return;
+        if (!embedder) {
+          sendJson(res, 200, { indexed: 0, skipped: 0, error: "vector-unavailable" });
+          return;
+        }
+        const url = new URL(req.url, "http://localhost");
+        const limit = Number(url.searchParams.get("limit") ?? 100);
+        // Unified re-index entry: works for both the legacy OpenAI embedder and
+        // the new local/ollama backends (which have no reindexMissing method).
+        const viaIndex = semantic?.vectorIndex && semantic?.vectorIndex.rebuildIndex;
+        const task = viaIndex
+          ? semantic.vectorIndex.rebuildIndex(embedder, { limit })
+          : embedder.reindexMissing ? embedder.reindexMissing(limit) : Promise.resolve({ indexed: 0, skipped: 0, error: "vector-unavailable" });
+        // Return the chain so awaiting callers (tests/health checks) observe the
+        // finished response rather than racing the async backfill.
+        return task.then((result) => {
+          sendJson(res, 200, result);
+        }).catch(() => {
+          sendJson(res, 200, { indexed: 0, skipped: 0, error: "vector-failed" });
+        });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- semantic pipeline status (model, index, reranker) ---
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/semantic",
+    handler(req, res) {
+      try {
+        const stats = semantic?.vectorIndex?.getStats?.() ?? null;
+        // #118: expose readiness so the status card can tell "initializing /
+        // unreachable" from "disabled". An embedder carrying no `ready` prop at
+        // all (test doubles, third-party adapters) is assumed usable.
+        // issue #135: 但「可用」和「配了」是两件事 —— legacy OpenAI embedder
+        // 此前恒报 ready，于是 baseUrl/apiKey/model 全空时状态卡依然是绿的
+        // （"绿色假阳性"）。下面把两者分开，并附上索引覆盖度。
+        const ready = embedder ? ("ready" in embedder ? embedder.ready === true : true) : null;
+        // local/ollama 的配置来自 settings 兜底，没有「未配置」态：它们的
+        // ready=false 表示「还在初始化」，不是「没配」。
+        const configured = embedder ? (typeof embedder.configured === "boolean" ? embedder.configured : true) : null;
+        const embedded = Number(stats?.embeddedCount ?? 0);
+        const total = Number(stats?.totalCount ?? 0);
+        sendJson(res, 200, {
+          ready,
+          configured,
+          // 机器可读的原因；null = 一切正常。面板按它选文案，不用猜布尔组合。
+          reason: !embedder ? "no-embedder"
+            : configured === false ? "not-configured"
+              : ready !== true ? "initializing"
+                : null,
+          // 空库时 ratio 必须是 null 而非 0 —— 0 会被读成「一条都没嵌上」。
+          coverage: { embedded, total, ratio: total > 0 ? embedded / total : null },
+          // 配了、库里有东西可嵌、却一条都没嵌上 = 真的降级（本机此前长期
+          // 处于 ready:true + embeddedCount:0，面板却显示正常）。
+          degraded: configured === true && total > 0 && embedded === 0,
+          // Explicit display name first: the legacy OpenAI-compatible embedder
+          // is an object literal, so constructor.name would be "Object".
+          embedProvider: embedder ? (embedder.name ?? embedder.constructor?.name ?? "unknown") : null,
+          modelHash: embedder?.modelHash ?? null,
+          dimension: embedder?.dimension ?? null,
+          reranker: semantic?.reranker ? "ready" : null,
+          index: stats,
+          // issue #131 / PR-A：自管运行时的状态。无论当前 provider 是不是 local
+          // 都报 —— 面板需要区分「已收编但没启用」和「要用本地却没收编」。
+          // 这里只做结构检查（不跑推理），失败降级成状态对象，不会让本端点 500。
+          // 必须经 publicRuntimeStatus 投影：本端点是免鉴权的，绝对路径与原始错误
+          // 文本不能出去（CWE-200）。完整诊断走本机 CLI 的 status。
+          localRuntime: publicRuntimeStatus(describeLocalRuntime({ runtimeDir: config?.runtimeDir ?? "" }))
+        });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- 一键取回自管运行时（issue #131 / PR-C：三档来源）-------------------------
+  // 写路径（requireAuth）：它会往 ~/.dsh/mneme/runtime/ 落一份依赖闭包。
+  //
+  // 第一档源固定取本 profile 的 node_modules（由插件自身路径推出），不接受客户端传源目录 ——
+  // 「从任意目录往运行时目录里搬东西」没有必要成为对外能力。
+  //
+  // 一律回 200：请求本身合法，成不成看 body 里的 ok/status/reason。让 HTTP 码只表达
+  // 「有没有鉴权/路径对不对」，结论交给 ok —— 面板要能显示「收编了但闭包缺件」这种
+  // 中间状态，用一个 4xx 概括会把信息压掉。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/runtime/provision",
+    handler(req, res) {
+      try {
+        if (req.method !== "POST") {
+          sendJson(res, 405, { error: "method-not-allowed" });
+          return;
+        }
+        if (!requireAuth(req, res, apiToken)) return;
+        return readBody(req).then(async (text) => {
+          const body = parseBody(text);
+          const runtimeDir = config?.runtimeDir ?? "";
+          // 三档来源依次尝试（收编本机 node_modules → 本地 .tgz 目录 → registry）。
+          // 所以这个端点会**联网**：它可能拉一份几十到上百 MB 的闭包。
+          const result = await provisionRuntime({
+            hostModulesDir: hostModulesDir(import.meta.url),
+            runtimeDir,
+            localTarballDir: config?.runtimeTarballDir ?? "",
+            mirror: config?.runtimeMirror ?? "",
+            overwrite: body?.overwrite === true
+          });
+          // 顺手带回最新状态，面板不必再打一次 /semantic（且它走的是免鉴权投影）。
+          sendJson(res, 200, {
+            ...result,
+            localRuntime: publicRuntimeStatus(describeLocalRuntime({ runtimeDir }))
+          });
+        });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- LLM audit trail (Bug8): paginated read + aggregate stats ---
+  // Read-only endpoints, so like list/search/semantic they stay open even when
+  // apiToken is set. The stats aggregate budget by source over the last N days.
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/semantic/llm-audit",
+    handler(req, res) {
+      try {
+        const url = new URL(req.url, "http://localhost");
+        const page = Math.max(1, Number(url.searchParams.get("page") ?? 1) || 1);
+        const pageSize = Math.min(200, Math.max(1, Number(url.searchParams.get("pageSize") ?? 50) || 50));
+        const source = url.searchParams.get("source") ?? undefined;
+        // 写入准入（#254）的 session_key 是内部计数键——session_id 不是对外的归属
+        // 标识（见 scope.js 文件头），而本端点按注释在 apiToken 之下也保持开放，所以
+        // 不回传它。按会话离线聚合直接读库。
+        const items = (service.listLlmAudits?.({ limit: pageSize, offset: (page - 1) * pageSize, source }) ?? [])
+          .map((row) => (row.session_key === undefined ? row : { ...row, session_key: undefined }));
+        const total = service.countLlmAudits?.({ source }) ?? items.length;
+        sendJson(res, 200, { items, total, page, pageSize });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/semantic/llm-audit/stats",
+    handler(req, res) {
+      try {
+        const url = new URL(req.url, "http://localhost");
+        const days = Math.max(1, Math.min(365, Number(url.searchParams.get("days") ?? 7) || 7));
+        const stats = service.getLlmAuditStats?.({ days }) ?? null;
+        sendJson(res, 200, stats ?? { error: "unavailable" });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- ego graph: 1-2 hop neighborhood of one entity (graph panel P1) ---
+  // Read-only like list/search/semantic, so it stays open when apiToken is set.
+  // BFS from the root entity over entity_relations (both directions; the
+  // idx_relations_from/to indexes keep a 2-hop walk in the tens of ms even
+  // for a few thousand nodes). `distance` on each node is the hop count from
+  // the root so the UI can shade the frontier. The API is graph-traversal
+  // only — nodes carry no attr payload; hover summaries come from
+  // /semantic/graph/entity-attrs.
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/semantic/graph/ego",
+    handler(req, res) {
+      try {
+        const url = new URL(req.url, "http://localhost");
+        const name = (url.searchParams.get("entity") ?? "").trim();
+        if (!name) {
+          sendJson(res, 400, { error: "missing-entity" });
+          return;
+        }
+        const root = service.findEntityByName?.(name);
+        if (!root) {
+          sendJson(res, 404, { error: "entity-not-found" });
+          return;
+        }
+        const depth = Math.max(1, Math.min(2, Number(url.searchParams.get("depth") ?? 1) || 1));
+        const limit = Math.max(1, Math.min(100, Number(url.searchParams.get("limit") ?? 40) || 40));
+
+        const nodes = new Map([[root.id, { ...root, distance: 0 }]]);
+        let frontier = [root.id];
+        for (let d = 1; d <= depth && nodes.size < limit; d++) {
+          const next = [];
+          for (const id of frontier) {
+            for (const rel of service.getRelations?.(id) ?? []) {
+              const other = rel.from_entity === id ? rel.to_entity : rel.from_entity;
+              if (nodes.has(other) || nodes.size >= limit) continue;
+              const entity = service.findEntityById?.(other);
+              if (!entity) continue;
+              nodes.set(other, { ...entity, distance: d });
+              next.push(other);
+            }
+          }
+          frontier = next;
+        }
+
+        // Collect every relation whose endpoints both survived the limit cut;
+        // each edge is visited twice (once per endpoint) so dedupe by id.
+        const edgeMap = new Map();
+        for (const id of nodes.keys()) {
+          for (const rel of service.getRelations?.(id) ?? []) {
+            if (nodes.has(rel.from_entity) && nodes.has(rel.to_entity)) {
+              edgeMap.set(rel.id, rel);
+            }
+          }
+        }
+
+        sendJson(res, 200, {
+          root: { id: root.id, name: root.name, type: root.type ?? null, mention_count: root.mention_count ?? 1 },
+          nodes: [...nodes.values()].map((n) => ({
+            id: n.id,
+            name: n.name,
+            type: n.type ?? null,
+            mention_count: n.mention_count ?? 1,
+            distance: n.distance,
+            // v0.7.0 实体热投影：实体热 = 关联记忆 heat 聚合（max），前端据此
+            // 缩放节点大小/明暗。heatEnabled=false 时 entityHeat 返回 null。
+            heat: service.entityHeat?.(n.id) ?? null
+          })),
+          edges: [...edgeMap.values()].map((e) => ({
+            id: e.id,
+            from: e.from_entity,
+            to: e.to_entity,
+            relation_type: e.relation_type,
+            memory_id: e.memory_id ?? null,
+            created_at: e.created_at
+          }))
+        });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- entity attrs: current valid attrs for one entity (graph hover panel) ---
+  // Read-only; mirrors getCurrentAttrs (valid_until IS NULL). Also used as the
+  // graph panel's fallback list when the ego graph is too sparse to draw.
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/semantic/graph/entity-attrs",
+    handler(req, res) {
+      try {
+        const url = new URL(req.url, "http://localhost");
+        const name = (url.searchParams.get("entity") ?? "").trim();
+        if (!name) {
+          sendJson(res, 400, { error: "missing-entity" });
+          return;
+        }
+        const entity = service.findEntityByName?.(name);
+        if (!entity) {
+          sendJson(res, 404, { error: "entity-not-found" });
+          return;
+        }
+        const attrs = service.getCurrentAttrs?.(entity.id) ?? [];
+        sendJson(res, 200, {
+          entity: { id: entity.id, name: entity.name, type: entity.type ?? null, mention_count: entity.mention_count ?? 1 },
+          attrs: Array.isArray(attrs)
+            ? attrs.map((a) => ({
+                key: a.attr_key,
+                value: a.attr_value,
+                confidence: a.confidence ?? null,
+                valid_from: a.valid_from ?? null
+              }))
+            : []
+        });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- entity directory (entity gene v0.3.0): read-only list for the UI ---
+  // Flat entity list (newest-seen first). Attributes and relations stay on
+  // their own per-entity endpoints, the directory only carries the columns
+  // the rail needs: name, type, mention_count, first/last seen.
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/entities",
+    handler(req, res) {
+      try {
+        const url = new URL(req.url, "http://localhost");
+        const limit = Math.min(1000, Math.max(1, Number(url.searchParams.get("limit") ?? 500) || 500));
+        const entities = service.listEntities?.({ limit }) ?? [];
+        sendJson(res, 200, { entities, total: entities.length });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- 一条记忆的关联实体（记忆详情侧栏）-------------------------------------
+  // 只读：entity_attrs.memory_id 反查实体（store 一条 JOIN 完成，按提及去重）。
+  // 与目录 /entities 不同，这里以记忆为锚点，回答"这条记忆提到了谁"。记忆不
+  // 存在或无关联一律返回空数组——详情侧栏不需要区分这两种情况。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/memories/entities",
+    handler(req, res) {
+      try {
+        const url = new URL(req.url, "http://localhost");
+        const memoryId = (url.searchParams.get("memoryId") ?? "").trim();
+        if (!memoryId) {
+          sendJson(res, 400, { error: "missing-memory-id" });
+          return;
+        }
+        sendJson(res, 200, { entities: service.entitiesForMemory?.(memoryId) ?? [] });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- 导出（面板备份/迁移）---------------------------------------------------
+  // 只读。json：全字段行（含 archived/forgotten，布尔化），updated_at DESC。
+  // markdown：一个文档 = 文档级 frontmatter + 按类型分节。frontmatter 只放文档
+  // 最前一份（分节不再各自带文件头，否则 9 个 `---` 块串在一个 .md 里，外部
+  // frontmatter 解析器只认第一段、其余降级成正文），它的 coverage: all 就是一句
+  // 看得见的声明：这份导出含 archived/forgotten（document 指针行除外——它两个
+  // 落点都不进，见 mirror.js 的 MIRROR_EXCLUDED_TYPES）。条目块与磁盘镜像同构（共用
+  // renderMirrorText 的条目渲染），所以导出文本能被 /import 原样吃回；「同构」
+  // 止于条目——磁盘镜像只写活跃集，导出的行集更宽，两者的覆盖声明因此不同
+  // （active-only vs all）。
+  // 全量一次性取回（store.all 按 updated_at DESC）——导出是一次性备份动作，
+  // 不需要流式。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/export",
+    handler(req, res) {
+      try {
+        const url = new URL(req.url, "http://localhost");
+        const format = url.searchParams.get("format") ?? "json";
+        if (format !== "json" && format !== "markdown") {
+          sendJson(res, 400, { error: "invalid-format" });
+          return;
+        }
+        const rows = service.all?.() ?? [];
+        const stamp = new Date().toISOString().slice(0, 10).replace(/-/g, "");
+        if (format === "json") {
+          sendJson(res, 200, {
+            exported_at: new Date().toISOString(),
+            version: PACKAGE_VERSION,
+            count: rows.length,
+            memories: rows
+          }, { "Content-Disposition": `attachment; filename="dsh-mneme-export-${stamp}.json"` });
+          return;
+        }
+        const byType = {};
+        for (const row of rows) {
+          // 没有 TYPE_FILE 条目的 type 不进 markdown 导出，只读 type（#296 的
+          // document 指针行）同样不进：导出/导入是 round-trip 通道，只读视图没有
+          // 可回填的正文。两个排除都在 mirror.js 那张表里（MIRROR_EXCLUDED_TYPES /
+          // MIRROR_READONLY_TYPES）。json 分支是全字段导出，不受影响。
+          if (TYPE_FILE[row.type] && !MIRROR_READONLY_TYPES.has(row.type)) (byType[row.type] ??= []).push(row);
+        }
+        const sections = [];
+        for (const type of Object.keys(TYPE_FILE)) {
+          if (byType[type]?.length) {
+            sections.push(renderMirrorText(type, byType[type], langOf(config), { fileHeader: false }));
+          }
+        }
+        const included = Object.values(byType).flat();
+        const docHeader = renderFileHeader("memory-export", included, "all");
+        sendAttachment(res, 200, "text/markdown; charset=utf-8", `dsh-mneme-export-${stamp}.md`, docHeader + sections.join("\n"));
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- 导入（Markdown 镜像回填）----------------------------------------------
+  // 写路径（requireAuth）。body {type, markdown}：type 是镜像 TYPE_FILE 键，
+  // markdown 是与镜像文件同构的文本。解析复用 readHumanEdits 的纯函数核心 parseHumanEdits（同一实现，
+  // 行为一致是硬约束），合并走 mergeHumanEdits（只吃 title/content；digest 命
+  // 中或无差异的条目在 service 侧自动跳过）。解析出 0 条不算错误。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/import",
+    handler(req, res) {
+      try {
+        if (req.method !== "POST") {
+          sendJson(res, 404, { error: "not-found" });
+          return;
+        }
+        if (!requireAuth(req, res, apiToken)) return;
+        return readBody(req).then((text) => {
+          const body = parseBody(text);
+          if (typeof body.type !== "string" || !Object.hasOwn(TYPE_FILE, body.type)) {
+            sendJson(res, 400, { error: "invalid-type" });
+            return;
+          }
+          // 只读 type 是合法 type，只是没有可回填的正文（#296）：单独的码，调用方
+          // 不会被误导去改 type 名重试。
+          if (MIRROR_READONLY_TYPES.has(body.type)) {
+            sendJson(res, 400, { error: "readonly-type" });
+            return;
+          }
+          if (typeof body.markdown !== "string" || !body.markdown.trim()) {
+            sendJson(res, 400, { error: "invalid-markdown" });
+            return;
+          }
+          const edits = parseHumanEdits(body.markdown);
+          service.mergeHumanEdits(body.type, edits);
+          sendJson(res, 200, { merged: edits.length, type: body.type });
+        });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- 巩固状态（dream 面板）--------------------------------------------------
+  // 只读：最近巩固运行（最多 5 条，created_at DESC）+ 未解决冲突队列。
+  // pendingMemoryIds 从未解决冲突行提取涉及的记忆 id，去重后上限 200，避免积
+  // 压很大时响应失控。listDreamRuns/listConflictPending 走 service 现成的审计
+  // 只读通道（bookkeeping 语义，不触发写钩子）。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/dream-status",
+    handler(req, res) {
+      try {
+        const runs = (service.listDreamRuns?.({ limit: 5 }) ?? []).map((r) => ({
+          created_at: r.created_at,
+          status: r.status,
+          provider: r.provider ?? null,
+          model: r.model ?? null,
+          error: r.error ?? null,
+          run_type: r.run_type ?? "auto",
+          // sleep 审计：heat/时间分层降级决策计数（工作动态可展示"降级 N 条"）。
+          // 数据来自 runSleep 写入的 decisions.demotion（{demoted,archived} 数组）。
+          demotion: r.decisions?.demotion
+            ? { demoted: (r.decisions.demotion.demoted ?? []).length, archived: (r.decisions.demotion.archived ?? []).length }
+            : null
+        }));
+        const pendingConflicts = service.countConflictPending?.() ?? 0;
+        const ids = new Set();
+        for (const row of service.listConflictPending?.({ limit: 200, includeResolved: false }) ?? []) {
+          if (ids.size >= 200) break;
+          if (row.memory_a) ids.add(row.memory_a);
+          if (ids.size >= 200) break;
+          if (row.memory_b) ids.add(row.memory_b);
+        }
+        sendJson(res, 200, {
+          lastRun: runs[0] ?? null,
+          runs,
+          pendingConflicts,
+          pendingMemoryIds: [...ids].slice(0, 200)
+        });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- 冲突集中处理（v0.8.0）：队列 + 人工确认 --------------------------------
+  // GET /conflicts：未解决冲突 + 双方当前内容（此前面板只有计数与徽章，
+  // resolveConflictPending 在存储层躺着没有任何调用方）。POST /conflicts/resolve：
+  // {id, winner: "a"|"b"|null, apply}——盖章审计，apply 时按 dream 非冻结
+  // conflict 路径同款处置（service.resolveConflictPending 内部复用
+  // applyDecisions：CAS + 事务 + 注记 + 幂等）。认证与其他写路由一致。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/conflicts",
+    handler(req, res) {
+      try {
+        if (req.method !== "GET") {
+          sendJson(res, 404, { error: "not-found" });
+          return;
+        }
+        if (!requireAuth(req, res, apiToken)) return;
+        const url = new URL(req.url, "http://localhost");
+        const limit = Math.min(200, Math.max(1, Number(url.searchParams.get("limit") ?? 50) || 50));
+        const offset = Math.max(0, Number(url.searchParams.get("offset") ?? 0) || 0);
+        const items = service.listConflictQueue({ limit, offset });
+        sendJson(res, 200, { items, total: service.countConflictPending?.() ?? items.length });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/conflicts/resolve",
+    handler(req, res) {
+      try {
+        if (req.method !== "POST") {
+          sendJson(res, 404, { error: "not-found" });
+          return;
+        }
+        if (!requireAuth(req, res, apiToken)) return;
+        return readBody(req).then((text) => {
+          let body = {};
+          try { body = JSON.parse(text || "{}"); } catch { /* 坏 JSON → 走缺参分支 */ }
+          const id = typeof body.id === "string" ? body.id.trim() : "";
+          if (!id) {
+            sendJson(res, 400, { error: "missing-id" });
+            return;
+          }
+          const winner = body.winner === "a" || body.winner === "b" ? body.winner : null;
+          const apply = body.apply !== false;
+          const result = service.resolveConflictPending(id, { winner, apply });
+          if (!result) {
+            sendJson(res, 404, { error: "not-found" });
+            return;
+          }
+          sendJson(res, 200, { ok: true, conflict: result });
+        }).catch(() => sendJson(res, 500, { error: "internal" }));
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- health: mirror sync state (F-NEW-03 / v0.3.6) ---
+  // Auth-gated; only returns a sanitized error code (never raw last_error which
+  // may leak paths/token-like strings/internal hosts). On state read failure it
+  // reports unknown/degraded (fail-closed) instead of a false dirty=false.
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/health",
+    handler(req, res) {
+      if (!requireAuth(req, res, apiToken)) return;
+      let state = null;
+      try {
+        state = service.getMirrorHealth?.() ?? null;
+      } catch {
+        // read failure is itself a health signal: do not report a false clean
+        sendJson(res, 200, { mirror: { dirty: null, status: "unknown", last_error: null, last_attempt: null, success_at: null } });
+        return;
+      }
+      if (!state) {
+        sendJson(res, 200, { mirror: { dirty: null, status: "unknown", last_error: null, last_attempt: null, success_at: null } });
+        return;
+      }
+      // Real read failure surfaces as dirty === null (peer blocker 5): report
+      // unknown explicitly instead of collapsing into a false "ok"/"degraded".
+      if (state.dirty === null) {
+        sendJson(res, 200, {
+          mirror: { dirty: null, status: "unknown", last_error: null, last_attempt: null, success_at: null }
+        });
+        return;
+      }
+      // Sanitized: boolean dirty + coarse status only; error string is mapped to
+      // a bounded code, never echoed verbatim.
+      let code = null;
+      if (state.last_error) {
+        const e = String(state.last_error);
+        code = /enospc|no space/i.test(e) ? "no-space" : /permission|eacces/i.test(e) ? "permission" : "sync-failed";
+      }
+      sendJson(res, 200, {
+        mirror: {
+          dirty: state.dirty === true,
+          status: state.dirty === true ? "degraded" : (code ? "degraded" : "ok"),
+          last_error: code,
+          last_attempt: state.last_attempt ?? null,
+          success_at: state.success_at ?? null
+        }
+      });
+    }
+  });
+
+  // --- panel mode (v0.7.12): light / standard -------------------------------
+  // Persists the Web panel's feature preset into the settings kv
+  // ("panel_mode"). PUT requires auth like every other settings write; the
+  // value is validated against the enum. index.js applies the light preset on
+  // the next boot (persisted mode wins over the bundle config).
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/mode",
+    handler(req, res) {
+      try {
+        if (req.method === "PUT" || req.method === "POST") {
+          if (!requireAuth(req, res, apiToken)) return;
+          return readBody(req).then((text) => {
+            const body = parseBody(text);
+            if (body.mode !== "light" && body.mode !== "standard") {
+              sendJson(res, 400, { error: "invalid-mode" });
+              return;
+            }
+            settings.setPanelMode(body.mode);
+            sendJson(res, 200, { mode: settings.getPanelMode() });
+          });
+        }
+        sendJson(res, 200, { mode: settings.getPanelMode() });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- external API settings (the standalone server's panel-facing config) ---
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/external-api",
+    handler(req, res) {
+      try {
+        if (req.method === "PUT" || req.method === "POST") {
+          if (!requireAuth(req, res, apiToken)) return;
+          return readBody(req).then((text) => {
+            const body = parseBody(text);
+            const patch = {};
+            if (body.enabled !== undefined) patch.enabled = body.enabled === true;
+            if (body.port !== undefined) {
+              const port = Number(body.port);
+              if (!Number.isInteger(port) || port < 1 || port > 65535) {
+                sendJson(res, 400, { error: "invalid-port" });
+                return;
+              }
+              patch.port = port;
+            }
+            if (body.host !== undefined) {
+              const host = String(body.host).trim();
+              if (!host || host.includes("://")) {
+                sendJson(res, 400, { error: "invalid-host" });
+                return;
+              }
+              patch.host = host;
+            }
+            sendJson(res, 200, { config: fullExternalConfig(settings.setExternalApi(patch)) });
+          });
+        }
+        // GET: always hand back a token — the standalone server generates its
+        // own on first enabled boot, but the panel displays it before that,
+        // so materialize and persist one here.
+        const kv = settings.getExternalApi() ?? {};
+        if (!kv.token) {
+          kv.token = randomBytes(24).toString("base64url");
+          settings.setExternalApi({ token: kv.token });
+        }
+        sendJson(res, 200, { config: fullExternalConfig(kv) });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- feature flags（功能开关：面板逐项开关后端能力）-------------------------
+  // 读路由保持开放（与 mode/list 一致），前端无需 token 即可渲染开关状态；PUT
+  // 与其他设置写一致走 requireAuth。空/非对象 patch 直接 400：它不携带任何
+  // 意图，静默成功只会掩盖前端 bug。校验失败（未知键/类型/越界）的 400 把键
+  // 名放进 error，前端能直接定位写坏的开关。生效节奏与 panel_mode 相同：
+  // 持久化后下次启动合并进 cfg，本次启动的运行时行为不变。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/features",
+    handler(req, res) {
+      try {
+        if (req.method === "PUT" || req.method === "POST") {
+          if (!requireAuth(req, res, apiToken)) return;
+          return readBody(req).then((text) => {
+            const body = parseBody(text);
+            if (typeof body !== "object" || body === null || Array.isArray(body) || Object.keys(body).length === 0) {
+              sendJson(res, 400, { error: "invalid-patch" });
+              return;
+            }
+            try {
+              settings.setFeatureFlags(body);
+            } catch (error) {
+              sendJson(res, 400, { error: error.message });
+              return;
+            }
+            sendJson(res, 200, featureSnapshot());
+          });
+        }
+        sendJson(res, 200, featureSnapshot());
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // 注入状态（#182，只读）：面板状态页据此决定是否渲染「极简模式下注入按宿主
+  // 设计关闭」提示卡。suppressed=true 需同时满足：autoInject 生效值开着（否则
+  // 关注入是用户自己的选择，不是黑盒）、会话头 preset 为 minimal、且 preset 确
+  // 实观测到了（null=未知会话/宿主不支持，面板静默不猜）。preset 值按字符串
+  // 精确匹配 "minimal"——host bundle 的 preset id（#175 逐字证据），不做模糊。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/inject-status",
+    handler(req, res) {
+      try {
+        if (req.method !== "GET") {
+          sendJson(res, 404, { error: "not-found" });
+          return;
+        }
+        const effective = featureSnapshot().effective;
+        const autoInject = effective.autoInject === true;
+        const agentPreset = lastAgentPreset;
+        const suppressed = autoInject && agentPreset === "minimal";
+        sendJson(res, 200, { autoInject, agentPreset, suppressed });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // 注入预览（#179，只读）：最近一帧 systemPrompt 组装的构成与体积——快照由
+  // inject.js 在真实渲染路径旁路缓存（同一份 candidates，不二次检索）。null =
+  // 尚未发生过渲染（新宿主/新会话）或 autoInject 关闭（注入器未注册），面板对
+  // 两者都以「暂无预览」呈现，不区分也不猜。
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/inject-preview",
+    handler(req, res) {
+      try {
+        if (req.method !== "GET") {
+          sendJson(res, 404, { error: "not-found" });
+          return;
+        }
+        sendJson(res, 200, { snapshot: getInjectionSnapshot() });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  // --- custom commands ---
+  register({
+    kind: "exact",
+    path: "/api/dsh-mneme/commands",
+    handler(req, res) {
+      try {
+        if (req.method === "POST") {
+          if (!requireAuth(req, res, apiToken)) return;
+          return readBody(req).then((text) => {
+            const body = parseBody(text);
+            try {
+              const command = commands.add({
+                name: body.name,
+                description: body.description,
+                instruction: body.instruction
+              });
+              sendJson(res, 200, { command });
+            } catch (error) {
+              sendJson(res, 400, { error: error.message });
+            }
+          });
+        }
+        if (req.method === "DELETE") {
+          if (!requireAuth(req, res, apiToken)) return;
+          const url = new URL(req.url, "http://localhost");
+          const id = url.searchParams.get("id");
+          const removed = id ? commands.remove(id) : false;
+          sendJson(res, 200, { removed });
+          return;
+        }
+        sendJson(res, 200, { commands: commands.list() });
+      } catch {
+        sendJson(res, 500, { error: "internal" });
+      }
+    }
+  });
+
+  return {
+    routes: 34,
+    dispose: () => {
+      for (const dispose of disposers) dispose();
+    }
+  };
+}