@modusensus/dsh-mneme 0.7.15 → 0.7.17

package/README.en.md

@@ -7,7 +7,7 @@ English | [简体中文](README.md)
 [![npm version](https://img.shields.io/npm/v/@modusensus/dsh-mneme?color=blue&label=npm)](https://www.npmjs.com/package/@modusensus/dsh-mneme)
 [![license](https://img.shields.io/badge/license-MIT-green)](LICENSE)
 [![Awesome](https://awesome-dsh-plugin.com/badge.svg)](https://github.com/awesome-dsh-plugin/awesome-dsh-plugin)
-[![tests](https://img.shields.io/badge/tests-645%20passed-success)](https://github.com/modusensus/dsh-mneme)
+[![tests](https://img.shields.io/badge/tests-664%20passed-success)](https://github.com/modusensus/dsh-mneme)
 
 > A cross-session memory plugin for DeepSeek Harness: it lets the Agent remember you, remember your projects, and organize memories automatically. **Mneme** (Μνήμη) — named after Mnemosyne, the Greek goddess of memory who presides over memory and dreams, just as autoDream consolidates memories in the background.
 
@@ -19,7 +19,7 @@ English | [简体中文](README.md)
 
 - **SQLite primary storage**: `~/.dsh/memory/memory.db`, built-in `node:sqlite`, zero native dependencies
 - **Markdown mirror**: `preferences.md` / `projects.md` / `decisions.md` / `history.md` / `summary.md` — human-readable and hand-editable (**manual edits take priority** when merged back into the store)
-- **4+1 memory types**: `preference` / `project` / `decision` / `history` / `summary`
+- **9 memory types**: `preference` / `project` / `decision` / `history` / `summary` / `pattern` + coding-retrospect types `rejected_solution` / `pitfall` / `constraint` (v0.7.13, `codingRetrospect` opt-in; the `user`/`fact` types were dropped in the v0.7.12 rewrite)
 - **Mirror sync state machine (v0.3.6+)**: the mirror is strongly consistent with the primary store, with sync debt modeled via `generation` (desired round) / `applied_generation` (applied round)
   - Business write operations **atomically increment** the desired generation **within their own transaction** — even a crash after COMMIT but before rendering recovers on restart from the durable debt, never silently skipped (v0.3.8)
   - `generation` is incremented with atomic SQLite statements — zero loss under multi-process concurrency; a `CHECK` upper bound rejects negative values/overflow
@@ -53,16 +53,16 @@ English | [简体中文](README.md)
   - `update` (v0.2.1): directly corrects outdated/incorrect content of a single memory (single id / must actually change / not `summary` / 24h protection / ≤2 per run)
 - **Failure tracking (v0.2.1)**: when the user corrects a memory, it is written to the `failure_memories` table (old value/new value), accumulating data for future self-evolution
 - **Summary generation**: after consolidation, a "memory store overview" (single instance) is generated, injected with priority in the next session
-- **Fail-safe**: illegal LLM output (unknown id / invalid action / cross-type merge / out-of-range importance) rejects the entire decision list — the memory store is never corrupted
+- **Fail-safe**: "individually invalid" decisions (unknown id / invalid action / cross-type merge / out-of-range importance, Issue #26) are skipped and the valid subset applied — the run is marked `degraded` and the memory store is never corrupted; only global errors (coverage shortfall / update overflow) reject the whole list
 - **Adjudication audit**: every run writes to the `dream_runs` audit table (input snapshot sha256 digest + full input snapshot + decision list + per-id disposition + receipt), replayable offline; merge / conflict / update are applied idempotently — replays and concurrent duplicate runs have no cumulative side effects; `update` records a `_before` snapshot
 
 #### dreamMaxTokens Tuning Guide
 
-The default `4096` covers ordinary memory stores. When the **memory volume is large** (tens of thousands of characters or more), the decision list and summary may exceed the default budget; scale it up by size:
+The default `8192` covers ordinary memory stores. When the **memory volume is large** (tens of thousands of characters or more), the decision list and summary may exceed the default budget; scale it up by size:
 
 | Memory store size | Recommended `dreamMaxTokens` |
 |-----------|----------------------|
-| Ordinary (<10k chars) | `4096` (default) |
+| Ordinary (<10k chars) | `8192` (default) |
 | Medium (10k–50k chars) | `65536` |
 | Large (>50k chars) | `131072` (cap) |
 
@@ -162,8 +162,24 @@ Every **background LLM call** (autoDream consolidation + summary, autoSummarize
 
 ## 🆕 Recent Release Highlights
 
+> ⚠️ **Archival note**: pre-v0.7.12 rows below record experimental features later removed in the v0.7.12 rewrite (heat model, Wiki-Link, tag system/directory/tag-boost, user/fact layered types, prefix-id resolution, /stats and /directory endpoints). They are version history only — **not current capability**. Current features are what this README's body and the [config table](#-configuration) describe.
+
 | Version | Highlights |
 |------|------|
+| **v0.7.17** | Panel polish: sidebar entry continuously aligned with the host (MutationObserver syncs the New-Session class + `width:100%` + native centering restored, resilient to async skin rewrites) + importance rendered as Lucide star glyphs (`ImportanceStars` filled/hollow component replacing text ★) + toolbar dropdown stacking fix (z-index moved to the container; export/import menu no longer painted under the sticky month header); 664 tests green |
+| **v0.7.16** | Fixed autoDream empty-body failures on thinking models (`no json array in llm output`): restored config-first routing (settings "consolidation model" wins, Issue #25) + reasoningEffort auto-retry without effort on rejection + honest `llm_audit` error on parse failure; backfilled API-route tests (/delete, /entities, /external-api) + lib runtime smoke; 662 tests green |
+| **v0.7.15** | Desktop adaptation: library panel redesign + 30-key feature-flag UI (features API) + status dashboard + import/export (mirror-isomorphic md golden loop) + token masking by default; 645 tests green |
+| **v0.7.14** | Security fix (CWE-200): distillation no longer collects private `reasoning` blocks — only public `text`; 617 tests green |
+| **v0.7.13** | Coding-memory distillation `codingRetrospect` (opt-in: full-transcript atomic memories, 3 new types) + 429 governor (global serial queue + exponential backoff); 616 tests green |
+| **v0.7.12** | Near-rewrite: inlined panel replaced by a pure HTTP API (127.0.0.1:8790 Bearer auth) + standalone zero-dep CLI `dsh-mneme` + lightMode; memory TYPES narrowed 8→6 (user/fact dropped) |
+| **v0.7.11** | Library panel redesign: monthly pagination + infinite scroll + global search + 30s silent refresh + two-step delete + issues #72/#59 fixes; 595 tests green |
+| **v0.7.10** | Web panel UX: memory-type color dots + graph canvas pan/zoom + settings re-grouping + sidebar tab-conflict fix + read-only `/entities` endpoint; 815 tests green |
+| **v0.7.9** | Issue #65 fix: the snapshotEvents shim only landed in src/, never the npm-loaded lib/ — synced lib + pre-publish src↔lib consistency gate (check-sync.js) + lib smoke tests; 815 tests green |
+| **v0.7.8** | DSH 0.1.2-rc.1 compatibility (issues #58 #59): `Session.events` → `snapshotEvents()` shim; autoSummarize & hot-context injection restored; 812 tests green |
+| **v0.7.5** | Layered memory types (user/fact) + Overview view + stats endpoint; 790 tests green |
+| **v0.7.0** | Self-evolving memory: heat power-law decay + per-type half-life + sleep dual-protection + entity heat projection; 757 tests green |
+| **v0.6.0** | Session lifecycle: `session_disposed_at` soft-hide (orthogonal to archived, recoverable) + `memory_delete` description delete; 628 tests green |
+| **v0.5.0** | Recall fusion & memory graph: BM25 three-way recall fusion + ego-graph API + zero-dep SVG force-directed graph + hot memory; 593 tests green |
 | **v0.4.2** | autoSummarize custom model: the `summarizeProvider`/`summarizeModel` config options let you independently designate a lightweight model (e.g. qwen3.6-plus) for session summaries, saving main-model tokens; 473 tests green |
 | **v0.4.0** | System-level Sleep Mode: idle-triggered four-phase deep maintenance (conflict resolution / archival demotion / pattern discovery / relation completion), interruptible, serially safe, fail-safe; tiered compression releases cold memories; 471 tests green |
 | **v0.3.9** | Fixed 4 FAILs from the third-party audit: CAS made atomic within the same transaction, mirror degraded-receipt passthrough, per-type physical terminal-state convergence, strict integer validation for generation and stabilized concurrent initialization |
@@ -187,7 +203,15 @@ Every **background LLM call** (autoDream consolidation + summary, autoSummarize
 | **v0.4.5** | ✅ Done | Epistemic trust + recall eval | Memory credibility grading `trustEpistemicWeighting` (observation>inferred>subjective: retrieval ranking favors high-credibility memories, injection tags `[verified]`, dream merge/conflict favors the more credible side; opt-in, off by default) + retrieval evaluation `evaluateRetrieval` persisted to `recall_evals` (`evalPersistTestResults` opt-in, off by default; production retrieval always goes through `recall_runs`, unconditionally isolated); 518 tests green |
 | **v0.4.6** | ✅ Done | 8 fixes (vector pipeline + injection/quality/audit) | Vector pipeline fixes (embedSingle adaptation / `autoReindexOnBoot` backfill of existing data / `vector_meta` metadata) + injection semantic recall `hybridInject` + same-title append `content_history` + injection length caps (300 per item / 1500 per block) + memory quality filter `memoryQualityFilter` + LLM usage audit `llmAudit` (table + instrumentation + read-only APIs); 553 tests green |
 | **v0.4.7** | ✅ Done | Idempotent schema migrations | When the same db is opened concurrently, the `PRAGMA table_info` check and ALTER are non-atomic and may repeat `ADD COLUMN`, failing with a duplicate column name; switched to an `addColumn` helper that swallows the race (try/catch), unifying all 12 migration sites |
-| **v0.5.0+** | 🚀 Long-term | Self-evolving memory | Interest drift tracking + cross-workspace memory sharing (pending DSH support) |
+| v0.5.0 | ✅ Done | Recall fusion & memory graph | BM25 three-way recall fusion + ego-graph API + zero-dep SVG force-directed graph + hot memory + recall benchmark; 593 tests green |
+| v0.6.0 | ✅ Done | Session lifecycle | `session_disposed_at` soft-hide (orthogonal to archived, recoverable) + `memory_delete` description delete + event circuit-breaker; 628 tests green |
+| v0.6.x | ✅ Done | Panel enhancements + fixes | 7 further 0.6.x releases: Wiki-Link/tag/directory experiments (later removed in v0.7.12), #25/#26 fixes, allowCrossTypeMerge, ID-exposure hardening, version-sync discipline; up to 735 tests green |
+| v0.7.0 | ✅ Done | Self-evolving memory | heat power-law decay + per-type half-life + sleep dual-protection + entity heat projection + recall_runs marking + 90-day cleanup; 757 tests green (heat model later removed in v0.7.12) |
+| v0.7.1–0.7.8 | ✅ Done | Issue fixes + graph backfill | tags↔entity_attrs bridge, inline-confirm delete, sidebar trigger toggle, brace escaping, user/fact layered types + stats endpoint (later removed), prefix-id resolution (later removed), sleep batch entity extraction, snapshotEvents() DSH compat; 764→812 tests green |
+| v0.7.9 | ✅ Done | lib-sync gate | Issue #65: src-only shim silently killed the shipped lib — synced lib + pre-publish src↔lib consistency check + lib smoke tests; 815 tests green |
+| v0.7.10–0.7.12 | ✅ Done | Panel redesign + near-rewrite | UX upgrades (color dots, pan/zoom, pagination, global search), then near-rewrite: inlined panel → pure HTTP API + standalone CLI + lightMode; TYPES narrowed 8→6 |
+| v0.7.13–0.7.17 | ✅ Done | Distillation, security, desktop, dream fix, panel polish | codingRetrospect + 429 governor; private reasoning blocks dropped (CWE-200); desktop panel redesign + feature flags; autoDream thinking-model empty-body fix + backfilled tests; sidebar entry alignment + star glyphs + menu stacking polish; 664 tests green |
+| **v0.8.0** | 🚧 Planned (late Sep) | Graph enhancement | Interest-drift visualization + scope isolation (issue #17) + cross-workspace sharing |
 
 > All new capabilities ship as **toggleable features** (enabled/disabled via configuration), conservatively on by default and never breaking existing behavior. The `failure_memories` table and the autoDream decision engine have already paved the way for future reflective growth.
 
@@ -256,8 +280,8 @@ It works out of the box with the defaults. To adjust, override in `~/.dsh/profil
 | `dreamThresholdCount` | `10` | Memory count threshold that triggers consolidation |
 | `dreamThresholdChars` | `5000` | Total character threshold that triggers consolidation |
 | `dreamDelayMs` | `2000` | Asynchronous consolidation delay (debounce) |
-| `dreamProvider` / `dreamModel` | empty | LLM route fallback for dream (defaults to the agent's default model) |
-| `dreamMaxTokens` | `4096` | Maximum tokens per dream LLM call (cap 131072; increase for large memory stores — see the tuning guide below) |
+| `dreamProvider` / `dreamModel` | empty | Explicit dream LLM route — config wins over the agent's default model (config-first, v0.7.16); left empty, the agent's default model is used |
+| `dreamMaxTokens` | `8192` | Maximum tokens per dream LLM call (cap 131072; increase for large memory stores — see the tuning guide below) |
 | `dreamReasoningEffort` | `none` | Reasoning-effort passthrough for the dream LLM: `low` / `medium` / `high` / `none` (`none` = omit the field and use the model default; set `low` when a reasoning model exhausts its budget on reasoning and produces an empty body) |
 | `apiToken` | empty | Optional API auth token; once set, write operations and key endpoints require `Authorization: Bearer <apiToken>` |
 | `embedProvider` | `openai` | Semantic backend: `openai` (default, v0.1-compatible) / `local` (ONNX offline) / `ollama` |
@@ -405,7 +429,7 @@ src/
 lib/
 ├── client.js         # Web 面板（手写 ModuleLoader bundle）
 └── *.js              # src 的同步分发产物
-test/                 # 645 个 node:test 测试（含审计与三轴线压测不变量）
+test/                 # 662 node:test tests (audit + three-axis stress invariants)
 scripts/              # e2e-dsh.js 端到端演示 · stress-dsh.js 三轴线压测 · sync-lib.js 同步
 ```
 
@@ -414,7 +438,7 @@ scripts/              # e2e-dsh.js 端到端演示 · stress-dsh.js 三轴线压
 ```bash
 cd dsh-mneme
 npm install        # 安装 peer 依赖（以 devDependencies 形式，用于本地测试）
-npm test           # 运行 450 个测试
+npm test           # 运行 662 个测试
 npm run stress     # 三轴线压测：长会话检索 / 冲突仲裁 / 多 Agent 并发（离线 mock LLM）
 npm run sync       # 把 src/ 同步到 lib/（发布时由 prepack 钩子自动执行）
 ```

package/README.md

@@ -5,7 +5,7 @@
 [![npm version](https://img.shields.io/npm/v/@modusensus/dsh-mneme?color=blue&label=npm)](https://www.npmjs.com/package/@modusensus/dsh-mneme)
 [![license](https://img.shields.io/badge/license-MIT-green)](LICENSE)
 [![Awesome](https://awesome-dsh-plugin.com/badge.svg)](https://github.com/awesome-dsh-plugin/awesome-dsh-plugin)
-[![tests](https://img.shields.io/badge/tests-645%20passed-success)](https://github.com/modusensus/dsh-mneme)
+[![tests](https://img.shields.io/badge/tests-664%20passed-success)](https://github.com/modusensus/dsh-mneme)
 [![CI](https://img.shields.io/github/actions/workflow/status/modusensus/dsh-mneme/ci.yml)](https://github.com/modusensus/dsh-mneme/actions)
 [![node](https://img.shields.io/badge/node-24%2B-blue)](https://nodejs.org)
 [![npm downloads](https://img.shields.io/npm/dm/@modusensus/dsh-mneme?color=blue&label=downloads)](https://www.npmjs.com/package/@modusensus/dsh-mneme)
@@ -41,8 +41,8 @@ dsh web
 ### 记忆存储（SQLite + Markdown 镜像）
 
 - **SQLite 主存储**：`~/.dsh/memory/memory.db`，`node:sqlite` 内置，零原生依赖
-- **Markdown 镜像**：`preferences.md` / `projects.md` / `decisions.md` / `history.md` / `summary.md` / `user.md` / `facts.md`，人类可读、可手工编辑（**人工修改优先**合并回库）
-- **8 种记忆类型**：用户可存 6 种（`preference` 偏好 / `project` 项目 / `decision` 决策 / `history` 历史 / `user` 用户画像 / `fact` 原子事实）+ AI 自动生成 2 种（`summary` 会话总览 / `pattern` 模式）（v0.7.5 起新增 user/fact 两个轻量分层，单表 `type` 字段扩展，不动表结构）
+- **Markdown 镜像**：`preferences.md` / `projects.md` / `decisions.md` / `history.md` / `summary.md`，人类可读、可手工编辑（**人工修改优先**合并回库）
+- **9 种记忆类型**：`preference` 偏好 / `project` 项目 / `decision` 决策 / `history` 历史 / `summary` 会话总览 / `pattern` 模式 + 编码记忆三型 `rejected_solution` 被否方案 / `pitfall` 踩坑 / `constraint` 约束（v0.7.13 起，`codingRetrospect` 默认关；v0.7.12 重写已收窄删除 user/fact 两型）
 - **镜像同步状态机（v0.3.6+）**：mirror 与主库强一致，用 `generation`（期望轮次）/ `applied_generation`（已应用轮次）建模同步债务
   - 业务写操作在**自身事务内原子递增** desired generation——崩溃在 COMMIT 后、渲染前，重启也能凭 durable 债务恢复，绝不静默跳过（v0.3.8）
   - `generation` 用 SQLite 原子语句递增，多进程并发零丢失；带 `CHECK` 上界，负数/溢出拒绝
@@ -57,7 +57,7 @@ dsh web
 | `memory_search` | 全文搜索（中文子串友好，可启用向量语义搜索） |
 | `memory_list` | 按类型分页列出（`include_archived=true` 可查看已归档） |
 | `memory_update` | 修改已有记忆 |
-| `memory_delete` | 删除记忆（v0.6.0 支持 `query` 按描述删除，不只能靠记忆 ID） |
+| `memory_delete` | 删除记忆（按记忆 ID 精确删除） |
 | `memory_forget` | 抑制注入（降权不删除，可恢复） |
 | `memory_archive` | 归档/恢复记忆（v0.2.5；归档后隐藏于列表/搜索/注入/整理，`archived=false` 可恢复） |
 
@@ -76,20 +76,20 @@ dsh web
   - `update`（v0.2.1）：直接修正单条记忆的过时/错误内容（单 id / 必须实际变化 / 非 summary / 24h 保护 / 每次 ≤2）
 - **失败追踪（v0.2.1）**：用户纠正记忆时写入 `failure_memories` 表（旧值/新值），为后续自进化积累数据
 - **摘要生成**：整理后生成"记忆库总览"（单一实例），作为下次会话的优先注入
-- **Fail-safe**：非法 LLM 输出（未知 id / 非法 action / 越界 importance 等）拒绝整单，绝不破坏记忆库；跨类型合并这类"单条非法"决策（Issue #26）默认跳过并应用合法子集（run 记为 `degraded`），`dreamSkipInvalid:false` 可恢复旧的整单拒绝
+- **Fail-safe**：非法 LLM 输出（未知 id / 非法 action / 越界 importance / 跨类型合并等"单条非法"决策，Issue #26）默认跳过该条并应用合法子集（run 记为 `degraded`），绝不破坏记忆库
 - **裁决审计**：每次运行写入 `dream_runs` 审计表（输入快照 sha256 digest + 完整输入快照 + 决策清单 + 逐 id 去向 + receipt），可离线回放；merge / conflict / update 幂等应用，重放/并发重复执行无累积副作用；update 记录 `_before` 快照
 
 #### dreamMaxTokens 调优指南
 
-默认 `4096` 已覆盖常规记忆库。当**记忆量大**（数万字符以上）时，决策清单与摘要可能超过默认预算，建议按规模调大：
+默认 `8192` 已覆盖常规记忆库。当**记忆量大**（数万字符以上）时，决策清单与摘要可能超过默认预算，建议按规模调大：
 
 | 记忆库规模 | 建议 `dreamMaxTokens` |
 |-----------|----------------------|
-| 常规（<1 万字） | `4096`（默认） |
+| 常规（<1 万字） | `8192`（默认） |
 | 中等（1 万-5 万字） | `65536` |
 | 大型（5 万字以上） | `131072`（上限） |
 
-> 若使用**思考型模型**（如 deepseek-v4-flash / DeepSeek-R1 类），模型可能把全部预算花在 reasoning 上导致正文为空（日志出现 `no json array in llm output`）。此时把 `dreamReasoningEffort` 设为 `off` 显式关闭思考（实测 8192 token 预算不再被推理吃光，输出恢复正常）；sleep 侧对应 `sleepReasoningEffort`。默认 `none` 不传该字段，完全沿用模型自身默认，行为与旧版本一致。
+> 若使用**思考型模型**（如 deepseek-v4-flash / DeepSeek-R1 类），模型可能把全部预算花在 reasoning 上导致正文为空（日志出现 `no json array in llm output`，v0.7.16 已修复）。此时把 `dreamReasoningEffort` 设为 `low` 显式压低思考（`none` 默认不传该字段，完全沿用模型自身默认，行为与旧版本一致）；即使被方舟拒绝该参数，v0.7.16 起也会自动去掉重试一次。sleep 侧对应 `sleepReasoningEffort`。
 
 ### Sleep Mode 系统级睡眠 💤（v0.4.0，opt-in）
 
@@ -116,22 +116,6 @@ dsh web
 - **幂等 + 熔断**：dispose/restore 状态守卫幂等（重复调用 no-op）；事件回调内部异常 catch 住，不抛进 DSH 会话清理流程
 - **恢复**：整会话 `service.restoreBySession(sessionId)` 一键还原；`service.listBySession(sessionId, { includeDisposed: true })` 可查看当前隐藏了哪些（DTO 带 `disposed` 标记）
 
-### Wiki-Link 双向链接 🔗（v0.6.1，opt-in）
-
-**默认关闭**（`wikiLinkEnabled: false`）。开启后，记忆正文支持 `[[target]]` / `[[显示|target]]` 双括号链接语法：保存时自动解析并记录跨记忆 `links_to` 关系（partial 唯一索引只对 links_to 去重），记忆详情旁出现反向链接面板，点击可跳转来源记忆。提供只读 API：backlinks / forward-links / wikilink-resolve（输出脱敏）。
-
-### Tag 系统 🏷️（v0.6.2，opt-in）
-
-`autoTagEnabled` 默认关。开启后：正文 `#标签` 即时标记（规则 `[a-zA-Z0-9_一-龥-]+`，≤20 字符，非法/超长自动丢弃）；autoDream 整理后 LLM 自动打 1-3 个标签（`autoTagMaxPerRun=10` 频控，fail-safe）；tags 存 `entity_attrs`；`tag:` 搜索前缀可与关键词/`entity:`/`attr:` 组合；mirror 顶部渲染 `#tag` 行；记忆面板标签可点击（过滤）、添加、移除（`manualTagEnabled` 默认开）。存储与搜索能力始终可用，`autoTagEnabled` 只控制自动打标。
-
-### 目录视图 📁（v0.6.3）
-
-记忆面板新增「目录」视图：以 Tag 为轴心的一级手风琴文件夹，无标签记忆自动沉底到「无标签」兜底组；组内按重要性/时间双降序；点击条目直达详情页。数据由 `GET /api/dsh-mneme/directory` 提供（只读、排除 disposed/archived/forgotten）。
-
-### Tag 加权召回 ⚖️（v0.6.4，opt-in）
-
-`tagBoostEnabled` 默认关。开启后，三路召回合并出候选后做 tag 加权重排：候选记忆 tags 与 Query 提取 tags（`#xxx` + 已知 tag 列表）交集 → 分数 `×1.15`；与当前 Session 热记忆 tags 交集 → `×1.08`（可叠加，上限 1.0）。关闭时行为与旧版完全一致，可用 `evaluateRetrieval` 开关对比调优。
-
 官方设置面板 → 「记忆库设置」→「记忆」标签：按类型浏览、全文搜索；启用向量搜索后可用「语义」切换做向量召回。
 
 ### 用户设置（画像 / 规则）与自定义指令 ⚙️
@@ -226,21 +210,14 @@ v0.3.0 起新增**记忆基因**层：从记忆里抽取**命名实体**、**带
 - **选择性注入**：query 向量可用时注入候选按主题相似度重排（`selectiveInjectEnabled` 可关）；**搜索时语义去重**为激进选项（`searchSemanticDedup=true` 显式开启，近重复行 Rerank 前丢弃）
 - **召回基准**（`scripts/benchmark-recall.js`）：标准查询集驱动，计算 Recall@5 与 MRR，`legacy`（三特性全关）vs `fused`（默认配置）双跑对比
 
-### 自进化记忆 🌡️（v0.7.0，默认保守开启）
-
-让记忆库从"存得准、召得回"进化为会自我衰减、识别兴趣漂移的智能体：
-
-- **heat 热度模型**（`src/heat.js`）：类遗忘曲线幂律衰减 `H = 1/(1+λ·Δt)^α`，per-type 差异化半衰期（TYPE_DECAY）：preference/pattern/summary 免疫（λ=0，热度恒 1.0），project 慢衰减（λ=0.0008），decision 中速（λ=0.002），history 较快（λ=0.006）；全局参数 `heatGlobalAlpha`(默认 1.2) 控制衰减速度
-- **sleep 热联合双保护**：降级需同时满足"冷"（heat < `sleepHeatThreshold` 0.05）+"非紧要"（importance < 5）+"非免疫类型"三重条件——冷但重要（如遗忘的高价值决策）与热但低值（如刚访问的闲聊）都受到保护，免疫类型（preference 等）永不降级
-- **updated_at 语义修正**：合并/更新刷新的 `updated_at` 不再计为访问（`last_accessed_at` 独立追踪），杜绝 autoDream 合并动作伪装成"刚被召回"；触达数据采集（touchRecalled）由 `heatEnabled` 门控
-- **recall_runs 数据前提**：检索记录默认开启（`recallRecordDefault: true`），candidates 打 `injected` 标记区分"被召回"（false）与"被注入上下文"（true）两个消耗强度；注入场景也记账（mode="inject"）；90 天滚动清理（`recallRetentionDays`）防膨胀
-- **实体热投影**（ego-graph API + 前端）：`entityHeat` 取关联记忆 heat 的 max 值，API 返回 `heat` 字段；前端节点大小/明暗随热度变化（`nodeRadius`/`fillOpacity`），兴趣漂移在图谱上可见
-- **配置**：`heatEnabled`(默认 true)、`heatTypeDecay`(per-type λ)、`heatGlobalAlpha`(1.2)、`sleepHeatThreshold`(0.05)、`recallRecordDefault`(true)、`recallRetentionDays`(90)
-
 ## 🆕 最近版本亮点
 
+> ⚠️ **历史存档提示**：下方及路线图表中 v0.7.12 之前的早期条目记录的实验功能（heat 热度模型、Wiki-Link、tag 系统/目录/tag 加权、user/fact 分层类型、前缀 id 解析、/stats 与 /directory 端点等）已在 v0.7.12 近重写时移除，仅作版本历史存档，**不代表当前能力**。当前特性以本 README 正文与 [配置表](#-配置) 为准。
+
 | 版本 | 亮点 |
 |------|------|
+| **v0.7.17** | 面板体验细化：侧边栏入口持续对齐宿主（MutationObserver 同步「新会话」类名 + `width:100%` + 交还原生居中，皮肤异步改写不再失配）+ 重要性星级换 Lucide 星形（`ImportanceStars` 实心/空心组件，替换文本 ★）+ 工具栏下拉层级修复（z-index 提到容器，导出/导入菜单不再被吸顶月份头遮挡）；664 测试全绿 |
+| **v0.7.16** | 修复 autoDream 在 thinking 模型上空体 failed（`no json array in llm output`）：恢复 config-first 路由（设置面板「巩固模型」生效，Issue #25）+ reasoningEffort 被拒自动去掉重试一次 + 解析失败如实记 llm_audit error 并带原始输出头日志；补测 API 路由空白（/delete、/entities、/external-api）+ lib 运行时冒烟（src↔lib 一致性）；662 测试全绿 |
 | **v0.7.15** | 桌面端适配：记忆库面板重设计（撤对话 tab → 居中非全屏 sheet、卡片网格/时间线双视图、右侧详情抽屉、编辑/归档/关联实体）+ 侧边栏入口上移工作区上方（借用宿主原生类名对齐、收起态零位移）+ 功能开关 30 键上 UI（features API，巩固模型与 embedding 提供方可配）+ 状态页工作台（巩固卡/工作动态流/沉淀记忆/归档恢复）+ 导入导出（镜像同构 md 黄金闭环）+ Token 面板默认遮蔽；645 测试全绿 |
 | **v0.7.14** | 安全修复（CWE-200）：蒸馏不再把助手 `reasoning` 私有推理块送进蒸馏上下文（只采公开 `text`，防止记忆沉淀私有思考链）；617 测试全绿 |
 | **v0.7.13** | 编码记忆蒸馏（`codingRetrospect` 默认关）：完整转录（用户+助手回答+工具调用/结果+代码执行）提炼原子记忆，新增 `rejected_solution`/`pitfall`/`constraint` 三类型，编码任务 `codingBoostFactor` 加权（cap 5）+ 智能调速器（蒸馏全局串行队列 + 429 指数退避自动重试）+ 原子记忆语义保留（宁可拆多条不合并丢细节，`distillMaxChars` 默认 24000）+ 修复 v0.7.12 CI 回归（恢复 c8 覆盖率）；616 测试全绿 |
@@ -309,7 +286,14 @@ v0.3.0 起新增**记忆基因**层：从记忆里抽取**命名实体**、**带
 | **v0.7.8** | ✅ 完成 | DSH 0.1.2-rc.1 兼容（issues #58 #59） | 官方移除 `Session.events` 属性、改用 `snapshotEvents()` 方法后 autoSummarize 与 hot-context 注入失效；改为兼容垫片 `session.snapshotEvents?.() ?? session.events`，新旧 DSH 通吃，老版本不受影响；新增 2 个回归用例；812 测试全绿 |
 | **v0.7.6** | ✅ 完成 | issue #48 修复 | 四工具统一 `service.resolveMemoryId`：截断/前缀短 id 也能精确操作（精确命中优先、唯一前缀解析、多命中拒绝并列出候选、`memory_delete` 未命中幂等返回 + `logger.warn` 留痕）；Web bundle `client.js` 改 src 正源；801 测试全绿 |
 | **v0.7.9** | ✅ 完成 | issue #65 修复 | v0.7.8 的 snapshotEvents 适配未同步 lib/（npm 实际加载 lib/）导致发布产物静默失效；补齐 lib 三处垫片 + 发布前 src↔lib 一致性校验（`scripts/check-sync.js`，root prepack 调用，漂移 exit 1）+ `test/lib-smoke.test.js` 从 lib 导入复跑 + 一致性断言；815 测试全绿 |
-| **v0.8.0** | 🚧 计划中（9 月末） | 图谱增强 | 兴趣漂移可视化 + scope 隔离（issue #17）+ 跨 workspace 记忆共享 + 更多 heat 信号 |
+| **v0.7.10** | ✅ 完成 | Web 面板体验升级 | 记忆类型色点体系 + 图谱画布平移/缩放（0.5x–3x 光标锚定）+ 设置页分区重排 + 侧边栏同标签冲突修复 + 详情 meta 精排 + 只读端点 `GET /api/dsh-mneme/entities`；815 测试全绿 |
+| **v0.7.11** | ✅ 完成 | 记忆库面板改版 | 按月分页 + 无限滚动 + 搜索全局化 + 30s 静默刷新 + 状态子页四卡 + 删除两步确认 + 重要性过滤芯片 + 内联 Lucide 图标；bundle 默认开实体抽取；修复 issue#72 图谱节点裁剪 + issue#59 autoSummarize 垫片；595 测试全绿 |
+| **v0.7.12** | ✅ 完成 | 近重写：纯 HTTP API + CLI | 内置面板 client.js 删除改纯 HTTP API（`127.0.0.1:8790` Bearer 鉴权，api.js/api-standalone.js）+ 独立 CLI `dsh-mneme`（bin/cli.mjs 零依赖）+ 轻量模式 lightMode + 设置面板新卡片；store TYPES 收窄 8→6（删 user/fact）、distill VALID →4、saveWithDedupe 改单参 |
+| **v0.7.13** | ✅ 完成 | 编码记忆蒸馏 + 调速器 | `codingRetrospect`（默认关：完整转录提炼原子记忆，新增 rejected_solution/pitfall/constraint 三类型）+ 智能调速器（蒸馏全局串行队列 + 429 指数退避重试）+ 语义保留（distillMaxChars 默认 24000）；616 测试全绿 |
+| **v0.7.14** | ✅ 完成 | 安全修复（CWE-200） | 蒸馏不再采集私有推理块：`collectMessages` 只采公开 text，防止记忆库沉淀模型私有思考链；617 测试全绿 |
+| **v0.7.15** | ✅ 完成 | 桌面端适配 | 记忆库面板重设计 + 功能开关 30 键 UI（features API）+ 状态页工作台 + 导入导出（镜像同构 md 黄金闭环）+ Token 面板默认遮蔽；645 测试全绿 |
+| **v0.7.16** | ✅ 完成 | autoDream thinking 模型空体修复 + 补测 | 恢复 config-first 路由（设置面板「巩固模型」生效，Issue #25）+ reasoningEffort 被拒自动去掉重试 + 解析失败如实记 llm_audit error；补测 API 路由空白（/delete、/entities、/external-api）+ lib 运行时冒烟；662 测试全绿 |
+| **v0.8.0** | 🚧 计划中（9 月末） | 图谱增强 | 兴趣漂移可视化 + scope 隔离（issue #17）+ 跨 workspace 记忆共享 |
 
 > 新能力一律做成**可开关的功能**（配置启用/关闭），默认保守开启、不破坏现有行为。`failure_memories` 表与 autoDream 决策引擎已为后续反思性成长铺好路。
 
@@ -382,9 +366,7 @@ dsh web
 | `dreamDelayMs` | `2000` | 整理异步延迟（去抖） |
 | `dreamProvider` / `dreamModel` | 空 | dream 的 LLM 路由覆盖（显式配置优先于 agent 默认模型；留空则回退到 agent 默认模型） |
 | `dreamMaxTokens` | `8192` | dream LLM 调用最大 token 数（上限 131072；大记忆量建议调大，见下方调优指南） |
-| `dreamReasoningEffort` | `none` | dream LLM 推理强度透传：`off` / `low` / `medium` / `high` / `none`（`off`=显式关闭思考，思考型模型（如 deepseek-v4-flash）必须用它，否则推理会烧光 token 预算导致正文为空；`none`=不传该字段，沿用模型默认） |
-| `dreamSkipInvalid` | `true` | 跳过非法决策（Issue #26，默认开）：跨类型 merge 等"单条非法"决策不再整单拒绝，而是跳过该决策、应用合法子集，run 记为 `degraded`（`applied>0`，autoTag 照常触发）；设为 `false` 恢复旧的"任意非法即整单拒绝"（`applied=0`）。防洗白语义不变——显式覆盖率不足 / update 超量等全局错误仍整单拒绝 |
-| `allowCrossTypeMerge` | `false` | 允许跨类型合并（Issue #26，默认关）：类型有语义边界（`preference` 注入权重更高、`decision`/`project` 注入上下文不同），跨类型合并默认视为非法并被跳过；显式开启后放宽该检查，跨类型 merge 可被应用，类型边界由用户自行承担 |
+| `dreamReasoningEffort` | `none` | dream LLM 推理强度透传：`low` / `medium` / `high` / `none`（`none`=不传该字段，沿用模型默认；思考型模型（如 deepseek-v4-flash）想压低思考可设 `low`；被方舟拒绝该参数时 v0.7.16 起自动去掉重试一次） |
 | `apiToken` | 空 | 可选 API 鉴权 token；设置后写操作与密钥接口要求 `Authorization: Bearer <apiToken>` |
 | `embedProvider` | `openai` | 语义后端：`openai`（默认，兼容 v0.1）/ `local`（ONNX 离线）/ `ollama` |
 | `localEmbedModel` | `Xenova/bge-small-zh-v1.5` | 本地 ONNX embedding 模型 |
@@ -541,10 +523,9 @@ src/
 ├── hot-memory.js     # 会话级短期热记忆（v0.5.0：滚动轮次 + token 预算）
 ├── embedding.js      # OpenAI 兼容 embeddings 客户端 + 向量检索
 ├── api.js            # HTTP 路由（Web 面板数据通道）
-├── client.js         # Web 面板 bundle（ModuleLoader 自注册；v0.7.6 起 src 正源）
 └── index.js          # 插件接线
-lib/                  # src 的同步分发产物（npm run sync；发布前由 root prepack 的 check-sync.js 校验一致性，无手写例外）
-test/                 # 645 个 node:test 测试（审计与三轴线压测不变量；src↔lib 一致性由 scripts/check-sync.js 发布闸门校验）
+lib/                  # src 的同步分发产物（npm run sync；发布前由 root prepack 的 check-sync.js 校验一致性；唯一手写例外 lib/client.js——Web 面板 bundle，sync 不覆盖）
+test/                 # 662 个 node:test 测试（审计与三轴线压测不变量；src↔lib 一致性由 scripts/check-sync.js 发布闸门校验）
 scripts/              # e2e-dsh.js 端到端演示 · stress-dsh.js 三轴线压测 · sync-lib.js 同步 · check-sync.js 发布闸门 · benchmark-recall.js 召回基准
 ```
 
@@ -553,14 +534,14 @@ scripts/              # e2e-dsh.js 端到端演示 · stress-dsh.js 三轴线压
 ```bash
 cd dsh-mneme
 npm install        # 安装 peer 依赖（以 devDependencies 形式，用于本地测试）
-npm test           # 运行 645 个测试
+npm test           # 运行 662 个测试
 npm run stress     # 三轴线压测：长会话检索 / 冲突仲裁 / 多 Agent 并发（离线 mock LLM）
 npm run sync       # 把 src/ 同步到 lib/（发布时由 prepack 钩子自动执行）
 ```
 
 > 压测（`npm run stress`）三条轴线：**长会话检索**（Recall@k、陈旧残留率）、**冲突裁决**（可重放仲裁集：审计快照 hash + receipt + 幂等回放）、**多 Agent 并发**（丢更新、重复合并、事务/崩溃恢复）。每次 autoDream 运行都会写入审计表 `dream_runs`（输入快照 digest + 决策清单 + 逐 id 去向 + receipt），让高通过率下也能定位静默错误。
 
-> `lib/` 是 `src/` 的同步分发产物（`npm run sync`）：v0.7.6 起全部由 src 复制而来，**不再有任何手写 lib 文件**——Web 面板 bundle 也以 `src/client.js` 为唯一正源。
+> `lib/` 是 `src/` 的同步分发产物（`npm run sync`）：除 **`lib/client.js`**（Web 面板 bundle）为 lib 独有的手写文件、sync 与 check-sync 均不触碰外，其余全部由 src 复制而来。改 src 必 `npm run sync` 并提交 lib，否则发布产物静默失效（issue #65 教训）。
 
 ## 📄 设计文档
 

package/lib/client.js

@@ -74,6 +74,24 @@ window.__ModuleLoader__.load({
       }, parts.map(([tag, attrs], i) => h(tag, { key: i, ...attrs })));
     };
 
+    // Lucide v1.42 star path（morphicons 官方配套的数据包）。morphicons 本体
+    // 是 ESM-only 的变形动画引擎、插件运行时不允许 require 第三方库，故按
+    // 既有惯例内联图标数据静态渲染；实心/空心由 fill 区分，空心降透明度。
+    const STAR_PATH_D = "M11.525 2.295a.53.53 0 0 1 .95 0l2.31 4.679a2.123 2.123 0 0 0 1.595 1.16l5.166.756a.53.53 0 0 1 .294.904l-3.736 3.638a2.123 2.123 0 0 0-.611 1.878l.882 5.14a.53.53 0 0 1-.771.56l-4.618-2.428a2.122 2.122 0 0 0-1.973 0L6.396 21.01a.53.53 0 0 1-.77-.56l.881-5.139a2.122 2.122 0 0 0-.611-1.879L2.16 9.795a.53.53 0 0 1 .294-.906l5.165-.755a2.122 2.122 0 0 0 1.597-1.16z";
+    const ImportanceStars = ({ value = 0, size = 13, className }) => {
+      const filled = Math.min(5, Math.max(0, Math.round(value || 0)));
+      return h("span", { className: `mneme-stars${className ? ` ${className}` : ""}`, role: "img", "aria-label": `${filled}/5` },
+        [0, 1, 2, 3, 4].map((i) => h("svg", { key: i, width: size, height: size, viewBox: "0 0 24 24", "aria-hidden": "true" },
+          h("path", {
+            d: STAR_PATH_D,
+            fill: i < filled ? "currentColor" : "none",
+            stroke: "currentColor",
+            strokeWidth: 1.6,
+            strokeLinejoin: "round",
+            opacity: i < filled ? 1 : 0.35
+          }))));
+    };
+
     // Unified API fetcher: attaches the optional apiToken (set in the settings
     // view, persisted in localStorage) as a Bearer header. When no token has
     // been configured the header is omitted and the API stays open (default).
@@ -695,7 +713,9 @@ window.__ModuleLoader__.load({
       ".mneme-vtab:hover{color:var(--dsw-alias-label-primary)}",
       ".mneme-vtab.mneme-active{color:var(--dsw-alias-state-business-primary)}",
       ".mneme-vtab.mneme-active::after{content:\"\";position:absolute;left:10px;right:10px;bottom:-1px;height:2.5px;border-radius:2px;background:var(--dsw-alias-state-business-primary)}",
-      ".mneme-xtools{position:absolute;right:14px;top:50%;transform:translateY(-50%);display:flex;align-items:center;gap:8px;padding:0}",
+      // transform 会自建层叠上下文：层级必须给在容器上——3 高于吸顶月份头(2)、
+      // 低于详情抽屉(6)，下拉菜单/导入对话框随容器整体上浮。
+      ".mneme-xtools{position:absolute;right:14px;top:50%;transform:translateY(-50%);display:flex;align-items:center;gap:8px;padding:0;z-index:3}",
       ".mneme-xcount{flex:none;font-size:12px;line-height:16px;color:var(--dsw-alias-label-tertiary);white-space:nowrap}",
       // --- three-column browse layout: hairline separators, no outer box ---
       ".mneme-xmain{flex:1;min-height:0;display:flex;flex-direction:row;overflow:hidden}",
@@ -824,10 +844,13 @@ window.__ModuleLoader__.load({
       // --- 侧边栏顶部入口：借宿主「新会话」按钮的原生类名对齐 ---
       // wrapper display:contents 隐身，按钮成为侧边栏弹性布局的直接子元素；
       // 盒模型/间距/收起态 rail 几何全部继承宿主，我们只覆盖配色为次级观感。
+      ".mneme-stars{display:inline-flex;align-items:center;gap:2px}",
+      // width:100% 与「新会话」同宽（同一弹性父容器）；文案对齐交给复制来的
+      // 原生类，不再用 flex:1/text-align:left 覆盖成左对齐。
       ".mneme-topentry{display:contents}",
-      ".mneme-topentry-native{background:var(--dsw-alias-bg-layer-1);border:1px solid var(--dsw-alias-border-l2);color:var(--dsw-alias-label-primary)}",
+      ".mneme-topentry-native{width:100%;background:var(--dsw-alias-bg-layer-1);border:1px solid var(--dsw-alias-border-l2);color:var(--dsw-alias-label-primary)}",
       ".mneme-topentry-native:hover{background:var(--dsw-alias-interactive-bg-hover);color:var(--dsw-alias-label-primary)}",
-      ".mneme-topentry-native .mneme-topentry-label{flex:1;min-width:0;text-align:left;white-space:nowrap;overflow:hidden;text-overflow:ellipsis}",
+      ".mneme-topentry-native .mneme-topentry-label{white-space:nowrap}",
       // --- 功能开关：一行一开关，Claude 式安静排版 ---
       ".mneme-featgroup{flex:none;font-size:12px;font-weight:600;letter-spacing:.02em;color:var(--dsw-alias-label-tertiary);margin:16px 0 2px}",
       ".mneme-featrow{display:flex;align-items:center;gap:14px;padding:11px 2px;border-bottom:1px solid var(--dsw-alias-border-l1)}",
@@ -2560,7 +2583,7 @@ window.__ModuleLoader__.load({
                   value: importance,
                   onChange: (e) => setImportance(Number(e.target.value))
                 }, [1, 2, 3, 4, 5].map((n) => h("option", { key: n, value: n }, "★".repeat(n))))
-              : h("span", { className: "mneme-dmetaval" }, "★".repeat(Math.min(5, Math.max(0, memory.importance || 0)))),
+              : h(ImportanceStars, { className: "mneme-dmetaval", value: memory.importance || 0 }),
             memory.source && h(react.Fragment, null,
               h("span", { className: "mneme-dmetakey" }, t("memory.explorer.source")),
               h("span", { className: "mneme-dmetaval", title: memory.source }, memory.source)),
@@ -3160,7 +3183,7 @@ window.__ModuleLoader__.load({
                         h("div", { className: "mneme-cardtitle" }, m.title || (m.content || "").slice(0, 60)),
                         h("div", { className: "mneme-cardexcerpt" }, m.content || ""),
                         h("div", { className: "mneme-cardfoot" },
-                          h("span", null, "★".repeat(Math.min(5, Math.max(0, m.importance || 0)))),
+                          h(ImportanceStars, { value: m.importance || 0, size: 12 }),
                           m.source && h("span", { className: "mneme-cardsrc", title: m.source }, m.source)
                         )
                       )),
@@ -3287,19 +3310,34 @@ window.__ModuleLoader__.load({
       const [nativeCls, setNativeCls] = useState("");
       useEffect(() => {
         if (!reactDom || typeof document === "undefined") return undefined;
-        let tries = 0, timer = null, created = null;
+        let tries = 0, timer = null, created = null, mo = null;
+        const findNative = (region) =>
+          region.parentElement.querySelector('[class*="newSession"]')
+          || region.previousElementSibling;
         const attempt = () => {
           const region = document.querySelector('[class*="regionArea"]');
           if (region && region.parentElement) {
             // 「新会话」按钮与 regionArea 同级且紧邻其前，借它的类名获得
             // 与原生侧边栏项完全一致的盒模型（含收起态 rail 几何）。
-            const ns = region.parentElement.querySelector('[class*="newSession"]')
-              || region.previousElementSibling;
+            // 宿主/皮肤会异步改写按钮类名，用 MutationObserver 持续同步，
+            // 否则 portal 按钮会停留在捕获时刻的旧类上（宽度/对齐失配）。
             created = document.createElement("div");
             created.dataset.pluginEntry = "@modusensus/dsh-mneme";
             region.parentElement.insertBefore(created, region);
             setHost(created);
-            setNativeCls(ns?.className || "");
+            setNativeCls(findNative(region)?.className || "");
+            mo = new MutationObserver(() => {
+              const cur = document.querySelector('[class*="regionArea"]');
+              if (!cur || !cur.parentElement) return;
+              const cls = findNative(cur)?.className || "";
+              setNativeCls((prev) => (prev === cls ? prev : cls));
+            });
+            mo.observe(region.parentElement, {
+              attributes: true,
+              attributeFilter: ["class"],
+              childList: true,
+              subtree: true
+            });
             return;
           }
           if (++tries > 40) return; // 放弃 portal，footer 回退保持可用
@@ -3308,6 +3346,7 @@ window.__ModuleLoader__.load({
         attempt();
         return () => {
           clearTimeout(timer);
+          if (mo) mo.disconnect();
           if (created && created.parentElement) created.parentElement.removeChild(created);
         };
       }, []);

package/lib/dream/sleep.js

@@ -17,7 +17,7 @@
 import { randomUUID, createHash } from "node:crypto";
 import { validateDecisions, applyDecisions } from "./decisions.js";
 import { findPotentialConflicts } from "./clustering.js";
-import { buildReceipt } from "../dream.js";
+import { buildReceipt, withEffortFallback } from "../dream.js";
 
 const SUMMARY_MAX = 120;
 // Conflict similarity threshold per strictness level (v0.4.0):
@@ -72,16 +72,18 @@ async function streamText(ctx, options) {
   return text;
 }
 
-/** LLM route: agent default model first, then sleepProvider/Model, then the
- *  dream route as a shared fallback. Sleep can pin a cheaper model for its
- *  bulk passes without disturbing the dream route. */
+/** LLM route (Issue #25): explicit sleepProvider/Model wins, then the dream
+ *  route as a shared explicit fallback, then the agent default model. Sleep
+ *  can pin a cheaper model for its bulk passes without disturbing the dream
+ *  route. Explicit config first — otherwise the config routes are dead code
+ *  whenever agentDefaultModel resolves (see resolveRoute in dream.js). */
 function resolveSleepRoute(ctx, config, logger) {
+  if (config.sleepProvider && config.sleepModel) return { provider: config.sleepProvider, model: config.sleepModel };
+  if (config.dreamProvider && config.dreamModel) return { provider: config.dreamProvider, model: config.dreamModel };
   try {
     const sel = ctx?.agentDefaultModel?.currentSelection?.();
     if (sel?.provider && sel?.model) return { provider: sel.provider, model: sel.model };
-  } catch { /* fall through to config route */ }
-  if (config.sleepProvider && config.sleepModel) return { provider: config.sleepProvider, model: config.sleepModel };
-  if (config.dreamProvider && config.dreamModel) return { provider: config.dreamProvider, model: config.dreamModel };
+  } catch { /* fall through to warn */ }
   logger?.warn?.("dsh-mneme sleep: no llm route available");
   return undefined;
 }
@@ -182,19 +184,19 @@ async function phaseConflicts(ctx, service, config, logger, runId, semantic = nu
   const listText = selected.map((p) =>
     `候选冲突：\nid=${p.a.id} | type=${p.a.type} | title=${p.a.title}\n${p.a.content}\n---\nid=${p.b.id} | type=${p.b.type} | title=${p.b.title}\n${p.b.content}\n（相似度 ${p.similarity.toFixed(2)}）`
   ).join("\n\n");
-  const text = await streamText(ctx, {
+  const sleepEffort = config.sleepReasoningEffort && config.sleepReasoningEffort !== "none" ? config.sleepReasoningEffort : null;
+  const runConflict = (withEffort) => streamText(ctx, {
     provider: route.provider,
     model: route.model,
     purpose: "sleep-conflict",
     maxTokens: 2048,
-    ...(config.sleepReasoningEffort && config.sleepReasoningEffort !== "none"
-      ? { reasoningEffort: config.sleepReasoningEffort }
-      : {}),
+    ...(withEffort && sleepEffort ? { reasoningEffort: sleepEffort } : {}),
     messages: [
       { role: "system", content: [{ type: "text", text: CONFLICT_PROMPT }] },
       { role: "user", content: [{ type: "text", text: listText }] }
     ]
   });
+  const text = await withEffortFallback(ctx, sleepEffort, () => runConflict(true), () => runConflict(false));
   if (text === undefined) return { status: "failed", error: "llm failed" };
   const decisions = parseJsonArray(text);
   if (!decisions) return { status: "failed", error: "invalid decisions json" };
@@ -282,19 +284,19 @@ async function phasePatterns(ctx, service, config, logger, runId, signal = null)
     .map((m) => `id=${m.id} | type=${m.type} | importance=${m.importance} | updated=${m.updated_at} | title=${m.title} | content=${m.content}`)
     .join("\n");
   const maxPatterns = config.sleepMaxPatternPerRun ?? 3;
-  const text = await streamText(ctx, {
+  const sleepEffort = config.sleepReasoningEffort && config.sleepReasoningEffort !== "none" ? config.sleepReasoningEffort : null;
+  const runPattern = (withEffort) => streamText(ctx, {
     provider: route.provider,
     model: route.model,
     purpose: "sleep-pattern",
     maxTokens: 2048,
-    ...(config.sleepReasoningEffort && config.sleepReasoningEffort !== "none"
-      ? { reasoningEffort: config.sleepReasoningEffort }
-      : {}),
+    ...(withEffort && sleepEffort ? { reasoningEffort: sleepEffort } : {}),
     messages: [
       { role: "system", content: [{ type: "text", text: PATTERN_PROMPT.replace("N", String(maxPatterns)) }] },
       { role: "user", content: [{ type: "text", text: listText }] }
     ]
   });
+  const text = await withEffortFallback(ctx, sleepEffort, () => runPattern(true), () => runPattern(false));
   if (text === undefined) return { status: "failed", error: "llm failed" };
   const decisions = parseJsonArray(text);
   if (!decisions || decisions.length === 0) return { status: "skipped", reason: "no patterns found" };