@modusensus/dsh-mneme 0.7.15 → 0.7.16

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-648%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.
 
@@ -405,7 +405,7 @@ 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/                 # 648 个 node:test 测试（含审计与三轴线压测不变量）
 scripts/              # e2e-dsh.js 端到端演示 · stress-dsh.js 三轴线压测 · sync-lib.js 同步
 ```
 

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-648%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)
@@ -241,6 +241,7 @@ v0.3.0 起新增**记忆基因**层：从记忆里抽取**命名实体**、**带
 
 | 版本 | 亮点 |
 |------|------|
+| **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；648 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 |
@@ -544,7 +545,7 @@ src/
 ├── 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）
+test/                 # 648 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
 ```
 
@@ -553,7 +554,7 @@ 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 645 个测试
+npm test           # 运行 648 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/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" };

package/lib/dream.js

@@ -1,7 +1,7 @@
 import { validateDecisions, applyDecisions } from "./dream/decisions.js";
 import { clusterMemories, findPotentialConflicts } from "./dream/clustering.js";
 import { createHash, randomUUID } from "node:crypto";
-export { validateDecisions, applyDecisions };
+export { validateDecisions, applyDecisions, withEffortFallback };
 
 
 // Extract the first JSON array from LLM output, tolerating markdown fences,
@@ -282,6 +282,15 @@ async function runAuditedLlm(ctx, service, config, spec, body) {
       // record it as error here so the audit shows the truth.
       status = "error";
       errorMessage = errorMessage ?? "llm stream aborted or errored";
+    } else if (typeof spec.auditError === "function") {
+      // A stream that returned text but yields nothing usable is still a
+      // failed call — record it as error, not the default success, so the
+      // audit no longer contradicts a failed run (dream "no json array").
+      const message = spec.auditError(result);
+      if (message) {
+        status = "error";
+        errorMessage = message;
+      }
     }
     return result;
   } catch (error) {
@@ -311,20 +320,50 @@ async function runAuditedLlm(ctx, service, config, spec, body) {
 }
 
 /**
- * Resolve the LLM route: agent default model (deployment) first, plugin config
- * (dreamProvider/dreamModel) as fallback. Falls through to undefined when no
- * route exists — runDream then fails safe. Fallback is logged so a silent
- * route switch is observable.
+ * Reasoning-effort rejection fallback (v0.8.1): a configured dreamReasoningEffort
+ * / sleepReasoningEffort may be rejected by the provider (volcano-engine returns
+ * UNSUPPORTED_REASONING_EFFORT for values it does not accept — "off" is known
+ * rejected there). When that happens, retry once WITHOUT the reasoning field
+ * instead of hard-failing the run, so effort config is safe to experiment with:
+ * accepted → reasoning capped; rejected → provider default (old behavior),
+ * logged so the rejection is observable.
+ */
+async function withEffortFallback(ctx, effort, attempt, fallback) {
+  if (!effort || effort === "none") return attempt();
+  try {
+    return await attempt();
+  } catch (error) {
+    const message = String(error?.message ?? error);
+    // matches both "reasoning effort" (natural language) and the bare
+    // "UNSUPPORTED_REASONING_EFFORT" error code (underscore).
+    if (!/reasoning[\s_]*effort/i.test(message)) throw error;
+    ctx.logger?.warn?.(`dsh-mneme dream: reasoningEffort "${effort}" rejected (${message}); retrying without it`);
+    return fallback();
+  }
+}
+
+/**
+ * Resolve the LLM route (Issue #25): an explicit plugin config
+ * (dreamProvider/dreamModel) is the user's declared override and wins; the
+ * agent default model (deployment) is only a fallback when no config route is
+ * set. In a standard DSH install agentDefaultModel always resolves, so without
+ * this ordering the config route would be dead code and dreamProvider/dreamModel
+ * could never take effect (v0.7.11 regressed this; README §config documents
+ * config-first). Falls through to undefined when no route exists — runDream
+ * then fails safe. A config→default switch is logged so it is observable.
  */
 function resolveRoute(ctx, config, logger) {
+  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 };
-    logger?.warn?.("dsh-mneme dream: agentDefaultModel unavailable, falling back to config route");
+    if (sel?.provider && sel?.model) {
+      logger?.info?.("dsh-mneme dream: no dreamProvider/dreamModel config, falling back to agent default");
+      return { provider: sel.provider, model: sel.model };
+    }
+    logger?.warn?.("dsh-mneme dream: agentDefaultModel unavailable, no config route either");
   } catch (error) {
-    logger?.warn?.(`dsh-mneme dream: agentDefaultModel lookup failed, falling back to config route: ${String(error)}`);
+    logger?.warn?.(`dsh-mneme dream: agentDefaultModel lookup failed: ${String(error)}`);
   }
-  if (config.dreamProvider && config.dreamModel) return { provider: config.dreamProvider, model: config.dreamModel };
   return undefined;
 }
 
@@ -585,28 +624,37 @@ export function createDreamScheduler({ onRun, thresholdCount = 10, thresholdChar
       ? CONSOLIDATION_PROMPT + `\n\n当前为「冲突冻结」模式：检测到内容矛盾的条目时，仍请输出 conflict，并以 winner/loser 作为候选、reason 说明理由；冲突不会被自动裁决，而会冻结待人工确认。`
       : CONSOLIDATION_PROMPT;
     let decisionText;
+    // 加固（v0.8.1）：配置的 reasoningEffort 被 provider 拒收时回退重试一次
+    // （不带该字段），避免 thinking 模型配置 low/medium 直接整单失败。解析放
+    // 在 auditError 检查器里、闭包交回主流程，避免二次解析；解析失败同时如实
+    // 记 audit error 并在日志带原始输出前 300 字节，便于定位"推理吞预算返回空体"。
+    const effort = config.dreamReasoningEffort && config.dreamReasoningEffort !== "none" ? config.dreamReasoningEffort : null;
+    let decisions = null;
+    const runConsolidation = (withEffort) => runAuditedLlm(ctx, service, config, {
+      triggerSource: "autoDream",
+      operationType: "dream_consolidate",
+      modelId: `${route.provider}:${route.model}`,
+      relatedMemoryIds: [...snapshot.keys()],
+      auditError: (text) => {
+        decisions = extractJsonArray(text);
+        return Array.isArray(decisions) ? null : "no json array in llm output";
+      }
+    }, (reportUsage) => streamText(ctx, {
+      provider: route.provider,
+      model: route.model,
+      purpose: "compaction",
+      maxTokens: config.dreamMaxTokens ?? 4096,
+      ...(withEffort && effort ? { reasoningEffort: effort } : {}),
+      messages: [
+        { role: "system", content: [{ type: "text", text: consolidationPrompt }] },
+        { role: "user", content: [{ type: "text", text: listText }] }
+      ]
+    }, reportUsage));
     try {
       // Bug8: the consolidation call is audited (tokens/time/status). A throw
       // re-propagates to the catch below; an aborted stream returns undefined
       // and is treated as a failed run after the check below.
-      decisionText = await runAuditedLlm(ctx, service, config, {
-        triggerSource: "autoDream",
-        operationType: "dream_consolidate",
-        modelId: `${route.provider}:${route.model}`,
-        relatedMemoryIds: [...snapshot.keys()]
-      }, (reportUsage) => streamText(ctx, {
-        provider: route.provider,
-        model: route.model,
-        purpose: "compaction",
-        maxTokens: config.dreamMaxTokens ?? 4096,
-        ...(config.dreamReasoningEffort && config.dreamReasoningEffort !== "none"
-          ? { reasoningEffort: config.dreamReasoningEffort }
-          : {}),
-        messages: [
-          { role: "system", content: [{ type: "text", text: consolidationPrompt }] },
-          { role: "user", content: [{ type: "text", text: listText }] }
-        ]
-      }, reportUsage));
+      decisionText = await withEffortFallback(ctx, effort, () => runConsolidation(true), () => runConsolidation(false));
     } catch (error) {
       logger?.warn?.(`dsh-mneme dream: consolidation llm call failed: ${String(error)}`);
       return finish({ ok: false, error: "llm failed", summary: false });
@@ -615,10 +663,9 @@ export function createDreamScheduler({ onRun, thresholdCount = 10, thresholdChar
       logger?.warn?.("dsh-mneme dream: consolidation llm stream aborted or errored");
       return finish({ ok: false, error: "llm failed", summary: false });
     }
-
-    const decisions = extractJsonArray(decisionText);
     if (!Array.isArray(decisions)) {
-      logger?.warn?.(`dsh-mneme dream: no json array in llm output (raw length ${decisionText?.length ?? 0})`);
+      const head = (decisionText ?? "").slice(0, 300).replace(/\s+/g, " ").trim();
+      logger?.warn?.(`dsh-mneme dream: no json array in llm output (raw length ${decisionText?.length ?? 0}; head: ${head})`);
       return finish({ ok: false, error: "no json array in llm output", summary: false });
     }
     const { ok, errors } = validateDecisions(decisions, snapshot, {
@@ -735,26 +782,25 @@ export function createDreamScheduler({ onRun, thresholdCount = 10, thresholdChar
     // Summary generation (second LLM call). A throwing stream is reported as
     // a failed run; summary:false marks a run that produced no summary.
     let summaryText;
+    const runSummary = (withEffort) => runAuditedLlm(ctx, service, config, {
+      triggerSource: "autoDream",
+      operationType: "dream_summarize",
+      modelId: `${route.provider}:${route.model}`,
+      relatedMemoryIds: []
+    }, (reportUsage) => streamText(ctx, {
+      provider: route.provider,
+      model: route.model,
+      purpose: "compaction",
+      maxTokens: config.dreamMaxTokens ?? 2048,
+      ...(withEffort && effort ? { reasoningEffort: effort } : {}),
+      messages: [
+        { role: "system", content: [{ type: "text", text: SUMMARY_PROMPT }] },
+        { role: "user", content: [{ type: "text", text: service.all().filter((m) => !m.archived && m.type !== "summary").map((m) => `- ${m.title}: ${m.content}`).join("\n") }] }
+      ]
+    }, reportUsage));
     try {
       // Bug8: the summary call is audited too (operation dream_summarize).
-      summaryText = await runAuditedLlm(ctx, service, config, {
-        triggerSource: "autoDream",
-        operationType: "dream_summarize",
-        modelId: `${route.provider}:${route.model}`,
-        relatedMemoryIds: []
-      }, (reportUsage) => streamText(ctx, {
-        provider: route.provider,
-        model: route.model,
-        purpose: "compaction",
-        maxTokens: config.dreamMaxTokens ?? 2048,
-        ...(config.dreamReasoningEffort && config.dreamReasoningEffort !== "none"
-          ? { reasoningEffort: config.dreamReasoningEffort }
-          : {}),
-        messages: [
-          { role: "system", content: [{ type: "text", text: SUMMARY_PROMPT }] },
-          { role: "user", content: [{ type: "text", text: service.all().filter((m) => !m.archived && m.type !== "summary").map((m) => `- ${m.title}: ${m.content}`).join("\n") }] }
-        ]
-      }, reportUsage));
+      summaryText = await withEffortFallback(ctx, effort, () => runSummary(true), () => runSummary(false));
     } catch (error) {
       logger?.warn?.(`dsh-mneme dream: summary llm call failed: ${String(error)}`);
       return finish({ ok: false, error: "llm failed", applied, decisions: auditDecisions, outcome, frozen: frozenCount, summary: false });

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@modusensus/dsh-mneme",
   "description": "Cross-session memory plugin for DeepSeek Harness with autoDream consolidation: SQLite store, Markdown mirrors, 7 model tools, automatic injection, session summarization, user profile/rules, custom slash commands, vector (semantic) search, and a Web GUI panel",
-  "version": "0.7.15",
+  "version": "0.7.16",
   "license": "MIT",
   "repository": {
     "type": "git",

package/src/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" };

package/src/dream.js

@@ -1,7 +1,7 @@
 import { validateDecisions, applyDecisions } from "./dream/decisions.js";
 import { clusterMemories, findPotentialConflicts } from "./dream/clustering.js";
 import { createHash, randomUUID } from "node:crypto";
-export { validateDecisions, applyDecisions };
+export { validateDecisions, applyDecisions, withEffortFallback };
 
 
 // Extract the first JSON array from LLM output, tolerating markdown fences,
@@ -282,6 +282,15 @@ async function runAuditedLlm(ctx, service, config, spec, body) {
       // record it as error here so the audit shows the truth.
       status = "error";
       errorMessage = errorMessage ?? "llm stream aborted or errored";
+    } else if (typeof spec.auditError === "function") {
+      // A stream that returned text but yields nothing usable is still a
+      // failed call — record it as error, not the default success, so the
+      // audit no longer contradicts a failed run (dream "no json array").
+      const message = spec.auditError(result);
+      if (message) {
+        status = "error";
+        errorMessage = message;
+      }
     }
     return result;
   } catch (error) {
@@ -311,20 +320,50 @@ async function runAuditedLlm(ctx, service, config, spec, body) {
 }
 
 /**
- * Resolve the LLM route: agent default model (deployment) first, plugin config
- * (dreamProvider/dreamModel) as fallback. Falls through to undefined when no
- * route exists — runDream then fails safe. Fallback is logged so a silent
- * route switch is observable.
+ * Reasoning-effort rejection fallback (v0.8.1): a configured dreamReasoningEffort
+ * / sleepReasoningEffort may be rejected by the provider (volcano-engine returns
+ * UNSUPPORTED_REASONING_EFFORT for values it does not accept — "off" is known
+ * rejected there). When that happens, retry once WITHOUT the reasoning field
+ * instead of hard-failing the run, so effort config is safe to experiment with:
+ * accepted → reasoning capped; rejected → provider default (old behavior),
+ * logged so the rejection is observable.
+ */
+async function withEffortFallback(ctx, effort, attempt, fallback) {
+  if (!effort || effort === "none") return attempt();
+  try {
+    return await attempt();
+  } catch (error) {
+    const message = String(error?.message ?? error);
+    // matches both "reasoning effort" (natural language) and the bare
+    // "UNSUPPORTED_REASONING_EFFORT" error code (underscore).
+    if (!/reasoning[\s_]*effort/i.test(message)) throw error;
+    ctx.logger?.warn?.(`dsh-mneme dream: reasoningEffort "${effort}" rejected (${message}); retrying without it`);
+    return fallback();
+  }
+}
+
+/**
+ * Resolve the LLM route (Issue #25): an explicit plugin config
+ * (dreamProvider/dreamModel) is the user's declared override and wins; the
+ * agent default model (deployment) is only a fallback when no config route is
+ * set. In a standard DSH install agentDefaultModel always resolves, so without
+ * this ordering the config route would be dead code and dreamProvider/dreamModel
+ * could never take effect (v0.7.11 regressed this; README §config documents
+ * config-first). Falls through to undefined when no route exists — runDream
+ * then fails safe. A config→default switch is logged so it is observable.
  */
 function resolveRoute(ctx, config, logger) {
+  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 };
-    logger?.warn?.("dsh-mneme dream: agentDefaultModel unavailable, falling back to config route");
+    if (sel?.provider && sel?.model) {
+      logger?.info?.("dsh-mneme dream: no dreamProvider/dreamModel config, falling back to agent default");
+      return { provider: sel.provider, model: sel.model };
+    }
+    logger?.warn?.("dsh-mneme dream: agentDefaultModel unavailable, no config route either");
   } catch (error) {
-    logger?.warn?.(`dsh-mneme dream: agentDefaultModel lookup failed, falling back to config route: ${String(error)}`);
+    logger?.warn?.(`dsh-mneme dream: agentDefaultModel lookup failed: ${String(error)}`);
   }
-  if (config.dreamProvider && config.dreamModel) return { provider: config.dreamProvider, model: config.dreamModel };
   return undefined;
 }
 
@@ -585,28 +624,37 @@ export function createDreamScheduler({ onRun, thresholdCount = 10, thresholdChar
       ? CONSOLIDATION_PROMPT + `\n\n当前为「冲突冻结」模式：检测到内容矛盾的条目时，仍请输出 conflict，并以 winner/loser 作为候选、reason 说明理由；冲突不会被自动裁决，而会冻结待人工确认。`
       : CONSOLIDATION_PROMPT;
     let decisionText;
+    // 加固（v0.8.1）：配置的 reasoningEffort 被 provider 拒收时回退重试一次
+    // （不带该字段），避免 thinking 模型配置 low/medium 直接整单失败。解析放
+    // 在 auditError 检查器里、闭包交回主流程，避免二次解析；解析失败同时如实
+    // 记 audit error 并在日志带原始输出前 300 字节，便于定位"推理吞预算返回空体"。
+    const effort = config.dreamReasoningEffort && config.dreamReasoningEffort !== "none" ? config.dreamReasoningEffort : null;
+    let decisions = null;
+    const runConsolidation = (withEffort) => runAuditedLlm(ctx, service, config, {
+      triggerSource: "autoDream",
+      operationType: "dream_consolidate",
+      modelId: `${route.provider}:${route.model}`,
+      relatedMemoryIds: [...snapshot.keys()],
+      auditError: (text) => {
+        decisions = extractJsonArray(text);
+        return Array.isArray(decisions) ? null : "no json array in llm output";
+      }
+    }, (reportUsage) => streamText(ctx, {
+      provider: route.provider,
+      model: route.model,
+      purpose: "compaction",
+      maxTokens: config.dreamMaxTokens ?? 4096,
+      ...(withEffort && effort ? { reasoningEffort: effort } : {}),
+      messages: [
+        { role: "system", content: [{ type: "text", text: consolidationPrompt }] },
+        { role: "user", content: [{ type: "text", text: listText }] }
+      ]
+    }, reportUsage));
     try {
       // Bug8: the consolidation call is audited (tokens/time/status). A throw
       // re-propagates to the catch below; an aborted stream returns undefined
       // and is treated as a failed run after the check below.
-      decisionText = await runAuditedLlm(ctx, service, config, {
-        triggerSource: "autoDream",
-        operationType: "dream_consolidate",
-        modelId: `${route.provider}:${route.model}`,
-        relatedMemoryIds: [...snapshot.keys()]
-      }, (reportUsage) => streamText(ctx, {
-        provider: route.provider,
-        model: route.model,
-        purpose: "compaction",
-        maxTokens: config.dreamMaxTokens ?? 4096,
-        ...(config.dreamReasoningEffort && config.dreamReasoningEffort !== "none"
-          ? { reasoningEffort: config.dreamReasoningEffort }
-          : {}),
-        messages: [
-          { role: "system", content: [{ type: "text", text: consolidationPrompt }] },
-          { role: "user", content: [{ type: "text", text: listText }] }
-        ]
-      }, reportUsage));
+      decisionText = await withEffortFallback(ctx, effort, () => runConsolidation(true), () => runConsolidation(false));
     } catch (error) {
       logger?.warn?.(`dsh-mneme dream: consolidation llm call failed: ${String(error)}`);
       return finish({ ok: false, error: "llm failed", summary: false });
@@ -615,10 +663,9 @@ export function createDreamScheduler({ onRun, thresholdCount = 10, thresholdChar
       logger?.warn?.("dsh-mneme dream: consolidation llm stream aborted or errored");
       return finish({ ok: false, error: "llm failed", summary: false });
     }
-
-    const decisions = extractJsonArray(decisionText);
     if (!Array.isArray(decisions)) {
-      logger?.warn?.(`dsh-mneme dream: no json array in llm output (raw length ${decisionText?.length ?? 0})`);
+      const head = (decisionText ?? "").slice(0, 300).replace(/\s+/g, " ").trim();
+      logger?.warn?.(`dsh-mneme dream: no json array in llm output (raw length ${decisionText?.length ?? 0}; head: ${head})`);
       return finish({ ok: false, error: "no json array in llm output", summary: false });
     }
     const { ok, errors } = validateDecisions(decisions, snapshot, {
@@ -735,26 +782,25 @@ export function createDreamScheduler({ onRun, thresholdCount = 10, thresholdChar
     // Summary generation (second LLM call). A throwing stream is reported as
     // a failed run; summary:false marks a run that produced no summary.
     let summaryText;
+    const runSummary = (withEffort) => runAuditedLlm(ctx, service, config, {
+      triggerSource: "autoDream",
+      operationType: "dream_summarize",
+      modelId: `${route.provider}:${route.model}`,
+      relatedMemoryIds: []
+    }, (reportUsage) => streamText(ctx, {
+      provider: route.provider,
+      model: route.model,
+      purpose: "compaction",
+      maxTokens: config.dreamMaxTokens ?? 2048,
+      ...(withEffort && effort ? { reasoningEffort: effort } : {}),
+      messages: [
+        { role: "system", content: [{ type: "text", text: SUMMARY_PROMPT }] },
+        { role: "user", content: [{ type: "text", text: service.all().filter((m) => !m.archived && m.type !== "summary").map((m) => `- ${m.title}: ${m.content}`).join("\n") }] }
+      ]
+    }, reportUsage));
     try {
       // Bug8: the summary call is audited too (operation dream_summarize).
-      summaryText = await runAuditedLlm(ctx, service, config, {
-        triggerSource: "autoDream",
-        operationType: "dream_summarize",
-        modelId: `${route.provider}:${route.model}`,
-        relatedMemoryIds: []
-      }, (reportUsage) => streamText(ctx, {
-        provider: route.provider,
-        model: route.model,
-        purpose: "compaction",
-        maxTokens: config.dreamMaxTokens ?? 2048,
-        ...(config.dreamReasoningEffort && config.dreamReasoningEffort !== "none"
-          ? { reasoningEffort: config.dreamReasoningEffort }
-          : {}),
-        messages: [
-          { role: "system", content: [{ type: "text", text: SUMMARY_PROMPT }] },
-          { role: "user", content: [{ type: "text", text: service.all().filter((m) => !m.archived && m.type !== "summary").map((m) => `- ${m.title}: ${m.content}`).join("\n") }] }
-        ]
-      }, reportUsage));
+      summaryText = await withEffortFallback(ctx, effort, () => runSummary(true), () => runSummary(false));
     } catch (error) {
       logger?.warn?.(`dsh-mneme dream: summary llm call failed: ${String(error)}`);
       return finish({ ok: false, error: "llm failed", applied, decisions: auditDecisions, outcome, frozen: frozenCount, summary: false });

package/test/dream.test.js

@@ -870,7 +870,7 @@ test("Bug8: runDream records llm_audit_logs rows for consolidation and summary",
   assert.deepEqual(summarize.related_memory_ids, [], "summary audit has no related ids");
   for (const row of rows) {
     assert.equal(row.status, "success");
-    assert.equal(row.model_id, "mock:stress-model");
+    assert.equal(row.model_id, "deepseek:deepseek-chat", "config-first route (Issue #25): dreamProvider/dreamModel wins");
     assert.ok(Number.isInteger(row.duration_ms) && row.duration_ms >= 0, "duration recorded");
     assert.equal(row.input_tokens, 0);
     assert.equal(row.output_tokens, 0);

package/test/llm-audit.test.js

@@ -50,10 +50,11 @@ test("autoDream writes llm_audit_logs rows for consolidation and summary", async
   assert.equal(summarize.trigger_source, "autoDream");
   assert.equal(consolidate.status, "success");
   assert.equal(summarize.status, "success");
-  // mockCtx resolves the route from agentDefaultModel (mock:stress-model), not
-  // the config fallback — assert the actually-used route.
-  assert.equal(consolidate.model_id, "mock:stress-model");
-  assert.equal(summarize.model_id, "mock:stress-model");
+  // config-first (Issue #25): dreamSetup sets dreamProvider/dreamModel, so it
+  // wins over mockCtx's agentDefaultModel (mock:stress-model) — assert the
+  // actually-used config route.
+  assert.equal(consolidate.model_id, "deepseek:deepseek-chat");
+  assert.equal(summarize.model_id, "deepseek:deepseek-chat");
   assert.ok(Array.isArray(consolidate.related_memory_ids) && consolidate.related_memory_ids.length === 2,
     "consolidation audit records the snapshot ids");
   assert.ok(consolidate.total_tokens >= 0 && summarize.total_tokens >= 0);
@@ -110,6 +111,35 @@ test("autoDream throwing LLM is audited as status=error", async () => {
   store.close();
 });
 
+test("autoDream parse failure is audited as status=error, not fake success", async () => {
+  const { store, service, config } = dreamSetup();
+  service.saveWithDedupe({ type: "project", title: "主题", content: "内容" });
+  // Stream completes fine but returns no JSON array — the run fails, and the
+  // audit must NOT claim success (the old bug: llm_audit said dream_consolidate
+  // success while the run recorded failed).
+  const warnings = [];
+  const ctx = {
+    logger: { warn: (m) => warnings.push(m) },
+    llm: {
+      async *stream() {
+        yield { type: "text-delta", index: 0, text: "抱歉，我无法解析成 JSON。" };
+        yield { type: "finish", reason: { kind: "stop" } };
+      }
+    }
+  };
+  const dream = createDreamScheduler({ thresholdCount: 1, thresholdChars: 0, delayMs: 0 });
+  const result = await dream.runDream(ctx, service, config);
+  assert.equal(result.ok, false);
+  assert.equal(result.error, "no json array in llm output");
+  const rows = service.listLlmAudits();
+  assert.equal(rows.length, 1, "failed consolidation still audited");
+  assert.equal(rows[0].operation_type, "dream_consolidate");
+  assert.equal(rows[0].status, "error", "stream succeeded but output unusable → audit error");
+  assert.match(rows[0].error_message, /no json array in llm output/);
+  assert.ok(warnings.some((m) => m.includes("head: 抱歉，我无法解析成")), "raw output head logged for diagnosis");
+  store.close();
+});
+
 test("autoDream audit is skipped when llmAudit.enabled === false", async () => {
   const { store, service, config } = dreamSetup({ llmAudit: { enabled: false } });
   service.saveWithDedupe({ type: "project", title: "主题", content: "内容" });

package/test/reasoning-effort.test.js

@@ -107,6 +107,70 @@ test("issue#9: dream forwards dreamReasoningEffort on both LLM calls", async ()
   store.close();
 });
 
+test("issue#25: dreamProvider/dreamModel config wins over the agentDefaultModel route", async () => {
+  const store = createStore(":memory:");
+  const service = createService({ store, mirror: null, config: {} });
+  const dream = createDreamScheduler({ onRun: () => Promise.resolve({ ok: true, skipped: true }) });
+  const { memory: a } = service.saveWithDedupe({ type: "project", title: "插件", content: "旧", importance: 3 });
+  const { memory: b } = service.saveWithDedupe({ type: "project", title: "插件2", content: "新细节", importance: 4 });
+  const captured = [];
+  const ctx = dreamCtx({
+    captured,
+    onConsolidation: () => JSON.stringify([
+      { action: "merge", ids: [a.id, b.id], keepSource: b.id, title: "插件总览", content: "合并内容", importance: 4 }
+    ])
+  });
+  // dreamCtx's agentDefaultModel resolves (mock:mock-model), but the explicit
+  // config route must win — otherwise dreamProvider/dreamModel is dead code in
+  // a standard DSH install and the dream can never be moved off a thinking model.
+  const result = await dream.runDream(ctx, service, { dreamProvider: "volcano", dreamModel: "deepseek-v3" });
+  assert.equal(result.ok, true);
+  assert.ok(captured.length >= 2, "consolidation + summary both hit the LLM");
+  for (const options of captured) {
+    assert.equal(options.provider, "volcano");
+    assert.equal(options.model, "deepseek-v3", "config route wins over agentDefaultModel (mock:mock-model)");
+  }
+  store.close();
+});
+
+test("issue#9: rejected reasoningEffort retries once without it and still consolidates", async () => {
+  const store = createStore(":memory:");
+  const service = createService({ store, mirror: null, config: {} });
+  const dream = createDreamScheduler({ onRun: () => Promise.resolve({ ok: true, skipped: true }) });
+  const { memory: a } = service.saveWithDedupe({ type: "project", title: "插件", content: "旧", importance: 3 });
+  const { memory: b } = service.saveWithDedupe({ type: "project", title: "插件2", content: "新细节", importance: 4 });
+  const calls = [];
+  const ctx = {
+    logger: { warn: () => {} },
+    agentDefaultModel: { currentSelection: () => ({ provider: "mock", model: "mock-model" }) },
+    llm: {
+      async *stream(options) {
+        calls.push(options);
+        const userText = options.messages.find((m) => m.role === "user")?.content?.[0]?.text ?? "";
+        if (userText.startsWith("id=")) {
+          // First attempt forwards reasoningEffort: the provider rejects it.
+          if (options.reasoningEffort) {
+            throw new Error("UNSUPPORTED_REASONING_EFFORT: DeepSeek does not support reasoning effort \"low\"");
+          }
+          yield { type: "text-delta", index: 0, text: JSON.stringify([
+            { action: "merge", ids: [a.id, b.id], keepSource: b.id, title: "合并标题", content: "合并内容", importance: 4 }
+          ]) };
+        } else {
+          yield { type: "text-delta", index: 0, text: "记忆库总览：用户偏好中文。" };
+        }
+        yield { type: "finish", reason: { kind: "stop" } };
+      }
+    }
+  };
+  const result = await dream.runDream(ctx, service, { dreamReasoningEffort: "low" });
+  assert.equal(result.ok, true, "run survives the effort rejection via the fallback retry");
+  assert.ok(result.applied > 0, "consolidation still lands changes");
+  assert.equal(calls.length, 3, "consolidation tried (rejected) + retried without effort + summary");
+  assert.equal(calls[0].reasoningEffort, "low", "first consolidation attempt forwards the effort");
+  assert.equal("reasoningEffort" in calls[1], false, "retry omits the rejected effort field");
+  store.close();
+});
+
 // ---------------------------------------------------------------- sleep passthrough
 
 function sleepSetup() {