dsh-decision-layer 0.6.0 → 0.6.1

package/README.en.md

@@ -4,7 +4,7 @@
 
 `dsh-decision-layer` adds structured, explicitly scoped decisions to the DeepSeek Harness (DSH) agent loop, backed by a pluggable adjudication backend. Dangerous-call gating, output self-check, tool narrowing, and loop prevention hook into the agent loop as automatic interventions.
 
-> **0.6.0 is a preview release.** The plugin works, but how much real value the adjudication interventions add is still being explored — every threshold is an unvalidated heuristic default. Treat it as something to observe, not a proven guarantee of impact. See [ROADMAP.md](<ROADMAP.md>) for scope and progress.
+> **0.6.1 is a preview release.** The plugin works, but how much real value the adjudication interventions add is still being explored — every threshold is an unvalidated heuristic default. Treat it as something to observe, not a proven guarantee of impact. See [ROADMAP.md](<ROADMAP.md>) for scope and progress.
 
 ## Installation
 
@@ -50,4 +50,4 @@ Restart `dsh web` and refresh after installing or updating: the host and client
 
 ## Safety boundary and development
 
-Automatic intervention only subtracts: the gate only denies, narrowing only removes tools, and neither grants new permissions. The gate defers to native DSH approval on backend failure rather than allowing; self-check, narrowing, and loop-guard errors skip intervention. The redacted summary sent to the backend carries only tool name and command/path with secret redaction — never file content. Node.js `>=20.11`; `npm test` and `npm run build:client` run local checks. Rebuild and commit `lib/client.js` with any client change. 0.6.0 is a preview: the interventions work, but their real value and thresholds are still being explored — feedback on misjudgments and feel observed in the decision log is welcome.
+Automatic intervention only subtracts: the gate only denies, narrowing only removes tools, and neither grants new permissions. The gate defers to native DSH approval on backend failure rather than allowing; self-check, narrowing, and loop-guard errors skip intervention. The redacted summary sent to the backend carries only tool name and command/path with secret redaction — never file content. Node.js `>=20.11`; `npm test` and `npm run build:client` run local checks. Rebuild and commit `lib/client.js` with any client change. 0.6.1 is a preview: the interventions work, but their real value and thresholds are still being explored — feedback on misjudgments and feel observed in the decision log is welcome.

package/README.md

@@ -4,7 +4,7 @@
 
 DeepSeek Harness（DSH）的结构化裁决插件。它把有明确判断标准的问题交给同协议的裁决后端，供 agent 在需要时使用，并在 agent loop 上接入危险门控、产出自检、工具收窄和防循环等自动介入。
 
-> **0.6.0 是体验版。** 插件功能已可用，但裁决介入能带来多少实际价值仍在探索中——各决策点的阈值都是未经验证的启发式默认值，建议先以观察心态体验，别当成经过验证的效果保证。功能与后续进度以 [ROADMAP.md](<ROADMAP.md>) 为准。
+> **0.6.1 是体验版。** 插件功能已可用，但裁决介入能带来多少实际价值仍在探索中——各决策点的阈值都是未经验证的启发式默认值，建议先以观察心态体验，别当成经过验证的效果保证。功能与后续进度以 [ROADMAP.md](<ROADMAP.md>) 为准。
 
 ## 安装
 
@@ -54,4 +54,4 @@ dsh plugin --profile web add file:/绝对路径/dsh-decision-layer
 
 ## 开发
 
-服务端使用 Node.js `>=20.11`（`.mjs`，无需构建）；客户端用 TypeScript + esbuild 构建。`npm test` 跑本地测试，`npm run build:client` 生成 `lib/client.js`。改动客户端后需重建并连同 `lib/client.js` 一起提交，否则 GitHub 安装拿到的是旧 bundle。0.6.0 为体验版：介入能力已可用，但实际价值与阈值仍在探索，欢迎反馈决策日志观察到的误判与体感。
+服务端使用 Node.js `>=20.11`（`.mjs`，无需构建）；客户端用 TypeScript + esbuild 构建。`npm test` 跑本地测试，`npm run build:client` 生成 `lib/client.js`。改动客户端后需重建并连同 `lib/client.js` 一起提交，否则 GitHub 安装拿到的是旧 bundle。0.6.1 为体验版：介入能力已可用，但实际价值与阈值仍在探索，欢迎反馈决策日志观察到的误判与体感。

package/lib/client.js

@@ -183,7 +183,10 @@ var dictionaries = {
     analysisAccuracy: "\u5224\u65AD\u51C6\u786E\u5EA6",
     analysisAccuracyEmpty: "\u5C1A\u672A\u6807\u6CE8\u4EFB\u4F55\u51B3\u7B56\u3002\u5728\u4E0B\u65B9\u51B3\u7B56\u5217\u8868\u9010\u6761\u6807\u6CE8\u300C\u5224\u5BF9 / \u5224\u9519 / \u8BF4\u4E0D\u51C6\u300D\u540E\uFF0C\u8FD9\u91CC\u7ED9\u51FA\u6C47\u603B\u3002",
     analysisLogNote: "\u6807\u6CE8\u5728\u6B64\u4E0E\u4ECB\u5165\u9762\u677F\u540C\u6E90\uFF1A\u4EFB\u4E00\u5904\u6807\u6CE8\u90FD\u5199\u5165\u540C\u4E00\u4EFD\u65E5\u5FD7\u3001\u6309\u51B3\u7B56 id \u5173\u8054\u3001\u53D6\u6700\u65B0\u3002",
-    analysisLogTruncated: "\u672C\u4F1A\u8BDD\u6682\u65E0\u53EF\u5C55\u793A\u7684\u51B3\u7B56\u8BB0\u5F55\uFF08\u5185\u5B58\u65E5\u5FD7\u53EF\u80FD\u5DF2\u6EDA\u52A8\u6E05\u7406\uFF09\u3002"
+    analysisLogTruncated: "\u672C\u4F1A\u8BDD\u6682\u65E0\u53EF\u5C55\u793A\u7684\u51B3\u7B56\u8BB0\u5F55\u3002",
+    pagerPrev: "\u4E0A\u4E00\u9875",
+    pagerNext: "\u4E0B\u4E00\u9875",
+    pagerStatus: "\u7B2C {page}/{pages} \u9875 \xB7 \u5171 {total} \u6761"
   },
   en: {
     summary: "A danger-call gate, output self-check, and loop guard driven by a pluggable adjudication model.",
@@ -335,7 +338,10 @@ var dictionaries = {
     analysisAccuracy: "Decision accuracy",
     analysisAccuracyEmpty: "No decisions rated yet. Rate rows below as correct / wrong / unsure and a summary appears here.",
     analysisLogNote: "Ratings here share one source with the intervention panel: a rating on either surface writes to the same log, keyed by decision id, latest wins.",
-    analysisLogTruncated: "No decision rows to show for this session (the in-memory log may have rolled over)."
+    analysisLogTruncated: "No decision rows to show for this session.",
+    pagerPrev: "Prev",
+    pagerNext: "Next",
+    pagerStatus: "Page {page}/{pages} \xB7 {total} total"
   }
 };
 
@@ -931,9 +937,11 @@ function SessionPanel({ sessionId, t }) {
 // src/client/analysis-view.tsx
 var import_react3 = require("react");
 var import_jsx_runtime4 = require("react/jsx-runtime");
+var LOG_PAGE_SIZE = 100;
 function AnalysisView({ sessionId, t }) {
   const [analysis, setAnalysis] = (0, import_react3.useState)(null);
-  const [log, setLog] = (0, import_react3.useState)([]);
+  const [rows, setRows] = (0, import_react3.useState)(null);
+  const [logPage, setLogPage] = (0, import_react3.useState)(1);
   const [loaded, setLoaded] = (0, import_react3.useState)(false);
   const [message, setMessage] = (0, import_react3.useState)("");
   const { ratings, seed, annotate } = useAnnotations(sessionId, () => setMessage(t("annotateFailed")));
@@ -941,23 +949,31 @@ function AnalysisView({ sessionId, t }) {
     let alive = true;
     setMessage("");
     setLoaded(false);
-    void Promise.allSettled([
-      request(`logs?sessionId=${encodeURIComponent(sessionId)}`).then((value) => {
-        if (alive) {
-          setAnalysis(value);
-          seed(value);
-        }
-      }),
-      request(`log?sessionId=${encodeURIComponent(sessionId)}`).then((value) => {
-        if (alive) setLog(value.entries);
-      })
-    ]).finally(() => {
+    setLogPage(1);
+    setRows(null);
+    void request(`logs?sessionId=${encodeURIComponent(sessionId)}`).then((value) => {
+      if (alive) {
+        setAnalysis(value);
+        seed(value);
+      }
+    }).catch(() => {
+    }).finally(() => {
       if (alive) setLoaded(true);
     });
     return () => {
       alive = false;
     };
   }, [sessionId, seed]);
+  (0, import_react3.useEffect)(() => {
+    let alive = true;
+    void request(`logrows?sessionId=${encodeURIComponent(sessionId)}&page=${logPage}&pageSize=${LOG_PAGE_SIZE}`).then((value) => {
+      if (alive) setRows(value);
+    }).catch(() => {
+    });
+    return () => {
+      alive = false;
+    };
+  }, [sessionId, logPage]);
   const profile = analysis?.profile;
   const backtest = analysis?.trendBacktest;
   const annotations = analysis?.annotations;
@@ -1021,8 +1037,13 @@ function AnalysisView({ sessionId, t }) {
       ] }),
       /* @__PURE__ */ (0, import_jsx_runtime4.jsxs)("section", { className: "decision-analysis-section", children: [
         /* @__PURE__ */ (0, import_jsx_runtime4.jsx)("h3", { children: t("log") }),
-        log.length === 0 ? /* @__PURE__ */ (0, import_jsx_runtime4.jsx)("p", { className: "decision-empty", role: "status", children: t("analysisLogTruncated") }) : /* @__PURE__ */ (0, import_jsx_runtime4.jsxs)(import_jsx_runtime4.Fragment, { children: [
-          /* @__PURE__ */ (0, import_jsx_runtime4.jsx)(DecisionLogList, { entries: log, ratings, onRate: (id, rating) => void annotate(id, rating), t }),
+        !rows || rows.total === 0 ? /* @__PURE__ */ (0, import_jsx_runtime4.jsx)("p", { className: "decision-empty", role: "status", children: t("analysisLogTruncated") }) : /* @__PURE__ */ (0, import_jsx_runtime4.jsxs)(import_jsx_runtime4.Fragment, { children: [
+          /* @__PURE__ */ (0, import_jsx_runtime4.jsx)(DecisionLogList, { entries: rows.entries, ratings, onRate: (id, rating) => void annotate(id, rating), t, newestFirst: false }),
+          rows.pages > 1 && /* @__PURE__ */ (0, import_jsx_runtime4.jsxs)("nav", { className: "decision-pager", "aria-label": t("log"), children: [
+            /* @__PURE__ */ (0, import_jsx_runtime4.jsx)("button", { type: "button", disabled: rows.page <= 1, onClick: () => setLogPage((page) => Math.max(1, page - 1)), children: t("pagerPrev") }),
+            /* @__PURE__ */ (0, import_jsx_runtime4.jsx)("span", { className: "decision-pager-status", children: t("pagerStatus", { page: rows.page, pages: rows.pages, total: rows.total }) }),
+            /* @__PURE__ */ (0, import_jsx_runtime4.jsx)("button", { type: "button", disabled: rows.page >= rows.pages, onClick: () => setLogPage((page) => page + 1), children: t("pagerNext") })
+          ] }),
           /* @__PURE__ */ (0, import_jsx_runtime4.jsx)("p", { className: "decision-muted decision-analysis-note", children: t("analysisLogNote") })
         ] })
       ] })
@@ -1566,13 +1587,13 @@ var css = `
 .decision-dot-deny{background:#c0392b}
 .decision-card-foot{display:flex;justify-content:space-between;gap:1rem;margin-top:.7rem;padding-top:.6rem;border-top:1px solid var(--dsw-alias-border-l2,#eef1f3);font-size:.76rem;color:var(--dsw-alias-label-secondary,#59636e)}
 .decision-card-foot b{color:var(--dsw-alias-label-primary,#202124);font-variant-numeric:tabular-nums;margin-left:.3rem}
-.decision-card-stats{display:flex;gap:.6rem}
-.decision-stat{flex:1 1 0;display:flex;flex-direction:column;align-items:center;gap:.2rem;padding:.5rem .3rem;border-radius:10px;background:var(--dsw-alias-bg-layer-1,#f6f8fa)}
-.decision-stat-num{font-size:1.35rem;font-weight:700;line-height:1;font-variant-numeric:tabular-nums;color:var(--dsw-alias-label-primary,#202124)}
+.decision-card-stats{display:grid;grid-auto-flow:column;grid-auto-columns:1fr;gap:.6rem;align-items:stretch}
+.decision-stat{min-width:0;display:flex;flex-direction:column;align-items:center;justify-content:space-between;gap:.2rem;padding:.5rem .3rem;border-radius:10px;background:var(--dsw-alias-bg-layer-1,#f6f8fa)}
+.decision-stat-num{font-size:1.35rem;font-weight:700;line-height:1;white-space:nowrap;font-variant-numeric:tabular-nums;color:var(--dsw-alias-label-primary,#202124)}
 .decision-stat-pct{font-size:.74rem;font-weight:600;line-height:1;color:var(--dsw-alias-label-tertiary,#77818b);font-variant-numeric:tabular-nums}
 .decision-stat-warn{color:#c98a1e}
 .decision-stat-muted{color:var(--dsw-alias-label-tertiary,#9aa5b1)}
-.decision-stat-label{font-size:.72rem;color:var(--dsw-alias-label-secondary,#59636e)}
+.decision-stat-label{font-size:.72rem;line-height:1.25;text-align:center;color:var(--dsw-alias-label-secondary,#59636e)}
 .decision-log{list-style:none;margin:0;padding:0;display:grid;gap:.5rem;max-height:13rem;overflow:auto}
 .decision-log-item{display:flex;align-items:baseline;gap:.6rem;font-size:.8rem;padding-left:.6rem;border-left:2px solid var(--dsw-alias-border-l2,#dce1e6);line-height:1.5}
 .decision-log-item.decision-log-deny{border-left-color:#c0392b}
@@ -1693,9 +1714,16 @@ var css = `
 .decision-analysis-section{border-top:1px solid var(--dsw-alias-border-l2,#eceff1);margin-top:1.1rem;padding-top:1rem}
 .decision-analysis-section>h3{font-size:.92rem;margin:0 0 .6rem}
 .decision-analysis-cards{display:grid;grid-template-columns:repeat(auto-fit,minmax(13rem,1fr));gap:.7rem}
-.decision-analysis-cards .decision-card-stats{flex-wrap:wrap}
 .decision-analysis-trend{max-width:26rem}
 .decision-analysis-note{margin:.5rem 0 0}
+/* The analysis tab's decision log shares the page scroll (no inner scrollbar);
+   the modal panel's .decision-log keeps its own capped scroll via the base rule. */
+.decision-analysis .decision-log{max-height:none;overflow:visible}
+.decision-pager{display:flex;align-items:center;justify-content:center;gap:.8rem;margin-top:.8rem}
+.decision-pager button{border:1px solid var(--dsw-alias-border-l2,#ccd3da);border-radius:7px;background:transparent;color:var(--dsw-alias-label-secondary,#59636e);padding:.3rem .85rem;font:inherit;font-size:.8rem;cursor:pointer}
+.decision-pager button:hover:not(:disabled){background:var(--dsw-alias-bg-layer-1,#f2f4f6);color:var(--dsw-alias-label-primary,#202124)}
+.decision-pager button:disabled{opacity:.45;cursor:default}
+.decision-pager-status{font-size:.8rem;color:var(--dsw-alias-label-secondary,#59636e);font-variant-numeric:tabular-nums}
 @media (prefers-color-scheme:dark){
   .decision-analysis{color:var(--dsw-alias-label-primary,#e6eaee)}
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-decision-layer",
-  "version": "0.6.0",
+  "version": "0.6.1",
   "description": "A structured decision layer for the DeepSeek Harness agent loop: a danger-call gate, output self-check, tool narrowing, and loop guard backed by a pluggable adjudication model.",
   "keywords": [
     "dsh",

package/service/dsh.mjs

@@ -11,7 +11,7 @@ import { createTaskCompletion } from './task-completion.mjs';
 import { createLoopGuard } from './loop-guard.mjs';
 import { createLogStore } from './log-store.mjs';
 import { createStateStore } from './state-store.mjs';
-import { analyzeSessionLogs, reconstructSessionSeed } from './log-analyze.mjs';
+import { analyzeSessionLogs, reconstructSessionSeed, listSessionDecisions } from './log-analyze.mjs';
 
 export const name = 'dsh-decision-layer';
 export const inject = ['tools', 'skills'];
@@ -261,7 +261,8 @@ export async function apply(ctx, options = {}) {
       port,
       trendConfig: (await resolveConfig({ path })).checkSettings,
     });
-    const unregister = createManagerRoutes({ path, sessions, backend, logStore, analyze, bindHost: web.webServer.host }).map(route => web.webServer.register(route));
+    const listRows = (sessionId, { page, pageSize } = {}) => listSessionDecisions(sessionId, { port, page, pageSize });
+    const unregister = createManagerRoutes({ path, sessions, backend, logStore, analyze, listRows, bindHost: web.webServer.host }).map(route => web.webServer.register(route));
     return () => { unregister.forEach(dispose => dispose()); const store = logStore; logStore = undefined; void store?.dispose(); };
   });
 }

package/service/log-analyze.mjs

@@ -230,3 +230,27 @@ export async function reconstructSessionSeed(sessionId, { dir, port, trendConfig
   if (decisions.size === 0) return undefined;
   return buildSeed(decisions, { maxLog, trendConfig });
 }
+
+// Read one session's persisted decision rows and return a single newest-first
+// page. Backs the analysis tab's full decision log (the dock panel still shows
+// only the in-memory recent rows). Decisions are folded by id (latest wins) so a
+// re-logged id is not double-counted. Rows carry only the panel fields — tool
+// name, outcome, counts — never gate command/path, which is never logged. The
+// internal sessionId tag is stripped before the row leaves the server.
+export async function listSessionDecisions(sessionId, { dir, port, page = 1, pageSize = 100, readStream = createReadableStream } = {}) {
+  if (typeof sessionId !== 'string' || !sessionId.trim()) throw new Error('Invalid session id');
+  const size = Number.isSafeInteger(pageSize) && pageSize > 0 && pageSize <= 500 ? pageSize : 100;
+  const records = [];
+  for await (const record of readStream({ dir, port })) {
+    if (!isObject(record)) continue;
+    if (record.sessionId === sessionId) records.push(record);
+  }
+  const { decisions } = foldSession(records);
+  const rows = [...decisions.values()].sort((a, b) => (b.at ?? 0) - (a.at ?? 0)); // newest first
+  const total = rows.length;
+  const pages = Math.max(1, Math.ceil(total / size));
+  const current = Math.min(Math.max(1, Number.isSafeInteger(page) ? page : 1), pages);
+  const start = (current - 1) * size;
+  const entries = rows.slice(start, start + size).map(({ sessionId: _sessionId, ...rest }) => rest);
+  return { entries, total, page: current, pageSize: size, pages };
+}

package/service/manager-api.mjs

@@ -37,7 +37,7 @@ async function readBody(req) {
   return value;
 }
 
-export function createManagerRoutes({ path, sessions, backend, logStore, analyze, env = process.env, bindHost = '127.0.0.1' }) {
+export function createManagerRoutes({ path, sessions, backend, logStore, analyze, listRows, env = process.env, bindHost = '127.0.0.1' }) {
   const route = (suffix, methods) => ({ kind: 'exact', path: `${API_PREFIX}/${suffix}`, handler: async (req, res) => {
     if (bindHost !== '127.0.0.1' || !authorizedBrowser(req)) return respond(res, 403, { ok: false, error: 'Local same-origin browser request required' });
     if (!Object.hasOwn(methods, req.method)) return respond(res, 405, { ok: false, error: 'Method not allowed' });
@@ -90,6 +90,21 @@ export function createManagerRoutes({ path, sessions, backend, logStore, analyze
         catch (error) { if (error?.code === 'BAD_REQUEST') throw error; const wrapped = new Error('analysis failed'); wrapped.code = 'INTERNAL'; throw wrapped; }
       },
     }),
+    // One newest-first page of this session's full persisted decision rows, for
+    // the analysis tab's paginated log. Rows carry only panel fields (tool name,
+    // outcome, counts) — never gate command/path, which is never logged — so this
+    // exposes nothing the recent-rows /log route does not already.
+    route('logrows', {
+      GET: async req => {
+        if (typeof listRows !== 'function') { const error = new Error('log rows unavailable'); error.code = 'BAD_REQUEST'; throw error; }
+        const id = sessionId(req);
+        const params = new URL(req.url ?? '/', 'http://localhost').searchParams;
+        const page = Number.parseInt(params.get('page') ?? '1', 10);
+        const pageSize = Number.parseInt(params.get('pageSize') ?? '100', 10);
+        try { return await listRows(id, { page: Number.isNaN(page) ? 1 : page, pageSize: Number.isNaN(pageSize) ? 100 : pageSize }); }
+        catch (error) { if (error?.code === 'BAD_REQUEST') throw error; const wrapped = new Error('log rows failed'); wrapped.code = 'INTERNAL'; throw wrapped; }
+      },
+    }),
     // Append one annotation event for a decision. Explicit user intent: validated
     // and written immediately; a write failure surfaces as an error, not silence.
     route('annotate', {