fix(memos-local-openclaw): Viewer API search 接口 limit 和 minScore 参数不生效

[zwcf5200]

问题描述

MemOS Viewer HTTP API 的 /api/search 端点存在以下问题：

1. limit 参数不生效

请求 /api/search?q=test&limit=3 返回所有匹配结果（可能 65+ 条），而不是限制为 3 条。

2. minScore 参数不生效

请求 /api/search?q=test&minScore=0.8 返回相似度低于 0.8 的结果，参数被完全忽略。

问题原因

在 apps/memos-local-openclaw/src/viewer/server.ts 的 serveSearch 方法中：

1. limit 参数从未被读取和使用
2. minScore 参数从未被读取，语义阈值使用硬编码值 0.64
3. 结果返回时没有截断，也没有在响应中包含参数信息

修复方案

// 添加参数读取
const limit = Math.min(100, Math.max(1, Number(url.searchParams.get("limit")) || 20));
const minScore = Math.max(0.35, Math.min(1, Number(url.searchParams.get("minScore")) || 0.64));

// 使用 minScore 替代硬编码阈值
const SEMANTIC_THRESHOLD = minScore;

// 结果截断
const results = (merged.length > 0 ? merged : ftsResults.slice(0, limit)).slice(0, limit);

// 返回参数信息
this.jsonResponse(res, {
  results,
  query: q,
  vectorCount: vectorResults.length,
  ftsCount: ftsResults.length,
  total: results.length,
  limit,
  minScore,
});

验证结果

测试	请求	返回	状态
limit=3	3 条	3 条	✅
limit=20	20 条	20 条	✅
minScore=0.7	≥70%	71%~78%	✅
minScore=0.8	≥80%	只返回高相似度结果	✅

相关文件

• apps/memos-local-openclaw/src/viewer/server.ts

影响

此问题影响所有通过 HTTP API 调用搜索功能的场景，包括：

• MCP 集成
• 第三方客户端
• 自动化脚本

[Memtensor-AI]

🤖 AutoDev has picked up this issue and started working on it.

Task ID: 31550d4b7a22aaf4
Working branch: bugfix/autodev-1372
Target branch: latest dev* branch
Workflow: opsp (analysis → coding → testing → PR)

I will post the PR link here once done. If I need more information, I will ask in the comments.

[Memtensor-AI]

✅ AutoDev task 31550d4b7a22aaf4 completed.

Summary: Fixes #1372 — viewer /api/search was silently dropping the limit and minScore query parameters. The handler in apps/memos-local-openclaw/src/viewer/server.ts never read either field: the merged result list was returned unbounded (or the fallback slice was hardcoded to 20), and the semantic-similarity gate stayed pinned at the constant 0.64.

The patch parses both params, clamps limit to [1, 100] (default 20, preserving the previous fallback slice size) and minScore to [0.35, 1] (default 0.64, preserving the previous semantic gate), uses the requested floor in place of the hardcoded SEMANTIC_THRESHOLD, truncates the final merged result list, and echoes both values back in the response JSON so callers can verify which thresholds were applied.

Regression coverage: new tests/viewer-search-params.test.ts adds four cases — limit truncation, limit clamping (zero/negative/over-100/NaN fallbacks), minScore raising the semantic gate (with a stub embedder producing deterministic cosine scores), and default-echo on a bare query. npx vitest run tests/viewer-search-params.test.ts → 4 passed. Full npx vitest run → 226 passed; the 5 remaining failures (accuracy semantic precision, summary length, skill-auto-install default, task-processor session-change, update-install SIGUSR1 timing) reproduce identically on the unchanged base branch (verified via git stash rerun) and are pre-existing — none touch serveSearch or /api/search. npx tsc --noEmit clean.

Base branch: main
Branch: bugfix/autodev-1372
Commit: af43a7e
PR: #1979
Assigned to: @MatthewZhuang, @CarltonXiang, @syzsunshine219, @World-controller
Reviewers: @MatthewZhuang, @CarltonXiang, @syzsunshine219, @World-controller