feat: add persistent local memory plugin for OpenClaw (memos-local-openclaw)

[tangbotony]

Description

Introduce MemOS (memos-local-openclaw-plugin), a fully local OpenClaw memory plugin that gives AI agents persistent memory, intelligent task summarization, self-evolving skills, and multi-agent collaboration.

Problem: OpenClaw agents have no persistent memory across sessions. Every conversation starts from scratch — the agent forgets previous interactions, repeats mistakes, and cannot learn from past experience. There is no mechanism to organize conversations into structured knowledge, distill successful executions into reusable skills, or share knowledge across multiple agents.

Solution — Four Intelligent Pipelines:

1. Memory Write Pipeline — Auto-captures every conversation via agent_end hook → semantic chunking → LLM summarization → vector embedding → smart deduplication (content-hash + Top-5 vector + LLM judge: DUPLICATE/UPDATE/NEW) → SQLite + FTS5 storage

2. Task Summarization Pipeline — Per-turn LLM topic boundary detection (warm-up: 1 user turn, 2h idle timeout) → structured summaries (Goal → Steps → Result → Key Details) → quality filtering → full CRUD from Viewer

3. Skill Evolution Pipeline — Rule filter → LLM evaluation → SKILL.md generation with scripts + evals → quality scoring (0-10) → auto-upgrade on similar tasks (refine/extend/fix) → version management → 3-level LLM fallback chain (skillSummarizer → summarizer → OpenClaw native model)

4. Smart Retrieval Pipeline — FTS5 + vector hybrid search → RRF fusion → MMR diversity reranking → recency decay → LLM relevance filter → auto-recall via before_agent_start hook

Additional capabilities:

• Multi-agent collaboration — Memory isolation via owner field + public memory + skill sharing/publishing
• Memory Migration (Reconnect 🦞) — One-click import from OpenClaw native memories with smart dedup, pause/resume, configurable agent parallelism (1-8)
• Memory Viewer — 7-page web dashboard (Memories, Tasks, Skills, Analytics, Logs, Import, Settings) with full CRUD, i18n, light/dark theme, password protection
• 12 smart agent tools — auto_recall, memory_search, memory_get, memory_timeline, memory_write_public, task_summary, skill_get, skill_install, skill_search, skill_publish, skill_unpublish, memory_viewer
• Tiered models — Embedding (lightweight) / Summarizer (mid-tier) / Skill generation (high-quality), each independently configurable
• Zero-config start — Falls back to local embedding + rule-based summarization when no API keys configured
• 100% local — All data in local SQLite, zero cloud uploads, localhost-only Viewer

Related Issue (Required): Fixes #1181

Type of change

• New feature (non-breaking change which adds functionality)
• Documentation update

How Has This Been Tested?

• Unit Test
• Test Script Or Test Steps (please provide)

Unit tests (npm test in apps/memos-local-openclaw/):

• tests/policy.test.ts — Retrieval strategy, search filtering, evidence extraction
• tests/recall.test.ts — RRF fusion, recency decay correctness
• tests/capture.test.ts — Message filtering, evidence block stripping, self-tool exclusion
• tests/storage.test.ts — SQLite CRUD, FTS5, vector storage, content hash dedup
• tests/chunker.test.ts — Semantic chunking for code blocks, paragraphs, function bodies
• tests/task-processor.test.ts — Task boundary detection, skip logic, summary generation
• tests/multi-agent.test.ts — Multi-agent memory isolation, owner filtering, public sharing
• tests/integration.test.ts — End-to-end ingestion and retrieval pipeline

Manual testing:

1. Install plugin via openclaw plugins install . → start gateway → verify Memory Viewer at http://127.0.0.1:18799
2. Have conversations with OpenClaw agent → verify auto-capture, dedup, task detection, skill generation
3. Test memory migration from OpenClaw native memories → verify dedup, resume, post-processing
4. Test multi-agent scenarios → verify memory isolation and skill sharing
5. Test task/skill CRUD operations from Viewer list cards (edit, delete, retry, visibility toggle)
6. Test LLM fallback chain by disabling primary model → verify automatic fallback to next level

Checklist

• I have performed a self-review of my own code | 我已自行检查了自己的代码
• I have commented my code in hard-to-understand areas | 我已在难以理解的地方对代码进行了注释
• I have added tests that prove my fix is effective or that my feature works | 我已添加测试以证明我的修复有效或功能正常
• I have created related documentation issue/PR in MemOS-Docs (if applicable) | 我已在 MemOS-Docs 中创建了相关的文档 issue/PR（如果适用）
• I have linked the issue to this PR (if applicable) | 我已将 issue 链接到此 PR（如果适用）
• I have mentioned the person who will review this PR | 我已提及将审查此 PR 的人

Reviewer Checklist

• closes feat: MemOS — persistent local memory plugin for OpenClaw with task summarization, skill evolution, and multi-agent collaborationClaw with task summarization, skill evolution, and multi-agent collaboration #1181 (Replace with actual issue number)
• Made sure Checks passed
• Tests have been provided