docs: Add FAQ section for common questions about MemOS Memory OS

[meichuanyi]

Summary

This PR adds a comprehensive FAQ section to the README, covering common questions about MemOS:

Topics Covered

General:

• What is MemOS
• Difference from embedding-based memory solutions
• Supported LLM providers (OpenAI, Azure, Qwen, DeepSeek, MiniMax, Ollama, HuggingFace, vLLM)

Getting Started:

• Quick start guide (Cloud vs Self-Hosted)
• Configuration requirements

Core Features:

• Multi-Modal Memory (text, images, tool traces, personas)
• Multi-Cube Knowledge Base Management
• Memory Feedback & Correction
• MemScheduler (async operations with Redis Streams)

Plugins & Integration:

• OpenClaw plugin (Cloud + Local)
• Hermes Agent plugin (memos-local-plugin 2.0)
• MCP support

Performance:

• Accuracy improvements (+43.70% vs OpenAI Memory, +40.43% LongMemEval)
• Token savings (35.24%)

Troubleshooting:

• API key errors
• Docker issues
• Local plugin troubleshooting

Help & Resources:

• Documentation links
• ArXiv paper
• Awesome-AI-Memory
• Community channels

This FAQ helps new users understand MemOS's value proposition, get started quickly, and troubleshoot common issues.

[shinetata]

Thanks @meichuanyi — the FAQ is a useful addition and most of it is accurate. I verified the benchmark numbers and the LLM-provider table (every MOS_CHAT_MODEL_PROVIDER value matches src/memos/llms/factory.py), and the plugin/deployment/links sections are consistent with the README.

Requesting changes on one item, plus a few small inline suggestions:

Must change — remove the "How does MemOS compare to other memory solutions?" table. It isn't sourced and contains factual errors about peers, which is risky in an official README. For example: mem0 does support multi-modal (images/documents), does have graph memory, and has a published paper (arXiv:2504.19413); the license column is also off (e.g. LangChain is MIT). I've left an inline suggestion replacing it with a neutral, sourced "Why MemOS?" list — feel free to use that or simply drop the section.

Minor (see inline suggestions):

Benchmark table: drop the empty Improvement column.
Self-host snippet: fix the .env copy path and add the actual start command.
Cloud dashboard link: align it with the Quick-start URL.
"Memory architecture": label it as the memos-local-plugin 2.0 self-evolving model, since it's not MemOS core's general storage.
Non-blocking follow-up (a separate PR/issue is fine): the FAQ is English-only — README_zh.md has no equivalent and is also missing the 2026-05-09 "memos-local-plugin 2.0" News entry.

[shinetata]

remove the "How does MemOS compare to other memory solutions?" table. It isn't sourced and contains factual errors about peers, which is risky in an official README.

Suggested change

	### How does MemOS compare to other memory solutions?
	| Feature | MemOS | mem0 | LangChain Memory | Letta |
	|---------|-------|------|------------------|-------|
	| Multi-Modal Memory | ✅ Text/Images/Tools | ❌ Text only | ❌ Text only | ❌ Text only |
	| Knowledge Base | ✅ Multi-Cube KB | ❌ No KB | ⚠️ RAG only | ❌ No KB |
	| Memory Feedback | ✅ Natural language | ❌ No | ❌ No | ❌ No |
	| Graph Memory | ✅ Inspectable/Editable | ❌ Black-box | ❌ Black-box | ❌ Limited |
	| Async Ingestion | ✅ MemScheduler | ❌ No | ❌ No | ❌ No |
	| Open Source | ✅ Apache 2.0 | ✅ MIT | ✅ Apache | ✅ MIT |
	| ArXiv Paper | ✅ 2507.03724 | ❌ No | ❌ No | ❌ No |
	### Why MemOS?
	- **Inspectable, editable graph memory** — not a black-box embedding store.
	- **Multi-modal**: text, images, tool traces, and personas in one system.
	- **Multi-Cube knowledge base** with isolation and controlled cross-user / project / agent sharing.
	- **Asynchronous ingestion** via MemScheduler with millisecond-level latency.
	- **Natural-language memory feedback** to correct, supplement, or replace memories over time.
	- **Open source (Apache 2.0)**, with a published method ([arXiv:2507.03724](https://arxiv.org/abs/2507.03724)).

[shinetata]

fix the .env copy path and add the actual start command.

Suggested change

	# Configure .env (OPENAI_API_KEY, etc.)
	cp docker/.env.example MemOS/.env
	# Start service
	# See docs for full setup
	# Configure .env (already inside the MemOS repo root after `cd MemOS`)
	cp docker/.env.example .env
	# Start the service (run from the docker directory)
	cd docker && docker compose up

[shinetata]

label it as the memos-local-plugin 2.0 self-evolving model, since it's not MemOS core's general storage.

Suggested change

	### What is the memory architecture?
	### What is the self-evolving memory architecture? (memos-local-plugin)
	> This is the self-evolving memory model used by `memos-local-plugin 2.0`, not MemOS core's general storage layer.