docs: improve structure, navigation, and coverage across documentation

[svidskiy]

This is a proposal, not a bug report. Happy to discuss and contribute PRs
for any of the points below if the direction looks good to maintainers.

Context

I went through all the docs pages (getting-started, architecture, python-api,
cli, faq, claude-plugin) and collected a list of concrete improvements that
would make memsearch easier to adopt for new users and easier to maintain
going forward.

Proposed improvements

1. Fix broken Integrations page (404)

mkdocs.yml has Integrations: integrations.md in the nav, but the file
does not exist. Anyone clicking that link on the docs site gets a 404.

Suggestion: either create a minimal integrations.md with LangChain /
LangGraph / LlamaIndex / CrewAI examples (the README already has stubs for
these), or remove the nav entry until the page is ready.

---

2. Add a zero-config quick-start

The most common entry point for new users — no API key, local embeddings,
local Milvus Lite — requires reading across several pages to get working.
A single self-contained section ("Up and running in 2 minutes, no API key
needed") would lower the barrier significantly:

pip install "memsearch[local]"

mem = MemSearch(paths=["./notes"], embedding_provider="local")
await mem.index()
results = await mem.search("your query")

---

3. Add platform support / requirements section

After issues #89 and #90 it is clear that Windows support is a frequent
question. Currently there is no single place in the docs that states which
platforms are supported. Suggested addition to the Installation section:

Platform support
Milvus Lite (the default backend) runs on Linux and macOS only.
On Windows, use Milvus Server via Docker or Zilliz Cloud — see
Milvus Backends.

---

4. Expand the FAQ

Currently only 3 questions. Suggested additions based on common issues:

Question	Answer summary
Does memsearch work on Windows?	Not with Milvus Lite — use Docker or WSL2
How do I wipe and rebuild the index?	memsearch reset --yes && memsearch index
How do I see what is indexed?	memsearch stats
Why is search returning irrelevant results?	Check embedding provider / try re-indexing
What is "dimension mismatch" and how do I fix it?	memsearch reset --yes when switching providers

---

5. Reduce duplication between docs/claude-plugin.md and ccplugin/README.md

Both files cover the same topics (architecture, quick start, troubleshooting).
When one is updated, the other goes stale. Possible approaches:

• Make ccplugin/README.md the canonical source for plugin users (people who
  clone the repo), and have docs/claude-plugin.md focus on concepts and
  link to it
• Or consolidate everything into docs/claude-plugin.md and make
  ccplugin/README.md a short pointer

---

6. Split docs/claude-plugin.md into sub-pages

At 930+ lines it covers too much in one place. Suggested split:

• claude-plugin/overview.md — what it is, comparison table, quick start
• claude-plugin/how-it-works.md — hooks lifecycle, progressive disclosure
• claude-plugin/configuration.md — config options, collection naming
• claude-plugin/troubleshooting.md — known issues and fixes

---

7. Add cross-links between pages

A few obvious missing links:

• python-api.md → architecture.md when mentioning chunk_hash and
  composite IDs
• cli.md → getting-started.md for first-time usage context
• getting-started.md → faq.md at the bottom ("Common questions →")

---

8. Add a troubleshooting section outside the plugin docs

Common errors (dimension mismatch, connection refused, API key missing) are
currently buried inside the plugin troubleshooting section. A top-level
Troubleshooting page (or section in FAQ) would help users of the Python
library and CLI who never use the plugin.

Summary

#	Change	Effort	Impact
1	Fix Integrations 404	Low	Medium
2	Zero-config quick-start	Low	High
3	Platform support section	Low	High
4	Expand FAQ	Low	High
5	Deduplicate plugin docs	Medium	Medium
6	Split claude-plugin.md	Medium	Medium
7	Add cross-links	Low	Medium
8	Troubleshooting page	Medium	Medium