cache stats output

[ousatov-ua]

Hi!

I'm just started to use lean-ctx and found strange thing, lean-ctx cache stats does not show that anything is cached:

lean-ctx cache stats

As I understood, this cache is used whenever read is proceeding with -m argument:

lean-ctx read <file> -m

I don't see cases when this happens by default.

Also, if I understood correctly, MCP has a separate cache and its status can be checked only inside MCP process.

I'm sorry if I missed RTFM.
Can you please point me to the doc or explain a little bit?

BTW! If I go by these steps: https://leanctx.com/docs/getting-started/#editor-setup (Verify) the output is not compressed because by default track mode is active.
I had to run lean-ctx-mode compress to see compressed output for such a commands like git status

[VectorPeak]

No worries - this is a docs/naming confusion, not RTFM.

-m only selects the read mode. It is not the thing that enables caching. Both of these go through the read pipeline:

lean-ctx read <file>
lean-ctx read <file> -m map

The cache win happens on a repeated read of the same unchanged file in the same session/conversation. The first read still has to render content; the second read can collapse to a small “already seen / unchanged” stub. Different mode, changed file contents, --fresh, or a different conversation can force a full render again.

lean-ctx cache stats is currently easy to misread because it does not necessarily reflect the live per-session MCP/read cache you are expecting to see. For testing the read-cache behavior, this is the better command:

lean-ctx verify-cache <some-file>

For MCP, yes: the live session cache is owned by that MCP process/session, so CLI-level stats may not show what the editor process has in memory. Inside MCP, the relevant check is the cache/status tool for that running session.

Your git status note looks like a real docs mismatch. The Verify step assumes shell compression mode, but if your hook is in track mode then command output is intentionally not shown/compressed the same way. Running this is the right workaround:

lean-ctx-mode compress

The docs should probably call out both points more explicitly: cache stats vs session read cache, and the fact that the Verify example depends on shell mode being compress, not track.

---

If my answer solved your problem, you can click answered the question. I'm really here to help, and along the way I'm also collecting Galaxy Brain badges haha 😆