#General support for Openclaw

1 messages · Page 1 of 1 (latest)

simple swan
#

This is general support chat I am opening for my questions.
What is the benefit of session-memory hook?

urban portalBOT
#

It’s basically a safety net when you start over.

session-memory runs when you use /new or /reset and saves the recent conversation into:

<workspace>/memory/YYYY-MM-DD-slug.md

Benefits:

  • preserves useful context before the session is cleared
  • makes old discussion searchable/recallable later via memory tools
  • gives the next session some recent background instead of losing everything
  • keeps it as plain Markdown, so you can inspect/edit/delete it
  • creates descriptive filenames like api-design, bug-fix, etc.

Main caveat: it only saves the last N user/assistant messages — default 15 — so it’s not a full transcript archive or a perfect long-term memory system.

Enable/check it with:

simple swan
#

this makes the /new or /reset very slow

urban portalBOT
#

Yep, that tracks. The slow part is probably the LLM-generated filename slug, not the markdown write itself.

On /new or /reset, session-memory reads the recent transcript, asks your configured model to create a nice slug like api-design / bug-fix, then writes the memory file. If your model is slow, local, overloaded, or rate-limited, /new feels slow.

Fastest fix:

openclaw hooks disable session-memory

If you want to keep it but reduce work, lower the captured messages:

{
  hooks: {
    internal: {
# 
      entries: {
        "session-memory": {
          enabled: true,
          messages: 5
        }
      }
    }
  }
}

There’s also a source-supported escape hatch to skip the LLM slug generation:

{
# 
  hooks: {
    internal: {
      entries: {
        "session-memory": {
          enabled: true,
          llmSlug: false
        }
      }
    }
  }
}

That should fall back to timestamp filenames and avoid the model call.

Also: what version are you on?

simple swan
#

OpenClaw 2026.4.27

urban portalBOT
#

2026.4.27 should be new enough for the fast workaround.

I’d try this first — keep the hook, but skip the LLM slug call:

{
  hooks: {
    internal: {
      entries: {
        "session-memory": {
          enabled: true,
          llmSlug: false
        }
      }
    }
  }
# 
}

Then restart the gateway:

openclaw gateway restart

That should make /new / /reset much faster while still saving the memory file, just with timestamp-ish filenames instead of nice generated names.

If it’s still slow after that, disable the hook entirely:

openclaw hooks disable session-memory
simple swan
#

Will that make model ineffective to search previous convos?

urban portalBOT
#

No — setting llmSlug: false should not make search ineffective.

It only changes the filename:

  • with LLM slug: 2026-04-30-api-debugging.md
  • without LLM slug: 2026-04-30-0026.md or similar timestamp fallback

The actual saved conversation content is still written into memory/*.md, and memory search searches the file contents. So model recall/search should still work basically the same.

The tradeoff is:

  • llmSlug: false → faster /new / /reset, uglier filenames
  • session-memory disabled entirely → no automatic saved excerpt on /new / /reset, so less previous-convo material unless you use transcript indexing / other memory systems

So I’d use:

simple swan
#

Hello....

#

Oh no @urban portal did the new update kill you too kek

simple swan
#

I want to exclude a particular discord channel from the server how can I do that?

urban portalBOT
#

Yep — for Discord guild channels, do it in channels.discord.guilds.<serverId>.channels.

If you want to allow the server generally but exclude one channel, use a wildcard allow plus disable the specific channel:

{
  channels: {
    discord: {
      groupPolicy: "allowlist",
      guilds: {
        "YOUR_SERVER_ID": {
          channels: {
            "*": { enabled: true },
            "CHANNEL_ID_TO_EXCLUDE": { enabled: false }
          }
        }