#Active memory plugin useful in my setup?

1 messages · Page 1 of 1 (latest)

torpid stump
#

I have openclaw 2026.4.15 running with QMD and local honcho memory plugin. Is there any additional value in activating active memory since honcho is already a fairly comprehensive system? And if so, what setup would you advise?

haughty rootBOT
mighty hingeBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> âś… Mark Solution

fluid adderBOT
#

Short version: probably not much extra value by default if Honcho is already doing the job you want.

Why:

  • Honcho already gives you proactive cross-session/user-model context before prompt build, plus semantic recall across conversations.
  • Active Memory is a separate blocking pre-reply pass that uses OpenClaw’s local memory stack, basically memory_search / memory_get, so with your setup that means mostly QMD-backed local recall.
  • So if Honcho already has the same facts/context, Active Memory can become mostly extra latency + duplicate recall.

Where Active Memory does help:

  • you want local QMD-only knowledge surfaced automatically, not just Honcho memory
  • QMD is indexing extra local paths / notes / docs / transcripts
  • you want a very small “check local memory first” pass for stable preferences / habits

My recommendation:

  1. Keep Honcho as primary
#
  1. Only add Active Memory as a narrow supplement
  2. Start DM-only, not everywhere

A safe starter config would be:

{
  plugins: {
    entries: {
      "active-memory": {
        enabled: true,
        config: {
          agents: ["main"],
          allowedChatTypes: ["direct"],
          queryMode: "recent",
          promptStyle: "balanced", // or "preference-only" if you want it narrower
# 
          timeoutMs: 8000,
          maxSummaryChars: 160,
          persistTranscripts: false,
          logging: true
        }
      }
    }
  }
}

A few practical notes:

  • Leave config.model unset at first, let it inherit the session model.
  • If it feels noisy, switch to promptStyle: "preference-only".
  • If it feels slow or redundant, turn it back off. That usually means Honcho is already covering the same ground well.
torpid stump
#

I let honcho manage all my extra notes, paths and transcripts. So less reason to use active memory. And why would I want qmd-only knowledge if the honcho plugin already uses qmd?

fluid adderBOT
#

Yeah, I think your read is basically right.

If Honcho is already managing the notes, paths, and transcripts you care about, then Active Memory is mostly not adding new knowledge. It is mostly adding a second recall behavior.

That’s the key distinction:

  • Honcho = memory system + proactive context injection
  • Active Memory = an extra blocking pre-reply recall pass that decides “should I inject a short memory summary right now?”

So if your Honcho setup is already surfacing the right stuff at the right time, I would not enable Active Memory. You’d mostly be stacking two proactive memory layers, which can mean:

  • more latency
  • more duplicate recall
  • more prompt noise

So to your “why would I want QMD-only knowledge?” point: you probably wouldn’t, if Honcho already has/uses that same corpus.

#

The only real reason to test Active Memory in your setup would be this:

“Honcho has the data, but in live chat I still want a very aggressive, tiny, pre-reply preference/habit recall pass.”

If you’re not seeing that gap, I’d skip it.

My practical recommendation:

  • Default: Honcho + QMD, no Active Memory
  • Only test Active Memory if you notice missed recall in real conversations
  • If you do test it, keep it very narrow:
    • allowedChatTypes: ["direct"]
    • queryMode: "message" or "recent"
    • promptStyle: "preference-only"
    • timeoutMs: 5000-8000
    • low maxSummaryChars