Duplicate entries in memory | Friends of the Crustacean 🦞🤝 | Page 1

lone ruin Apr 20, 2026, 6:46 AM

#

I am getting duplicate memory entries in Dreaming mode

ionic ginkgoBOT Apr 20, 2026, 6:46 AM

#

covert valleyBOT Apr 20, 2026, 6:47 AM

#

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

outer elmBOT Apr 20, 2026, 6:49 AM

#

That’s probably version-sensitive.

I checked current source, and newer builds make deep promotion replay-safe, so reruns should reconcile existing MEMORY.md promotions instead of appending duplicates.

First, what does this show?

openclaw --version

Also, can you tell me where the duplicates are showing up:

MEMORY.md
DREAMS.md

If you’re already on a recent build, paste these too:

#

openclaw status --all
openclaw memory status --deep --agent main

And for a safe repair pass, try:

openclaw doctor --fix
openclaw memory status --fix --agent main

Relevant docs:

#

If your version is older than 2026.4.5, I’d update first, because that’s around where the duplicate MEMORY.md promotion replay fix landed.

lone ruin Apr 20, 2026, 6:50 AM

#

dreams.md

outer elmBOT Apr 20, 2026, 6:51 AM

#

Got it, that’s a different bucket.

If the duplicates are in dreams.md / DREAMS.md, that’s the Dream Diary lane, not long-term MEMORY.md.

I checked current source, and there are now two relevant fixes:

exact Dream Diary dedupe
dreaming artifact repair for contaminated/self-ingested dream state

First thing I still want is:

openclaw --version

Then please paste:

#

openclaw status --all
openclaw memory status --deep --agent main

And if you can, a small redacted snippet of 2 duplicated diary entries from dreams.md.

Why I’m asking: there are two different failure modes here:

exact duplicate blocks in DREAMS.md
near-duplicate / repeated dream runs, which usually means dreaming state got weird and needs repair

Useful docs:

#

As a safe first pass, you can also run:

openclaw doctor --fix
openclaw memory status --fix --agent main

That won’t tell me everything yet, but it often clears stale dreaming state.

lone ruin Apr 20, 2026, 6:58 AM

#

By source:
memory · 126/130 files · 1008 chunks
Vector: unknown
FTS: ready
Embedding cache: enabled (0 entries)
Batch: disabled (failures 0/2)
Recall store: 5284 entries · 4 promoted · 5284 concept-tagged · 2823 spaced · scripts=5283 latin, 1 mixed
Recall path: /var/lib/openclaw/data/openclaw/workspace/memory/.dreams/short-term-recall.json
Recall updated: 2026-04-20T06:12:38.587Z
Dreaming artifacts: diary present · 17 corpus files · ingestion state present
Dream corpus: /var/lib/openclaw/data/openclaw/workspace/memory/.dreams/session-corpus
Dream ingestion: /var/lib/openclaw/data/openclaw/workspace/memory/.dreams/session-ingestion.json
Dream diary: /var/lib/openclaw/data/openclaw/workspace/DREAMS.md

outer elmBOT Apr 20, 2026, 6:59 AM

#

Thanks, that helps.

That output confirms the dreaming artifacts exist, but it does not tell us yet whether this is:

exact duplicate Dream Diary blocks, or
repeated/self-ingested dreaming runs

I still need the version first:

pnpm openclaw --version

Then run:

#

pnpm openclaw doctor --fix
pnpm openclaw memory status --fix --agent main

Please paste anything those commands say about:

dreaming
dream diary
session corpus
self-ingested narrative

And if you can, paste one duplicated block from DREAMS.md (redacted is fine).
If the two blocks are byte-for-byte identical, that points one way. If they differ a bit, that points another.

If this fixed it, please right-click the solution message → Apps → ✅ Mark Solution so Answer Overflow can index it. 🎉

lone ruin Apr 20, 2026, 6:59 AM

#

🦞 OpenClaw 2026.4.15 (041266a) — Type the command with confidence—nature will provide the stack trace if needed.

outer elmBOT Apr 20, 2026, 7:00 AM

#

Nice, that helps.

2026.4.15 is new enough that the old duplicate MEMORY.md promotion bug should already be fixed. So this looks more like a DREAMS.md / Dream Diary problem.

Next step:

pnpm openclaw doctor --fix
pnpm openclaw memory status --fix --agent main

Then paste the output, especially if it mentions:

dreaming artifacts
session corpus
self-ingested narrative
dream diary

#

Also, two quick checks:

Do you have heartbeat enabled?
Are the duplicates exact copies (same timestamp + same text), or same text with different timestamps?

If you can paste one duplicated diary block from DREAMS.md (redacted is fine), I can narrow it down fast.

lone ruin Apr 20, 2026, 7:05 AM

#

Memory Search (main)
Provider: none (requested: auto)
Model: none
Sources: memory
Indexed: 126/130 files · 1008 chunks
Dirty: no
Store: ~/.openclaw/memory/main.sqlite
Workspace: /var/lib/openclaw/data/openclaw/workspace
Dreaming: 0 3 * * * · limit=10 · minScore=0.8 · minRecallCount=3 · minUniqueQueries=3 · recencyHalfLifeDays=14 · maxAgeDays=30
By source:
memory · 126/130 files · 1008 chunks
Vector: unknown
FTS: ready
Embedding cache: enabled (0 entries)
Batch: disabled (failures 0/2)
Recall store: 5284 entries · 4 promoted · 5284 concept-tagged · 2823 spaced · scripts=5283 latin, 1 mixed
Recall path: /var/lib/openclaw/data/openclaw/workspace/memory/.dreams/short-term-recall.json
Recall updated: 2026-04-20T06:12:38.587Z
Dreaming artifacts: diary present · 17 corpus files · ingestion state present
Dream corpus: /var/lib/openclaw/data/openclaw/workspace/memory/.dreams/session-corpus
Dream ingestion: /var/lib/openclaw/data/openclaw/workspace/memory/.dreams/session-ingestion.json
Dream diary: /var/lib/openclaw/data/openclaw/workspace/DREAMS.md
Repair: no changes

outer elmBOT Apr 20, 2026, 7:06 AM

#

Thanks, that’s useful.

Also, Provider: none / Vector: unknown is not the problem here. That part is unrelated to duplicate DREAMS.md entries.

The important bit is this:

pnpm openclaw memory status --fix only checks recall-store repair / stale promotion locks
it does not fully rule out dreaming artifact / session-corpus contamination

So the next command I want is still:

pnpm openclaw doctor --fix

Please paste that output.

#

That is the path that checks the broader dreaming-artifact problems, including the self-ingested/session-corpus weirdness I’m worried about.

If doctor --fix also says nothing changed, then this is starting to look like a real DREAMS.md duplication bug.

Temporary workaround while testing:

/dreaming off

so it stops adding more duplicates.

One more thing: is heartbeat enabled on this agent?

lone ruin Apr 20, 2026, 7:07 AM

#

its longer than 2000 characters

outer elmBOT Apr 20, 2026, 7:07 AM

#

Yep, that’s fine.

Easiest options:

upload it as a doctor.txt file, or
paste only the lines that mention:

dream
diary
session
ingest
repair
warning
archive

If you want to filter it first:

#

pnpm openclaw doctor --fix 2>&1 | grep -Ei 'dream|diary|session|ingest|repair|warning|archive'

Also, is heartbeat enabled on this agent?

lone ruin Apr 20, 2026, 7:07 AM

#

yes heartbeat is enabled

outer elmBOT Apr 20, 2026, 7:08 AM

#

Ah, that narrows it a lot.

With:

2026.4.15
duplicates in DREAMS.md
memory status --fix showing no changes
heartbeat enabled

I strongly suspect you’re hitting a dreaming + heartbeat regression, not a bad memory-search config.

My recommendation:

Keep heartbeat on, turn dreaming off with /dreaming off
(best immediate workaround, stops new dupes)
If you need dreaming more than heartbeat, disable heartbeat instead
If you want to help pin it down, grab the filtered doctor --fix output and file a bug

#

Bug report link:
https://github.com/openclaw/openclaw/issues

Useful details to include:

pnpm openclaw --version
heartbeat enabled
duplicates are in DREAMS.md
pnpm openclaw memory status --deep --fix --agent main
one redacted duplicated diary block
optionally:

pnpm openclaw doctor --fix 2>&1 | grep -Ei 'dream|diary|session|ingest|repair|warning|archive'

If you paste that filtered doctor output here, I can help you write a tight issue report.

lone ruin Apr 20, 2026, 7:15 AM

#

[svc_admin_ansible@openclaw ~]$ openclaw doctor --fix 2>&1 | grep -Ei 'dream|diary|session|ingest|repair|warning|archive
'
◇ Config warnings ─────────────────────────────────────────────────╮
◇ Doctor warnings ────────────────────────────────────────────────────────╮
│ These agents can still have sessions/auth state on disk, but │
│ - 1/5 recent sessions are missing transcripts. │
│ Verify sessions in store: openclaw sessions --store │
│ "/home/svc_admin_ansible/.openclaw/agents/main/sessions/sessions.json" │
│ Preview cleanup impact: openclaw sessions cleanup --store │
│ "/home/svc_admin_ansible/.openclaw/agents/main/sessions/sessions.json" │
│ Prune missing entries: openclaw sessions cleanup --store │
│ "/home/svc_admin_ansible/.openclaw/agents/main/sessions/sessions.json" │
│ - Found 1 orphan transcript file in ~/.openclaw/agents/main/sessions. │
│ These .jsonl files are no longer referenced by sessions.json, so │
│ they are not part of any active session history. │
│ Doctor can archive them safely by renaming each file to │
│ - Archived 1 orphan transcript file in │
│ ~/.openclaw/agents/main/sessions as .deleted timestamped backups. │
Session store (main): /home/svc_admin_ansible/.openclaw/agents/main/sessions/sessions.json (1059 entries)

agent:main:dreaming-narrative-deep-5714cc07e417-1776667657331 (20m ago)
Config warnings:\n- plugins.entries.memory-postgres: plugin disabled (disabled in config) but config is present

outer elmBOT Apr 20, 2026, 7:15 AM

#

Yep, this is enough to call bug territory.

What I’m seeing:

2026.4.15
heartbeat enabled
duplicates are in DREAMS.md
memory status --fix said Repair: no changes
doctor --fix did not find dreaming-artifact corruption, it only cleaned up session-store/orphan-transcript stuff

So the orphan transcript warning looks like side noise here, not the root cause.

My read is: this is most likely a heartbeat-triggered dreaming duplication bug.

Immediate workaround

Use:

#

/dreaming off

to stop new duplicate diary entries.

Please file a GitHub issue

https://github.com/openclaw/openclaw/issues

You can pretty much paste this:

Title: [Bug]: DREAMS.md gets duplicate entries when dreaming is enabled with heartbeat on

Version:
OpenClaw 2026.4.15 (041266a)

#


What happens:
- Dreaming is enabled
- Heartbeat is enabled
- Duplicate entries show up in DREAMS.md

What I checked:
- `openclaw memory status --deep --fix --agent main`
  - Repair: no changes
- `openclaw doctor --fix`
  - did not report dreaming-artifact/session-corpus repair
  - only reported missing/orphan transcript cleanup

Notes:
- Duplicates are in `DREAMS.md`, not `MEMORY.md`

#

- Heartbeat is enabled
- This looks like dreaming may be retriggering from heartbeat rather than a memory-search config issue

If helpful, I can attach a redacted duplicate diary block.

If you want, paste the final issue text here before you submit it and I’ll tighten it up.

lone ruin Apr 20, 2026, 7:17 AM

#

is there an existing issue with that?

outer elmBOT Apr 20, 2026, 7:18 AM

#

Yep, I found two closely related issues:

#65705, Dreaming sweep re-triggers on every heartbeat instead of once per cron schedule
https://github.com/openclaw/openclaw/issues/65705
#65652, Dreaming still triggers on every heartbeat in 2026.4.11, creating empty light/rem/heartbeat sessions
https://github.com/openclaw/openclaw/issues/65652

Your case sounds closest to #65705.

I’d comment on that issue first with:

OpenClaw 2026.4.15 (041266a)
heartbeat enabled
duplicates are in DREAMS.md
openclaw memory status --fix --agent main → Repair: no changes
openclaw doctor --fix only found orphan/missing transcript cleanup

#Duplicate entries in memory

Immediate workaround

Please file a GitHub issue