Honcho connection failed 'HonchoClientConfig' object has no attribute 'memory_mode' | Nous Research | Page 1

severe belfry Apr 6, 2026, 4:54 PM

#

marc@theden:~$ hermes honcho status

Honcho status
────────────────────────────────────────
  Host:           hermes
  Enabled:        True
  API key:        not set
  Workspace:      hermes
  Config path:    /home/marc/.hermes/honcho.json
  AI peer:        hermes
  User peer:      marc
  Session key:    marc
  Recall mode:    hybrid
  Observation:    user(me=True,others=True) ai(me=True,others=True)
  Write freq:     async

  Connection... OK

  No peer data yet (accumulates after first conversation)

marc@theden:~$ hermes memory status

Memory status
────────────────────────────────────────
  Built-in:  always active
  Provider:  honcho

  Plugin:    installed ✓
  Status:    available ✓

  Installed plugins:
    • byterover  (requires API key)
    • hindsight  (API key / local)
    • holographic  (local)
    • honcho  (API key / local) ← active
    • mem0  (API key / local)
    • openviking  (API key / local)
    • retaindb  (API key / local)

When I run hermes doctor I get this:
✗ Honcho connection failed 'HonchoClientConfig' object has no attribute 'memory_mode'

I've tried re-running the setup with hermes memory setup and hermes honcho setup but neither helped. I tried manually adding in memory_mode, memorymode, and memoryMode (one at a time) in the .hermes/honcho.json but it didn't help. I tried putting that config from honcho.json into the config.yaml both in the {} and as just line items under the Honcho section.

Anyone have any pointers?

simple spear Apr 6, 2026, 5:01 PM

#

@sullen crown @glass glacier

sullen crown Apr 6, 2026, 5:04 PM

#

severe belfry ``` marc@theden:~$ hermes honcho status Honcho status ─────────────────────────...

hi there! memory_mode is dead code. i think it still exists in the hermes doctor. will open a cleanup PR . but workaround could be asking hermes to give placeholder config value memory_mode: hybrid in the meantime. see if that helps

#

looks like honcho is active and you set everything up correctly!

#

when did you last run hermes update ?

severe belfry Apr 6, 2026, 5:05 PM

#

Nice. That doctor output had me worried. I just started setting things up last night. Thanks for the quick answer!

#

About an hour ago

sullen crown Apr 6, 2026, 5:05 PM

#

ok! thanks that helps (: are you using the cloud or running honcho locally?

severe belfry Apr 6, 2026, 5:07 PM

#

locally

sullen crown Apr 6, 2026, 5:09 PM

#

heroic!!

severe belfry Apr 6, 2026, 9:32 PM

#

Hey @sullen crown I appear to still be stuck. I can run one prompt - sometimes two - and then it goes to "Initializing agent..." even though I have plently of context left. Any thoughts on what else I can check on?

●   /hermes-agent see how the memory system is setup. See if there are issue and if you find any let me know
────────
● [SYSTEM: The user has invoked the "hermes-agent" skill, indicating they want you to follow its instructions. The full skill content is loaded below.] (+658 lines)

:zap: Loading skill: hermes-agent

────────

  ┊ :computer: preparing terminal…
  ┊ :computer: $         hermes honcho status  0.4s
  ┊ :computer: preparing terminal…
  ┊ :computer: $         grep -A 5 "^honcho:" ~/.hermes/config.yaml  0.2s

╭─ ⚕ Hermes ──────────╮
Memory system looks healthy now:

**Honcho status:**
- :white_check_mark: Connected and OK
- :white_check_mark: Enabled with hybrid mode
- :white_check_mark: API key not set (you're running local Honcho instance at localhost:8000)
- :white_check_mark: Configuration in `honcho.json` is correct

**Config.yaml:**
- :white_check_mark: Honcho section properly configured with `enabled: true` and `mode: hybrid`

**No issues found.** The Honcho memory backend is working correctly, and your config is fixed from the earlier empty `honcho: {}` problem.

The files you updated (.hermes/memories/MEMORY.md, USER.md, SOUL.md) are being picked up properly through the local Honcho instance. Everything's connected and running.
╰─────────────────────╯
────────
● great.  Who are you and who am I?

Initializing agent...
─────────

  ┊ :zap: preparing honcho_profile…
  ┊ :zap: honcho_pr   0.0s

╭─ ⚕ Hermes ───────────╮
It looks like I don't have any existing profile information for you, so I can't provide any personalized insights.

How can I help you today? Are you looking to:

1.  **Get information** about a specific topic?
...

sullen crown Apr 6, 2026, 9:47 PM

#

the honcho profile takes some time to be synthesized in the system. you can have hermes agent write your own peer card or create conclusions -- or prompt it to interview you and save it to honcho -- would try that route first!

sullen crown Apr 6, 2026, 9:47 PM

#

severe belfry Hey <@429384745374187523> I appear to still be stuck. I can run one prompt - so...

^

severe belfry Apr 6, 2026, 9:51 PM

#

.hermes/memories/MEMORY.md, USER.md, SOUL.md - these are all populated porperly already. The "Initializing agent..." part just seems to keep popping up in the session and its like a full session reset every time.

severe belfry Apr 7, 2026, 2:49 AM

#

sullen crown Apr 8, 2026, 3:42 PM

#

severe belfry .hermes/memories/MEMORY.md, USER.md, SOUL.md - these are all populated porperly ...

thanks will look into this to confirm whether or not it is honcho related or something dealing with the overall hermes memory plugins architecture. lmk if you still experience it after the latest hermes update

severe belfry Apr 8, 2026, 4:05 PM

#

I tried ripping out both hermes and honcho and setting them up again. I'm not getting the "Initializing agent..." part every time anymore and it is saving the memories to the /.hermes/memories/memory.md and user.md now. But it still is not using Honcho. Hermes was installed using the quick install script (curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash) and honcho was installed locally by cloning the repo, copying and editing the compose.yaml and .env files and then bringing it up in docker. I have no auth set on the Honcho right now and the docker logs show no issues.

severe belfry Apr 8, 2026, 7:28 PM

#

What was broken:

Honcho could write to the API, but inserts were failing because the app expected 1536-dim vectors while 'nomic-embed-text-v2-moe:latest' embedding endpoint returns 768.
Dialectic reads were failing because Honcho was trying to write traces.jsonl in /app and the container user didn’t have permission.

What needed to be changed:

Patched /home/marc/honcho/src/models.py to use settings.VECTOR_STORE.DIMENSIONS instead of hardcoded 1536
Altered Postgres columns to vector(768):
- public.documents.embedding
- public.message_embeddings.embedding
Patched /home/marc/honcho/.env:
- LOCAL_METRICS_FILE=/tmp/metrics.jsonl
- REASONING_TRACES_FILE=/tmp/traces.jsonl
Restarted honcho-api and honcho-deriver
Aligned ~/.hermes/config.yaml so Honcho user peer is Marc

Verified working:

honcho_conclude now succeeds
honcho_context now returns stored memory
hermes memory status shows Honcho available
hermes honcho status shows connection OK

#Honcho connection failed 'HonchoClientConfig' object has no attribute 'memory_mode'