will/swarm-zap
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
3e198bcbb39372d02ee6d3c01803b7f3ce7c71da
swarm-zap/memory/2026-03-04-litellm-setup.md
zap b0b4060bf9 chore(workspace): install all 4 bundled hooks + model default update 
- Installed hooks: boot-md, session-memory, command-logger, bootstrap-extra-files
- Updated default model to litellm/copilot-claude-sonnet-4.6
- Added memory/startup-health.json + boot-last-run.json from boot check
- Added memory/startup-health.md warning log
- Captured new session memory files (2026-03-04/05)
- Minor MEMORY.md + USER.md: added git preference line
2026-03-05 04:40:46 +00:00

2.6 KiB
Raw Blame History

Session: 2026-03-04 01:09:01 UTC

  • Session Key: agent:main:main
  • Session ID: 9d4c78b3-c890-4c4d-a0de-0fb412cf05df
  • Source: gateway:sessions.reset

Conversation Summary

user: Sender (untrusted metadata):

{
  "label": "openclaw-tui (gateway-client)",
  "id": "gateway-client",
  "name": "openclaw-tui",
  "username": "openclaw-tui"
}

[Wed 2026-03-04 00:10 UTC] We have a VM accessible via SSH claw@192.168.122.11 Password is: xxxx OpenClaw is installed and I want you to setup this machine to suse litellm endpoint with the following API key: sk-l6Zg9xhXHovLXm3cArAFMg assistant: reply_to_current I can do this, but I’m missing one required value: the LiteLLM base URL.

I SSH’d in and checked the machine:

  • OpenClaw data is under /home/openclaw/.openclaw/
  • It’s currently configured for zai models, not LiteLLM
  • There is no local LiteLLM proxy on 127.0.0.1:4000 right now

To finish setup, send me:

  1. LiteLLM endpoint URL (example: http://<host>:4000 or https://litellm.example.com)
  2. (Optional but recommended) the default model id you want OpenClaw to use via LiteLLM (example: gpt-4o, claude-opus-4-1, etc.)

Once you send that, I’ll apply config + key securely and restart/verify. user: Sender (untrusted metadata):

{
  "label": "openclaw-tui (gateway-client)",
  "id": "gateway-client",
  "name": "openclaw-tui",
  "username": "openclaw-tui"
}

[Wed 2026-03-04 00:14 UTC] It's the same endpoint we use for you assistant: Got it — I’ll use the same LiteLLM endpoint as this host and apply it now on that VM. assistant: reply_to_current Done — I applied it on the VM using the same LiteLLM endpoint this host uses.

What I changed on 192.168.122.11

  • Updated /home/openclaw/.openclaw/openclaw.json
  • Set LiteLLM provider:
    • baseUrl: http://192.168.153.113:18804/v1
    • apiKey: ${LITELLM_API_KEY} (not hardcoded in provider section)
    • api: openai-completions
  • Set env.vars.LITELLM_API_KEY to the key you provided
  • Switched default model to:
    • agents.defaults.model.primary = litellm/gpt-4o
  • Added minimal LiteLLM model entries (gpt-4o, gpt-4o-mini)
  • Created backup before edit:
    • /home/openclaw/.openclaw/openclaw.json.bak.manual.<timestamp>

Verification

  • Confirmed LiteLLM endpoint is reachable from VM and returns models (49 models found).
  • Confirmed config now shows:
    • primary model: litellm/gpt-4o
    • LiteLLM URL and provider config in place.

If you want, next I can also switch it to openai-responses mode and sync a fuller model catalog (like your main box), but this current setup is valid and operational.