Skip to main content

Troubleshooting

Dashboard shows "ES: red" or "ES: unreachable"​

The dashboard cannot connect to Elasticsearch.

  1. Check Elasticsearch is running: sudo systemctl status lme-elasticsearch
  2. Test connectivity: curl -sk https://localhost:9200
  3. Restart if needed: sudo systemctl restart lme-elasticsearch

Dashboard shows "LLM: unreachable"​

The LiteLLM proxy is not responding.

  1. Check the service: sudo systemctl status lme-litellm
  2. Check llama.cpp (LiteLLM depends on it): sudo systemctl status lme-llama-cpp
  3. Restart the AI stack: 
    sudo systemctl restart lme-llama-cpp
    sleep 5
    sudo systemctl restart lme-litellm

Dashboard shows "pgvector: unreachable"​

The vector database is not responding.

  1. Check the service: sudo systemctl status lme-pgvector
  2. Restart: sudo systemctl restart lme-pgvector
  3. If RAG was working before, re-ingest docs from Settings > Documents

AI responses are slow or low quality​

The default 1.2B model is small and fast but has limited reasoning ability. Options:

  • Download a larger local model (see Managing Models)
  • Connect a cloud model for better analysis quality

"Certificate error" when accessing the dashboard​

LME uses self-signed TLS certificates. This is expected.

  • Chrome: Click "Advanced" then "Proceed to site"
  • Firefox: Click "Advanced" then "Accept the Risk and Continue"
  • Edge: Click "Continue to site"