Fathom VM Recovery Guide

Backup timestamp: 20260417_053924 HF dataset: umer07/vllm-deployement-backup

This backup matches the live VM after the current vLLM, backend, dashboard, log endpoint, classifier, RAG-context, demo HTML, and upload-flow changes.

Current Runtime

Dashboard: port 3000
Backend: port 7860
vLLM OpenAI-compatible endpoint: port 8000
vLLM service: fathom-vllm.service
vLLM start script: /usr/local/bin/fathom-vllm-start
vLLM model id exposed to backend: fathom-base
vLLM max context configured: 32768
Main app root: /opt/fathom
Public demo HTML: /opt/fathom/dashboard/public/demo.html
Root dashboard route redirects to /demo.html

Backup Files

vllm_deployement_backup_20260417_053924.tar.gz
fathom_minio_data_20260417_053924.tar.gz
fathom_neo4j_data_20260417_053924.tar.gz
fathom_neo4j_logs_20260417_053924.tar.gz
fathom_hf_cache_20260417_053924.tar.gz
fathom_adapters_20260417_053924.tar.gz
vllm_deployement_backup_manifest_20260417_053924.txt
restore_fathom_vm.sh

Restore

Use the restore script on a ROCm/vLLM-capable VM:

apt update
apt install -y docker.io docker-compose-v2 python3 python3-pip curl
python3 -m pip install --break-system-packages -U huggingface_hub

HF_TOKEN='<your token>' bash restore_fathom_vm.sh --download-latest --start

The restore expects the DigitalOcean vLLM ROCm base container named rocm or an equivalent ROCm/vLLM container that supports /opt/venv/bin/vllm. The large Hugging Face model cache under /shared-docker/hf-cache is intentionally excluded and will be repopulated from Hugging Face.

Verify

systemctl status fathom-vllm.service --no-pager
curl http://127.0.0.1:8000/v1/models
curl http://127.0.0.1:7860/health
curl -I http://127.0.0.1:3000/
curl http://127.0.0.1:3000/demo.html