Skip to content

Teardown

Teardown removes all containers, volumes, and optionally images created by hemlock-lab.

Quick Teardown

# Stop containers, remove volumes
docker compose down -v

Full Teardown

# Stop containers, remove volumes AND images
docker compose down --rmi all -v
Flag Purpose
-v Remove named volumes (chromadb-data)
--rmi all Remove all images built or pulled by Compose

What Gets Destroyed

Resource down -v down --rmi all -v
Running containers ✓ ✓
ChromaDB data volume ✓ ✓
Pipeline images — ✓
ChromaDB image — ✓
hemlock-net network ✓ ✓

Irreversible

docker compose down -v deletes the ChromaDB data volume. All ingested collections and test data are permanently removed.

What Survives

These items are on your host, not in containers:

Resource Location Survives?
hemlock-lab repo ~/projects/hemlock-lab/ ✓
Local reports reports/ directory ✓
Ollama models ~/.ollama/ on host ✓
.env overrides Project root ✓

Rebuild After Teardown

docker compose up -d       # Pull images, create containers
docker compose ps           # Verify all running

If you did --rmi all, images will be re-pulled/rebuilt on the next up.

Partial Cleanup

Remove specific containers without tearing down everything:

# Rebuild a single pipeline
docker compose up -d --build langchain-rag

# Remove only ChromaDB data
docker volume rm hemlock-lab_chromadb-data

Or use docker compose down -v && docker compose up -d as a full reset (see Reset).

Clean Up Ollama (Host)

Since Ollama runs on the host, it's not affected by Compose teardown:

# Remove a model
ollama rm smollm2:135m
ollama rm nomic-embed-text

# Remove Ollama entirely (macOS)
brew uninstall ollama
rm -rf ~/.ollama

Next Steps

Legacy: Proxmox Teardown

The original Proxmox deployment used qm stop 260 --skiplock && qm destroy 260 --purge via make teardown to destroy the entire VM, including disk images and snapshots. The VM template 1101 was preserved for rebuilding.