Verify Installation

Health Checks

The backend exposes two health endpoints:

Liveness

Returns 200 OK when the process is running:

curl https://your-domain/api/v1/health/live

Readiness

Returns 200 OK when the database connection and migrations are complete:

curl https://your-domain/api/v1/health/ready

If readiness fails, check that the database container is healthy:

docker compose ps

Access the Web UI

Open your browser to your configured domain (or http://localhost if using the default SITE_ADDRESS).

You should see the MemGhost setup wizard on first visit. If the frontend is not yet ready, give it a moment — the web container starts after the API is available.

Test the API

List notes (should return an empty array on a fresh install):

curl -s https://your-domain/api/v1/notes | head -c 200

Verify AI Features

If you enabled the ai profile, check that Ollama is running and models are loaded:

# Check Ollama is responding
docker compose exec ollama ollama list

# Test the health endpoint
curl -s https://your-domain/api/v1/health/ready

If models haven’t been pulled yet, see First Run for the model download commands.

Verify Voice Features

If you enabled the voice profile:

# Check TTS service
docker compose logs kokoro | tail -5

# Check STT service
docker compose logs whisper | tail -5

The voice toggle buttons appear in the chat UI header when TTS and STT services are available.

Container Status

View all running containers and their health:

docker compose ps

Expected output for a full install (all profiles):

Service	Status	Notes
db	Up (healthy)	PostgreSQL with pgvector
api	Up	Go backend
web	Up	Next.js frontend
caddy	Up	Reverse proxy
ollama	Up	LLM inference (ai profile)
kokoro	Up	TTS (voice profile)
whisper	Up	STT (voice profile)

View Logs

# All services
docker compose logs -f

# Single service
docker compose logs -f api

Troubleshooting

API container won’t start

Check the migrate service completed successfully:

docker compose logs migrate

Common causes:

Database not yet ready — the API waits for the migrate service, which waits for db to be healthy.
Migration conflict — run docker compose down -v to reset the database and try again.

Frontend shows blank page or errors

The frontend proxies API requests through Caddy. If using the standalone profile, make sure Caddy is running:

docker compose logs caddy

Ollama out of memory

The default chat model (qwen3:8b) needs ~5 GB RAM. On memory-constrained systems, try a smaller model:

docker compose exec ollama ollama pull qwen3:4b

Then update AI_LLM_MODEL=qwen3:4b in your .env and restart the API.

Port conflicts

If ports 80 or 443 are already in use, change them in .env:

PORT=8080
HTTPS_PORT=8443