The AI stack (p4n4-ai) provides local LLM inference (Ollama), agent memory (Letta),
and workflow automation (n8n). It attaches to p4n4-net as an external network.
| Service | Image | Port | Role |
|---|---|---|---|
| Ollama | ollama/ollama:latest |
11434 | Local LLM runtime |
| Letta | letta/letta:latest |
8283 | AI agent framework with memory |
| n8n | n8nio/n8n:latest |
5678 | Workflow automation |
The IoT stack must be running (to provide p4n4-net) or the network must exist:
docker network create p4n4-net # if not using p4n4-iot
Or start everything with: p4n4 up (starts all enabled stacks in dependency order, so p4n4-iot creates the network first)
./scripts/pull-models.sh llama3.2
# or
docker exec p4n4-ollama ollama pull llama3.2
Starter workflows are in config/n8n/workflows/:
| Workflow | Description |
|---|---|
alert-enrichment.json |
Subscribes to inference/+/result; sends low-confidence results to Ollama for analysis |
scheduled-digest.json |
Hourly telemetry summary from InfluxDB via Ollama |
device-onboarding.json |
Listens on devices/+/register; registers new devices and confirms over MQTT |
incident-escalation.json |
Listens on alerts/+/critical; classifies severity via Ollama and publishes to alerts/escalated |
Import them via the n8n UI or mount the directory as a volume.
Uncomment the deploy.resources block in docker-compose.override.yml for NVIDIA GPU.
| Variable | Description |
|---|---|
N8N_BASIC_AUTH_USER / _PASSWORD |
n8n UI credentials |
N8N_ENCRYPTION_KEY |
n8n data encryption key |
LETTA_SERVER_PASSWORD |
Letta API password |
INFLUXDB_TOKEN / INFLUXDB_ORG / INFLUXDB_BUCKET |
Shared with IoT stack (must match) |