silo

README

Silo CLI

CLI tool to deploy and manage Silo Box on customer hardware.

Silo Box is a self-hosted AI chat application. This CLI installs and orchestrates Docker containers: PostgreSQL, backend, frontend, llama.cpp inference engine, and proxy agent.

Requirements

• Docker 20.10+ with Compose v2
• Linux (Debian/Ubuntu) or macOS
• User in docker group (Linux) or Docker Desktop (macOS)
• 5GB+ free disk space

Installation

Quick Install

curl -fsSL https://raw.githubusercontent.com/EternisAI/silo/main/scripts/install.sh | bash

Build from Source

git clone https://github.com/EternisAI/silo.git
cd silo
make build
cp bin/silo ~/.local/bin/

Add to PATH if needed (~/.bashrc or ~/.zshrc on macOS):

export PATH="$HOME/.local/bin:$PATH"

Commands

Start Services

silo up                    # first run installs, subsequent runs start
silo up --port 8080        # custom port (first install only)
silo up --image-tag 0.1.3  # specific version

Services auto-restart on system reboot (uses restart: unless-stopped).

Stop Services

silo down                  # stops containers, preserves data

Status

silo status                # show deployment and service status

Logs

silo logs                  # view all logs
silo logs -f               # follow logs in real-time
silo logs --tail 50        # show last 50 lines
silo logs backend          # logs for specific service

Upgrade

silo upgrade               # pull latest images and recreate containers
silo upgrade --json        # JSON output for automation

Important: Upgrade only updates Docker images (backend, frontend). It preserves:

• Configuration files
• Database volumes
• Model files
• Application data

To change the LLM model after upgrade:

1. Place new .gguf file in ~/.local/share/silo/data/models/
2. Edit inference_model_file in ~/.config/silo/config.yml
3. Run silo down && silo up

Check Configuration

silo check                 # validate config and installation state

Version

silo version               # show CLI and application versions
silo version --json        # JSON output

Daemon (Remote Control)

HTTP API for remote management over LAN:

make build-daemon
./bin/silod                                    # localhost only
SILO_DAEMON_BIND_ADDRESS=0.0.0.0 ./bin/silod  # LAN access

API Endpoints

curl http://localhost:9999/health                                      # Health check
curl http://localhost:9999/status | jq                                 # Container status
curl http://localhost:9999/api/v1/version | jq                         # Version info
curl -X POST http://localhost:9999/api/v1/up                           # Start/install
curl -X POST http://localhost:9999/api/v1/down                         # Stop
curl -X POST http://localhost:9999/api/v1/restart -d '{"service":"backend"}'
curl -X POST http://localhost:9999/api/v1/upgrade                      # Upgrade
curl "http://localhost:9999/api/v1/logs?service=backend&lines=50"
curl http://localhost:9999/api/v1/check | jq                           # Validate config

Configuration

Edit ~/.config/silo/config.yml then restart:

silo down && silo up

Key Settings

Setting	Default	Description
port	80	Frontend port
image_tag	0.1.2	Docker image version
inference_model_file	GLM-4.7-Q4_K_M.gguf	LLM model file
inference_gpu_layers	999	GPU layers (999 = all)
inference_context_size	8192	Token context window
inference_gpu_devices	"0", "1", "2"	GPU device IDs

Directory Structure

~/.local/bin/silo                      # CLI binary
~/.config/silo/config.yml              # Configuration file
~/.local/share/silo/
├── docker-compose.yml                 # Generated compose file
├── state.json                         # Installation state
└── data/
    ├── models/                        # LLM model files
    └── postgres/                      # Database data

Global Flags

-v, --verbose              # Enable debug logging
--config-dir PATH          # Custom config directory

Uninstall

silo down
rm -rf ~/.config/silo ~/.local/share/silo ~/.local/bin/silo
docker volume prune

Development

make build                 # Build CLI binary to bin/silo
make build-daemon          # Build daemon binary to bin/silod
make test                  # Run tests
make fmt                   # Format code
make lint                  # Run golangci-lint
make dev ARGS="up"         # Run CLI without building

Releasing

Releases are automated via GitHub Actions. Do not manually create tags.

Creating a Release

1. Push your changes to main (via PR or direct push)
2. Go to Actions → Release → Run workflow
3. Choose version bump type:
   • patch (default): Bug fixes, minor updates (0.1.2 → 0.1.3)
   • minor: New features, backward compatible (0.1.2 → 0.2.0)
   • major: Breaking changes (0.1.2 → 1.0.0)
4. Click Run workflow

What Happens

1. Tag created - Auto-increments version based on bump type
2. GoReleaser runs - Builds binaries for Linux/macOS (amd64/arm64)
3. GitHub Release created - With binaries, checksums, and changelog
4. Install script updated - Users running curl | bash get the new version

Release Artifacts

Each release includes:

• silo_<version>_linux_amd64.tar.gz
• silo_<version>_linux_arm64.tar.gz
• silo_<version>_darwin_amd64.tar.gz
• silo_<version>_darwin_arm64.tar.gz
• checksums.txt

CI Pipeline

Pull requests to main trigger CI checks:

• Build (both silo and silod)
• Tests
• Linting (golangci-lint)

Related Projects

• silo_box/ — The application this CLI deploys (Go backend + Next.js frontend)