Skip to content

01-installation.md

Latest commit

 

History

History
364 lines (279 loc) · 10.5 KB

01-installation.md

File metadata and controls

364 lines (279 loc) · 10.5 KB
Raw

Installation Guide

Complete installation instructions for AutoBot. Two deployment methods: bare-metal (systemd) or Docker.

Prerequisites

System Requirements

Component Minimum Recommended
OS Debian/Ubuntu or WSL2 Ubuntu 22.04 LTS
RAM 16 GB 32 GB+
CPU 8 cores 16+ cores
Storage 50 GB 200 GB+ (model + knowledge base storage)
GPU — NVIDIA (CUDA) or Intel NPU for accelerated inference
Python 3.12 3.12 (installed automatically by installer)
Node.js 18 20 (installed automatically by installer)

Required Permissions (Bare-Metal Install)

  • Root access (sudo) — the installer must run as root
  • systemd — required for service management (PostgreSQL, nginx, SLM backend)
  • Internet access — outbound HTTPS to github.com and package repositories (deb.nodesource.com, PPA)

WSL2 users: systemd must be enabled before running the installer. Add the following to /etc/wsl.conf and restart WSL:

[boot]
systemd=true

What the Installer Creates

Resource Purpose
autobot system user Runs all services (with passwordless sudo)
/opt/autobot/ Application base directory
/var/log/autobot/ Installation and service logs
/etc/autobot/ Secrets and configuration
SSH key pair Fleet management (/home/autobot/.ssh/autobot_key)
Self-signed TLS cert HTTPS for nginx

Installation Methods

Method 1: Bare-Metal Install (Recommended)

Installs all services directly on the host using systemd:

git clone https://github.com/mrveiss/AutoBot-AI.git
cd AutoBot-AI
sudo ./install.sh              # Interactive install
sudo ./install.sh --unattended # Unattended (CI/automation)

Options:

Flag Description
--unattended No prompts, use all defaults
--reinstall Force reinstall over existing installation
--branch=BRANCH Git branch to install (default: Dev_new_gui)
--admin-pass=PASS SLM admin password (auto-generated if not set)

The installer will:

  1. Verify root access, systemd, disk space (5 GB+), memory (2 GB+), and internet
  2. Install system packages (Python 3.14, Node.js 20, nginx, Ansible, build tools, libpq)
  3. Create the autobot system user with passwordless sudo
  4. Clone the repository and distribute code to service directories
  5. Run Ansible deployment for the SLM stack (PostgreSQL, backend, nginx)
  6. Verify all services are running and healthy
  7. Display admin credentials and save them to /root/autobot-credentials.txt

Takes 10-20 minutes depending on internet speed. Logs: /var/log/autobot/.

Post-install access:

Service URL
SLM Admin https://<server-ip>/
Service Logs journalctl -u autobot-slm-backend -f

Method 2: Docker (Single Node)

Run the entire stack on one machine with Docker Compose. No system users or packages installed on the host — Docker handles all isolation.

Requirements: Docker Engine 24+ and Docker Compose v2.

git clone https://github.com/mrveiss/AutoBot-AI.git
cd AutoBot-AI

# Core services (backend, SLM, frontend, Redis, PostgreSQL, ChromaDB)
docker compose --env-file docker/.env.docker up -d --build

# Include local Ollama LLM
docker compose --env-file docker/.env.docker --profile ollama up -d --build

# Include Prometheus + Grafana monitoring
docker compose --env-file docker/.env.docker --profile monitoring up -d --build
Service URL
Frontend http://localhost
SLM Admin http://localhost/slm
Backend API http://localhost/api
RedisInsight http://localhost:8001
Ollama http://localhost:11434 (if --profile ollama)
Grafana http://localhost:3000 (if --profile monitoring)

Dev mode with hot reload:

# Optional: cp docker-compose.override.example.yml docker-compose.override.yml
docker compose up -d

LLM Provider Setup

Choose one or more LLM providers for AutoBot:

Option A: Ollama (Local, Recommended)

Installation:

# Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh

# Start Ollama service
ollama serve &

# Pull recommended models
ollama pull tinyllama:latest    # Fast, low resource
ollama pull phi:2.7b           # Better quality
ollama pull llama2:7b          # High quality (requires 8GB+ RAM)

Configuration: Ollama runs on http://localhost:11434 by default - no additional config needed.

Option B: OpenAI API

Setup:

# Set API key as environment variable
export OPENAI_API_KEY="your-openai-api-key-here"  # pragma: allowlist secret

# Or add to your shell profile
echo 'export OPENAI_API_KEY="your-api-key"' >> ~/.bashrc  # pragma: allowlist secret
source ~/.bashrc

Configuration: Configure via .env or environment variables:

llm_config:
  default_llm: "openai_gpt35"
  openai:
    api_key: ""  # Uses environment variable
    models:
      - "gpt-3.5-turbo"
      - "gpt-4"

Option C: Anthropic Claude

Setup:

# Set API key
export ANTHROPIC_API_KEY="your-anthropic-api-key-here"  # pragma: allowlist secret
echo 'export ANTHROPIC_API_KEY="your-api-key"' >> ~/.bashrc  # pragma: allowlist secret

Configuration: Configure via .env or environment variables:

llm_config:
  default_llm: "anthropic_claude"
  anthropic:
    api_key: ""  # Uses environment variable
    models:
      - "claude-3-sonnet-20240229"
      - "claude-3-haiku-20240307"

GPU Acceleration Setup (Optional)

NVIDIA GPU Setup

Install CUDA:

# Download and install CUDA 11.8 (or latest compatible)
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin
sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/11.8.0/local_installers/cuda-repo-ubuntu2004-11-8-local_11.8.0-520.61.05-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu2004-11-8-local_11.8.0-520.61.05-1_amd64.deb
sudo cp /var/cuda-repo-ubuntu2004-11-8-local/cuda-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update
sudo apt-get -y install cuda

Install PyTorch with CUDA:

# Activate the AutoBot virtual environment
source /opt/autobot/autobot-slm-backend/venv/bin/activate

# Install PyTorch with CUDA support
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

AMD GPU Setup (ROCm)

Install ROCm:

# Add ROCm repository
wget -q -O - https://repo.radeon.com/rocm/rocm.gpg.key | sudo apt-key add -
echo "deb [arch=amd64] https://repo.radeon.com/rocm/apt/debian/ ubuntu main" | sudo tee /etc/apt/sources.list.d/rocm.list
sudo apt update
sudo apt install rocm-dkms rocm-dev rocm-libs

Install PyTorch with ROCm:

source /opt/autobot/autobot-slm-backend/venv/bin/activate
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.4.2

Intel NPU Setup (OpenVINO)

Install OpenVINO:

# Download and install OpenVINO toolkit
wget https://storage.openvinotoolkit.org/repositories/openvino/packages/2024.0/linux/l_openvino_toolkit_ubuntu20_2024.0.0.14509.34caeefd078_x86_64.tgz
tar -xzf l_openvino_toolkit_ubuntu20_2024.0.0.14509.34caeefd078_x86_64.tgz
sudo mv l_openvino_toolkit_ubuntu20_2024.0.0.14509.34caeefd078 /opt/intel/openvino_2024
echo 'source /opt/intel/openvino_2024/setupvars.sh' >> ~/.bashrc

Verification

Bare-Metal Install

The installer runs its own verification (Phase 5). After completion, confirm services are healthy:

# Check systemd services
systemctl status autobot-slm-backend --no-pager
systemctl status postgresql --no-pager
systemctl status nginx --no-pager

# Verify HTTPS endpoint
curl -sk https://127.0.0.1/api/health | jq

# Check recent logs for errors
journalctl -u autobot-slm-backend -n 20 --no-pager

Access the web interface:

Service URL
SLM Admin https://<server-ip>/

Credentials are displayed at the end of the install and saved to /root/autobot-credentials.txt.

Docker Install

# Verify all containers are running
docker compose ps

# Check backend health
curl -s http://localhost/api/health | jq

# View logs
docker compose logs -f --tail=50

Access the web interface:

Service URL
Frontend http://localhost
SLM Admin http://localhost/slm
Backend API http://localhost/api

GPU Verification (Optional)

# NVIDIA
nvidia-smi

# PyTorch CUDA check (inside the AutoBot venv)
source /opt/autobot/autobot-slm-backend/venv/bin/activate
python -c "import torch; print(f'CUDA available: {torch.cuda.is_available()}')"

Troubleshooting Installation

Common Issues

Issue: Permission denied running install.sh

# The installer requires root
chmod +x install.sh
sudo ./install.sh

Issue: systemd not running (WSL2)

# Enable systemd in WSL2
sudo tee /etc/wsl.conf <<'EOF'
[boot]
systemd=true
EOF
# Then restart WSL from PowerShell: wsl --shutdown

Issue: Python 3.14 not found

# Install Python 3.14 via deadsnakes PPA
sudo add-apt-repository ppa:deadsnakes/ppa
sudo apt-get update
sudo apt-get install -y python3.14 python3.14-venv python3.14-dev

Issue: Node.js/npm not found

# Install Node.js 20.x LTS
curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt-get install -y nodejs

Issue: Ansible deployment failed

# Check the install log for details
tail -100 /var/log/autobot/install-*.log

# Re-run just the Ansible phase
cd /opt/autobot/code_source/autobot-slm-backend/ansible
ansible-playbook -i inventory/localhost.yml playbooks/deploy-slm-manager.yml --skip-tags seed,provision

Issue: GPU not detected

# Check GPU status
nvidia-smi  # For NVIDIA
rocm-smi    # For AMD

# Verify drivers
lsmod | grep nvidia  # Should show nvidia modules

Getting Help

If installation fails:

  1. Check logs: tail -100 /var/log/autobot/install-*.log
  2. Verify prerequisites: Ensure all system requirements are met
  3. Try Docker: Use Method 2 as an alternative to bare-metal
  4. GitHub Issues: Report installation problems with system details at https://github.com/mrveiss/AutoBot-AI/issues

Next Steps

After successful installation:

  1. Quick Start Guide - Your first conversation
  2. Settings Guide - Configuration and preferences
  3. User Guides - Complete user documentation