AI-Suite

Name: ai-suite
Author: trevorsandy

AI-Suite is intended to provide an end-to-end path from zero to working
AI workflows and agents for developers and those who want to enable a local,
private AI solution.

It provides an open, curated, pre-configured Docker Compose configuration
file along with Python and Bash shell scripts that bootstraps fully featured
Local AI Agents, personal AI Assistant and a Low/No Code environment on a self-hosted
platform enabling users to focus on building solutions that employ robust AI workflows
and agents.

Portions of AI-Suite extends Cole Medin's
Self-hosted AI Package
which is built on the n8n-io
Self-hosted AI Starter Kit.

n8n.io - n8n

Curated by Trevor SANDY - https://github.com/trevorsandy.

What’s included

✅ Self-hosted n8n - Automation platform with over 400
integrations and advanced AI components.

✅ Self-hosted OpenClaw - a personal AI assistant
you run on your own devices.

✅ Open WebUI - ChatGPT-like interface to
privately interact with your local models and N8N agents.

✅ OpenCode - open source agent that helps you write
code in your terminal.

✅ Ollama - Cross-platform LLM platform to install
and run the latest LLMs.

✅ LLaMA.cpp - Cross-platform LLaMA.cpp
HTTP Server platform to install and run the latest LLMs in gguf format.

✅ Supabase - Open source database as a service,
most widely used database for AI agents.

✅ Flowise - No/low-code AI agent builder that
pairs very well with n8n.

✅ Qdrant - Open source, high performance vector
store with an comprehensive API.

✅ PostgreSQL - Workhorse of the Data
Engineering world, backend for Langfuse.

✅ MCP Gateway - Reverse proxy
and management layer for MCP servers.

✅ Neo4j - Knowledge graph engine that powers tools
like GraphRAG, LightRAG, and Graphiti.

✅ Redis (Valkey) - High-performance key/value datastore,
supports caching and message queues workloads.

✅ SearXNG - Open source internet metasearch
engine, aggregates results from up to 229 search services.

✅ Langfuse - Open source LLM engineering platform
for agent observability.

✅ MinIO - High-performance, S3-compatible object
storage solution.

✅ ClickHouse - Open source, database management
system that can generate analytical data reports in real-time.

✅ Caddy - Managed HTTPS/TLS for custom domains.

✅ Nginx - HTTPS/TLS server, reverse proxy, TCP/UDP
proxy server

✅ Authelia - Authentication and authorization
server, identity and access management (IAM).

Prerequisites

System specifications:

32GB RAM recommended (8GB minimum)
40GB free disk space

Before you begin, make sure you have the following software installed:

Git - For easy repository management.
Python 3.10+ - To run the setup script.
Node 22.16+ - For auto-configuration and OpenClaw runtime.

Docker 20.10+ - Required to
setup and run all AI-Suite services.

Docker Compose commands

By default, AI-Suite automatic configuration will validate and, if needed,
install Docker. However, you can also manually install Docker in advance.

If you are using a machine without the docker compose application available
by default, run these commands to install Docker compose:

DOCKER_COMPOSE_VERSION=$(curl -s https://api.github.com/repos/docker/compose/
releases/latest | grep 'tag_name' | cut -d\\" -f4)
sudo curl -L "https://github.com/docker/compose/releases/download/${DOCKER_COMPOSE_VERSION}/docker-compose-linux-x86_64" -o /usr/local/bin docker-compose
sudo chmod +x /usr/local/bin/docker-compose
sudo mkdir -p /usr/local/lib/docker/cli-plugins
sudo ln -s /usr/local/bin/docker-compose /usr/local/lib/docker/cli-plugins/docker-compose

Developers may also consider the following optional software:

VSCode - Python and Bash shell development.
GitKraken - Superior Git SCM platform

Installation

Step 1: Clone the repository and set environment variables

Clone the repository and navigate to the project directory:

git clone https://github.com/trevorsandy/ai-suite.git
cd ai-suite

AI-Suite will automatically configure the settings and environment variables
including generation of secret passwords, tokens and keys needed to successfully
perform installation and startup operations.

However, it is good practice to review your .env environment variables before
running install or update taking into account your installation platform
components, specifications and requirements. Particularly pay attention to the
Ollama or LLaMA.cpp (depending on which LLM you are using) configuration
settings and Windows WSL (posix) versus native paths.

Environment variables

Optionally, you can make a copy of .env.example renamed to .env in the
project directory.

cp .env.example .env

Credentials

If you install Supabase, all credentials will be auto-generated. If you
prefer to manually setup the Supabase credentials, you may use the their
self-hosting guide.

.env.example file

# Change this file  name to .env after updating it if not using auto-configuration!

############
# Auto-Configuration:
#   AI-Suite uses this file as the .env template. You should update default settings
#   you wish to set before running install or update using suite_services.py.
#
#   If an existing .env is encountered during auto-configuration, defaults from this
#   file are overlayed with the existing .env values. This means secrets for variables
#   in the existing .env will not be generated by AI-Suite during install or update.
#
#   Variables that hold generated secrets will have a specific default value
#   format: <variable>=generate using <generator>[:<argument>]
#   Examples: N8N_RUNNERS_AUTH_TOKEN=generate using gen_hex:32
#             SERVICE_ROLE_KEY=generate using gen_token:service_role_sym
#             PROXY_AUTH_PASSWORD=generate using gen_bcrypt
#
# Generating Credentials:
#   All secrets are generated when using auto-configure except N8N_ENCRYPTION_KEY
#   which can also be set by exporting the environment variable or by placing the
#   key=value pair in n8n/.n8n.encryption.key.

#   When using your existing n8n encryption key placed in n8n/.n8n.encryption.key,
#   be sure to properly terminate the line entry with a new line (hit enter key to
#   move your cursor to the next line). Also ensure the file format is LF (Unix)
#   and not CRLF (Windows).
#
# OpenSSL: Available by default on Linux/Mac via command `openssl rand -hex 32`
#   For Windows, use 'WSL2', 'Git Bash' terminal installed with git or from cmd
#   run the command: python -c "import secrets; print(secrets.token_hex(32))"
#
# Password: Use Python command to generate 16-character strong password:
#   python3 -c "import secrets;import string; alphabet = string.ascii_letters + string.digits;\
#               password = ''.join(secrets.choice(alphabet) for i in range(16));\
#               print(password)"
#
# JWT Tokens: Use https://jwtsecrets.com/#generator to generate keys and tokens
#   ranging from 8 to 128 characters long.
############

############
# [required for Auto-Configuration] - automatically set when enabled (AC=True)
# Access Control - Proxy, Identity and Access Management configuration
############

# Enable proxy, identity and access auto-configure mode -credentials are auto-generated
AC=True
# Your public/private domain name. An arbitrary name is allowed for private domain
AC_DOMAIN=local.pc
# Configure AI-Suite as a local (private) vs. global (public) installation
AC_LOCAL=True
# The reverse proxy to use (Caddy or Nginx)
AC_PROXY=caddy
# User name for PROXY configuration (alphanumeric characters only)
AC_USERNAME=AISuiteProxyUser
# User password for PROXY configuration
# Keep default '*******' to trigger password prompt during setup
AC_PASSWORD='*******'
# Send confirmation email on user registration - SMTP server required
AC_CONFIRM=False
# Enable Authelia 2FA (two factor authentication) support
AC_WITH_AUTHELIA=True
# User email address for Authelia - required if AC_WITH_AUTHELIA=True
[email protected]
# User display name for Authelia - required if AC_WITH_AUTHELIA=True (alphanumeric chars and spaces only)
AC_DISPLAY_NAME='AI Suite Authelia User'
# Use Redis with Authelia - recommended if AC_WITH_AUTHELIA=True and public
AC_WITH_REDIS=False
# Auto-configuration runtime log relative path without filename
AC_LOG_PATH=./access

############
# [required] - automatically set when auto-configure (AC=True) is enabled
# n8n credentials - use OpenSSL `openssl rand -hex 32` for all
############

# Master key used to encrypt sensitive credentials that n8n stores
N8N_ENCRYPTION_KEY=generate using gen_n8ncrypt
# Shared secret between n8n containers and runners sidecars
N8N_RUNNERS_AUTH_TOKEN=generate using gen_hex:32
# Specific JWT secret. By default, n8n generates one on start
N8N_USER_MANAGEMENT_JWT_SECRET=generate using gen_hex:32

############
# [required] - automatically set when auto-configure (AC=True) is enabled
# PostgreSQL database user password - use OpenSSL `openssl rand -hex 16`
############

POSTGRES_PASSWORD=generate using gen_hex:16

# Following settings are required if you enable the respective module.

   #
   #
#######
 #####
   #

############
# [required for Supabase] - automatically set when auto-configure (AC=True) is enabled
# Supabase Secrets

# Read these docs for any help: https://supabase.com/docs/guides/self-hosting/docker
# For the JWT Secret and keys, see: https://supabase.com/docs/guides/self-hosting/docker#generate-api-keys
# For the other secrets, see: https://supabase.com/docs/guides/self-hosting/docker#update-secrets

# Note that using special symbols (like '%') can complicate things a bit for your Postgres password.
# If you use special symbols in your Postgres password, you must remember to percent-encode your password later if using the
# Postgres connection string, for example, postgresql://postgres.projectref:p%[email protected]:6543/postgres
#
# To enable the new asymmetric key support, uncomment these lines in docker-compose.yml:
#   Auth    : GOTRUE_JWT_KEYS: ${JWT_KEYS:-[]}
#   Realtime: API_JWT_JWKS: ${JWT_JWKS:-{"keys":[]}}
#   Storage : JWT_JWKS: ${JWT_JWKS:-{"keys":[]}}
############

# Legacy symmetric HS256 key
JWT_SECRET=generate using gen_key:secret
# Legacy symmetric API key (HS256-signed JWT) for anon role.
ANON_KEY=generate using gen_key:anon_sym
# Legacy symmetric API key (HS256-signed JWT) for service role.
SERVICE_ROLE_KEY=generate using gen_key:service_role_sym
# Pre-signed ES256 JWT "API key" for anon role.
ANON_KEY_ASYMMETRIC=generate using gen_key:anon_asym
# Pre-signed ES256 JWT "API key" for service role.
SERVICE_ROLE_ASYMMETRIC=generate using gen_key:service_role_asym
# Opaque API key for client-side use (anon role).
SUPABASE_PUBLISHABLE_KEY=generate using gen_key:client
# Opaque API key for server-side use (service_role). Never expose in client code.
SUPABASE_SECRET_KEY=generate using gen_key:server
# JSON array of signing JWKs (EC private + legacy symmetric).
# Used by Auth.
JWT_KEYS=generate using gen_key:keys
# JWKS for token verification (EC public + legacy symmetric).
# Used by PostgREST, Realtime, Storage to verify tokens.
JWT_JWKS=generate using gen_key:jwks
# Used by Realtime and Supavisor
SECRET_KEY_BASE=generate using gen_token:48
# Used by Supavisor
VAULT_ENC_KEY=generate using gen_hex:16
# Used by Studio to access Postgres via postgres-meta
PG_META_CRYPTO_KEY=generate using gen_token:24
# Used by Kong dashboard user
DASHBOARD_PASSWORD=generate using gen_hex:16

############
# [required for Supabase] - automatically set when auto-configure (AC=True) is enabled
# Logs - Configuration for Supabase Analytics
# Please refer to https://supabase.com/docs/reference/self-hosting-analytics/introduction
############

# Change vector.toml sinks to reflect this change
# These cannot be the same value
# Must be at least 32 characters; generate with 'openssl rand -base64 24'
LOGFLARE_PUBLIC_ACCESS_TOKEN=generate using gen_token:24
LOGFLARE_PRIVATE_ACCESS_TOKEN=generate using gen_token:24

############
# [required for Supabase S3] - automatically set when auto-configure (AC=True) is enabled
# S3 - Supabase alternative storage
############

S3_PROTOCOL_ACCESS_KEY_ID=generate using gen_hex:16
S3_PROTOCOL_ACCESS_KEY_SECRET=generate using gen_hex:32

############
# [required for Supabase and Langfuse] - automatically set when auto-configure (AC=True) is enabled
# MinIO - authentication configuration - use OpenSSL `openssl rand -hex 16`
############

MINIO_ROOT_PASSWORD=generate using gen_hex:16

############
# [required for Flowise] - automatically set when auto-configure (AC=True) is enabled
# Flowise - authentication configuration - use OpenSSL `openssl rand -hex 16`
############

FLOWISE_PASSWORD=generate using gen_hex:16

############
# [required for Neo4j] - automatically set when auto-configure (AC=True) is enabled
# Neo4j admin username and password
# The admin username must remain "neo4j".
# Replace "password" with your chosen password.
# Keep the "/" as a separator between the two.
############

NEO4J_PASSWORD=generate using gen_hex:16
NEO4J_AUTH=neo4j/${NEO4J_PASSWORD}

############
# [required for Langfuse] - automatically set when auto-configure (AC=True) is enabled
# Langfuse credentials
# Each of the secret keys you can set to whatever you want, just make it secure!
# For salt, secret and encryption key, use OpenSSL command specified above
############

CLICKHOUSE_PASSWORD=generate using gen_hex:16
LANGFUSE_SALT=generate using gen_hex:16
NEXTAUTH_SECRET=generate using gen_token:32
ENCRYPTION_KEY=generate using gen_hex:32

# Following settings are required for production.

   #
   #
#######
 #####
   #

############
# [required for production if using Authelia]
# Automatically set when auto-configure (AC=True) is enabled
# Authelia Config
############

AUTHELIA_SESSION_SECRET=generate using gen_hex:32
AUTHELIA_STORAGE_ENCRYPTION_KEY=generate using gen_hex:32
AUTHELIA_IDENTITY_VALIDATION_RESET_PASSWORD_JWT_SECRET=generate using gen_hex:32

AUTHELIA_SCHEMA=authelia

############
# [required for production if using TLS Proxy]
# Automatically set when auto-configure (AC=True) is enabled
# Config for optional Caddy or Nginx reverse proxy with Let's Encrypt
############

# Generated bcrypt password for basic authentication without Authelia
PROXY_AUTH_PASSWORD=generate using gen_bcrypt
PROXY_AUTH_USERNAME=AISuiteProxyUser

############
# [required for production if using Caddy]
# Automatically set when auto-configure (AC=True) is enabled
# Caddy Config

# By default listen on https://localhost:[service port] and don't use an email for SSL
# To change this for production:
# Uncomment all of these environment variables for the services you want exposed
# Note that you might not want to expose Ollama or SearXNG since they aren't secured by default
############

# Domain name for the proxy (must point to your server)
PROXY_DOMAIN=${AC_DOMAIN}

# WEBUI_HOSTNAME=openwebui.${AC_DOMAIN}
# N8N_HOSTNAME=n8n.${AC_DOMAIN}
# OPENCLAW_HOSTNAME=openclaw.${AC_DOMAIN}
# FLOWISE_HOSTNAME=flowise.${AC_DOMAIN}
# SUPABASE_HOSTNAME=supabase.${AC_DOMAIN}
# LANGFUSE_HOSTNAME=langfuse.${AC_DOMAIN}
# OLLAMA_HOSTNAME=ollama.${AC_DOMAIN}
# LLAMACPP_HOSTNAME=llamacpp.${AC_DOMAIN}
# SEARXNG_HOSTNAME=searxng.${AC_DOMAIN}
# NEO4J_HOSTNAME=neo4j.${AC_DOMAIN}
# WEBHOOK_URL=https:n8n.${AC_DOMAIN}
# LETSENCRYPT_EMAIL=${AC_EMAIL}

############
# [required for production if using Nginx]
# Automatically set when auto-configure (AC=True) is enabled
# Ngnix Config
############

NGINX_SERVER_NAME=generated-primary-hostname
[email protected]
# This must be set to 0 for public installation
USE_LOCAL_CA=1

# Everything below this point is optional.
# Default values will suffice unless you need more features/customization.

...

OpenClaw setup .env configuration

############
# OpenClaw - Setup
############

# Set to specific release tag, 'commit' for latest commit or empty for latest release
OPENCLAW_RELEASE=''
# Set to 1 to enable onboarding during setup
OPENCLAW_ONBOARDING=0
# Set to 0 for default Docker setup
OPENCLAW_DOCKER_SANDBOX=1
# Set to 0 for pre-built image download when sandbox is disabled
OPENCLAW_DOCKER_LOCAL_IMAGE=1
# Preserve locally updated files when performing clone operation
OPENCLAW_KEEP_LOCAL_UPDATES=0

Ollama .env configuration

############
# Ollama - LLM
############

OLLAMA_PORT=11434

# Docker backend connect when running Ollama in the Host:
#OLLAMA_HOST=host.docker.internal:${OLLAMA_PORT}
# When accessing Ollama from the Host:
OLLAMA_HOST=localhost:${OLLAMA_PORT}
# When running Ollama in Docker:
#OLLAMA_HOST=ollama:${OLLAMA_PORT}

# Tuning
OLLAMA_CONTEXT_LENGTH=4096
OLLAMA_FLASH_ATTENTION=1
OLLAMA_KV_CACHE_TYPE=q4_0
OLLAMA_MAX_LOADED_MODELS=2

# Models
OLLAMA_DEFAULT_MODEL=llama3.2
OLLAMA_SUPPLEMENT_MODEL=qwen3:8b
OLLAMA_EMBEDDING_MODEL=nomic-embed-text

# Ollama server arguments - use ollama serve --help for available 'serve' arguments
OLLAMA_SERVER_ARGS=serve

############
# LLAMA (Ollama/LLaMA.cpp) - Shared environment variables
############

# Application Installation path
# Set for LLaMA.cpp or if using custom Ollama installation path
# e.g. LLAMA_PATH=~\Projects\ai-suite\llama.cpp\bin\llama-server.exe
# Omit '<value>' to return 'False' when queried
LLAMA_PATH=

LLaMA.cpp .env configuration

############
# LLaMA.cpp - LLM
############

LLAMA_ARG_PORT=8040

# Docker backend connect when running LLaMA.cpp in the Host:
#LLAMA_ARG_HOST=host.docker.internal
# When running LLaMA.cpp in Docker:
LLAMA_ARG_HOST=0.0.0.0

# Backend connect
LLAMACPP_HOST=${LLAMA_ARG_HOST}:${LLAMA_ARG_PORT}

# Model names - Dictionary keys for model download identifier values below.
# Keys, and values below include an empty slot for a user-defined model
LLAMACPP_MODEL_GEMMA=gemma-4b  # Default
LLAMACPP_MODEL_DEEPSEEK=deepseek-7b
LLAMACPP_MODEL_MISTRAL=mistral-7b
LLAMACPP_MODEL_LLAMA=llama-8b
LLAMACPP_MODEL_QWEN=qwen-8b
LLAMACPP_MODEL_USER=

# Model download identifier - Dictionary values for model keys above.
# Model selected by 'best match' to LLAMACPP_MODEL_NAME
# To specify a local model, change '-hf' to '-m' in LLAMACPP_SERVER_ARGS below
# and replace the respective model id value below with 'models/<model filename>'.
LLAMACPP_MODEL_GEMMA_ID=ggml-org/gemma-3-4b-it-GGUF
LLAMACPP_MODEL_DEEPSEEK_ID=mradermacher/DeepSeek-R1-Distill-Qwen-7B-Uncensored-i1-GGUF
LLAMACPP_MODEL_MISTRAL_ID=bartowski/mistralai_Ministral-3-8B-Instruct-2512-GGUF
LLAMACPP_MODEL_LLAMA_ID=bartowski/allura-forge_Llama-3.3-8B-Instruct-GGUF
LLAMACPP_MODEL_QWEN_ID=bartowski/Qwen_Qwen3-8B-GGUF
LLAMACPP_MODEL_USER_ID=

# Model and paths
LLAMACPP_PATH=llama.cpp
LLAMACPP_DEFAULT_MODEL=${LLAMACPP_MODEL_GEMMA}  # IMPORTANT: should reasonably match dictionary model name above.
LLAMACPP_MODELS_DIR=${LLAMACPP_PATH}/models
LLAMACPP_MODEL_PATH=${LLAMACPP_MODELS_DIR}/${LLAMACPP_DEFAULT_MODEL}

# Model management - automatically download specified model if not downloaded.
LLAMA_ARG_HF_REPO=${LLAMACPP_MODEL_GEMMA_ID}

# Tuning
LLAMA_ARG_CTX_SIZE=4096
LLAMA_ARG_FLASH_ATTN=1
LLAMA_ARG_N_GPU_LAYERS=0
LLAMA_ARG_THREADS=4
LLAMA_ARG_MODELS_MAX=4

# LLaMA.cpp server arguments - use 'llama-server --help' for available arguments
# To specify a local model, append '-m' or '––model'.
# To auto-download model (if not already downloaded) and if LLAMA_ARG_HF_REPO is
# not used (commented), append '-hf' or '--hf-file'.
LLAMACPP_SERVER_ARGS=--jinja

Shared Ollama and LLaMA.cpp .env configuration

############
# LLAMA (Ollama/LLaMA.cpp) - Shared environment variables
############

# Application Installation path
# Set for LLaMA.cpp or if using custom Ollama installation path
# e.g. LLAMA_PATH=~\Projects\ai-suite\llama.cpp\bin\llama-server.exe
# Omit '<value>' to return 'False' when queried
LLAMA_PATH=

# Conecting to LLAMA using OpenAI API connection
# When running Ollama:    ${OLLAMA_HOST}
# When running LLaMA.cpp: ${LLAMACPP_HOST}
OPENAI_API_BASE_URL=${OLLAMA_HOST}
#OPENAI_API_KEY -  OpenAI API key declared below at Studio

You may also choose to generate the AI-Suite requirements.txt file and install
missing Python modules to ensure your environment meets the Python module dependencies.

Generate and install requirements.txt

Optionally, setup a Python virtual environment under ai-suite.

pip install -U virtualenv

Unix

python3 -m venv .venv
source ./.venv/bin/activate

Windows

python -m venv .venv

Powershell

.venv/Scripts/activate.ps1

Console

.venv\Scripts\activate

Then run the following commands from ai-suite:

pip install pipreqs
pipreqs --encoding=utf8 .
pip install -r requirements.txt

[!IMPORTANT]
Make sure to generate secure random values for all secrets. Never use the
example values in production.

Step 2: Run suite_services.py

AI-Suite uses the suite_services.py script for the installation command
that handles the AI-Suite functional module selection, LLAMA (Ollama/LLaMA.cpp)
CPU/GPU configuration, and starting Supabase, OpenClaw and Open WebUI Filesystem
when specified.

Additionally, this script is used to perform operational actions such as stopping
or pausing the running suite stack, OpenClaw clawdock operations and updating container
images.

Command syntax:

suite_services.py --profile <arguments> --environment <argument> --operation
<argument> --log <argument>

[!NOTE]
The following example commands will use the n8n and OpenCode functional
modules. Simply substitute these modules for your desired options if you elect
to use these examples in your environment.

The profile command arguments

Both installation and operation commands utilize the optional --profile
arguments to specify which AI-Suite functional modules and which LLAMA CPU/GPU
configuration to use. When no functional profile argument is specified, the
default functional module open-webui is used, Likewise, if no CPU/GPU configuration
profile argument is specified, it is assumed LLAMA is being run from the Host.
Multiple profile arguments for functional modules are supported.

suite_services.py --profile functional module arguments

Argument	Functional Module
`n8n`	n8n - automation platform
`openclaw`	OpenClaw - your own personal AI assistant
`opencode`	OpenCode - low-code, no-code agent
`open-webui`	Open WebUI - chatbot interface
`open-webui-mcpo`	Open WebUI MCPO - MCP to OpenAPI translator
`open-webui-pipe`	Open WebUI Pipelines - agent tools and functions
`flowise`	Flowise - complementary agent builder
`supabase`	Supabase - alternative database
`searxng`	SearXNG - internet metasearch
`langfuse`	Langfuse - agent observability platform
`neo4j`	Neo4j - knowledge graph
`caddy`	Caddy - managed https/tls server
`n8n-all`	n8n - complete bundle
`open-webui-all`	Open WebUI - complete bundle
`ai-all`	AI-Suite full stack - all modules

suite_services.py --profile LLAMA CPU/GPU in Docker argument:

Argument	LLAMA CPU/GPU
`cpu`	Ollama - run on CPU
`gpu-nvidia`	Ollama - run on Nvidia GPU
`gpu-amd`	Ollama - run on AMD GPU
`cpp-cpu`	LLaMA.cpp - run on CPU
`cpp-gpu-nvidia`	LLaMA.cpp - run on Nvidia GPU
`cpp-gpu-amd`	LLaMA.cpp - run on AMD GPU

Example command:

# Ollama
python suite_services.py --profile n8n opencode gpu-nvidia
# LLaMA.cpp
python suite_services.py --profile n8n opencode cpp-gpu-nvidia

suite_services.py --profile LLAMA running on Host argument:

Argument	LLAMA CPU/GPU
`ollama`	Ollama - run on Host (Default)
`llama.cpp`	LLaMA.cpp - run on Host

Example command:

# Ollama - As the default LLAMA option, the argument is not required
python suite_services.py --profile n8n opencode
# LLaMA.cpp
python suite_services.py --profile n8n opencode llama.cpp

If you intend to install Supabase, before running suite_services.py, setup
the Supabase environment variables using their self-hosting guide.

For Docker LLAMA with Nvidia GPU users

# Ollama
python suite_services.py --profile gpu-nvidia n8n opencode
# LLaMA.cpp
python suite_services.py --profile cpp-gpu-nvidia n8n opencode

[!NOTE]
If you have not used your Nvidia GPU with Docker before, please follow the
Ollama Docker instructions.
LLaMA.cpp Docker instructions

For Docker LLAMA with AMD GPU users

# Ollama
python suite_services.py --profile gpu-amd n8n opencode
# LLaMA.cpp
python suite_services.py --profile cpp-gpu-amd n8n opencode

For LLAMA on Mac running Apple Silicon users

If you're using a Mac with an M1 or newer processor, you cannot expose your GPU
to the Docker instance, unfortunately. There are two options in this case:

Run ai-suite fully on CPU:

# Ollama
python suite_services.py --profile cpu n8n opencode
# LLaMA.cpp
python suite_services.py --profile cpp-cpu n8n opencode

Run LLAMA on your Host for faster inference, and connect to that from the
n8n instance:
```
# Ollama
python suite_services.py --profile n8n opencode
# LLaMA.cpp
python suite_services.py --profile n8n opencode llama.cpp
```
If you want to run LLAMA on your Mac, check the Ollama homepage
or LLaMA.cpp install
for installation instructions.

For LLAMA running on the Host users

If you're running LLAMA on your Host (not in Docker), the suite_services.py
script will automatically set your OLLAMA_HOST/LLAMA_ARG_HOST environment
variable in the .env file. Using interpolation, these settings will also be set
for the n8n service configuration.

To manually configure the Ollama settings and update the x-n8n section in
your .env file:

Manual Ollama .env Host configuration

OLLAMA_HOST=host.docker.internal:11434
#OLLAMA_HOST=ollama:11434

# ... other configurations ...

# When running Ollama in the Host and Open WebUI in Docker:
OLLAMA_BASE_URL=http://host.docker.internal:11434
#OLLAMA_BASE_URL=http://localhost:11434

... or youe Docker Compose file:

x-n8n: &service-n8n
  # ... other configurations ...
  environment:
    # ... other environment variables ...
    - OLLAMA_HOST=host.docker.internal:11434

The suite_services.py script will similarly set the OPENAI_API_BASE_URL
environment variable to use the HOST and PORT of the selected LLAMA LLM
(Ollama/LLaMA.cpp). This option will enable n8n backend connections to
LLaMA.cpp.

Manual LLaMA.cpp .env Host configuration

LLAMA_ARG_PORT=8040

# When running LLaMA.cpp in the host:
#LLAMA_ARG_HOST=host.docker.internal
# When running LLaMA.cpp in Docker:
#LLAMA_ARG_HOST=0.0.0.0
LLAMA_ARG_HOST='host.docker.internal'

# Backend connect
LLAMACPP_HOST=${LLAMA_ARG_HOST}:${LLAMA_ARG_PORT}

# ... other configurations ...

# Conecting to LLAMA using OpenAI API connection
# When running Ollama:    ${OLLAMA_HOST}
# When running LLaMA.cpp: ${LLAMA_ARG_HOST}:${LLAMA_ARG_PORT}
OPENAI_API_BASE_URL='${LLAMACPP_HOST}'

For everyone else (...using CPU)

# Ollama
python suite_services.py --profile n8n opencode cpu
# LLaMA.cpp
python suite_services.py --profile n8n opencode cpp-cpu

[!NOTE]
Script examples beyond this point will use Ollama or LLaMA.cpp interchangeably.

The operation command argument

There are also operation commands that start, stop, stop-llama, pause,
unpause, update and install the AI-Suite services using the optional
--operation argument. A LLAMA (Ollama/LLaMA.cpp) check is performed when
it is assumed LLAMA is running from the Host. If LLAMA is determined to be
installed but not running, an attempt to launch the Ollama/LLaMA.cpp service
is executed on install, start and unpause. The check will also attempt to
stop the running LLAMA service (in addition to stopping the AI-Suite services)
when the stop-llama operational command argument is specified.

suite_services.py ... --operation argument:

Argument	Operation
`start`	Start - start the previously stopped, specified profile containers
`stop`	Stop - shut down the specified profile containers
`stop-llama`	Stop - perform `stop` and shut down Ollama/LLaMA.cpp on Host
`pause`	Pause - pause the specified profile containers
`unpause`	Unpause - unpause the previously paused profile containers
`backup-data`	Backup Data - backup volume mount data to backup file
`restore-data`	Restore Data - restore volume mount data from backup file

Example command:

# Ollama
python suite_services.py --profile n8n opencode gpu-nvidia --operation stop
# LLaMA.cpp
python suite_services.py --profile n8n opencode llama.cpp --operation stop-llama

OpenClaw operation commands are available using clawdock.

suite_services.py --operation clawdock argument

[!NOTE]
Clawdoc commands that require an argument are executed in two parts. First,
the clawdock command: suite_services.py --operation clawdock-approve
will, second, generate an argument prompt: Enter a device pairing: <id>

Basic Operations

Argument	Operation
`clawdock-start`	Start the gateway
`clawdock-stop`	Stop the gateway
`clawdock-restart`	Restart the gateway
`clawdock-status`	Check container status
`clawdock-logs`	View live logs (follows output)

Container Access

Command	Description
`clawdock-shell`	Interactive shell inside the gateway container
`clawdock-cli <command>`	Run OpenClaw CLI commands
`clawdock-exec <command>`	Execute arbitrary commands in the container

Web UI & Devices

Command	Description
`clawdock-dashboard`	Open web UI in browser with authentication
`clawdock-devices`	List device pairing requests
`clawdock-approve <id>`	Approve a device pairing request

Setup & Configuration

Command	Description
`clawdock-fix-token`	Configure gateway authentication token (run once)

Maintenance

Command	Description
`clawdock-update`	Pull latest, rebuild image, and restart (one command)
`clawdock-rebuild`	Rebuild the Docker image only
`clawdock-clean`	Remove all containers and volumes (destructive!)

Utilities

Command	Description
`clawdock-health`	Run gateway health check
`clawdock-token`	Display the gateway authentication token
`clawdock-cd`	Jump to the OpenClaw project directory
`clawdock-config`	Open the OpenClaw config directory
`clawdock-show-config`	Print config files with redacted values
`clawdock-workspace`	Open the workspace directory
`clawdock-help`	Show all available commands with examples

Example command:

python suite_services.py --operation clawdock-status

OpenClaw guide (auth, pairing, channels): Getting started

The environment command argument

The --environment command allows the installation to be defined as private
(default) or public. A public install restricts the communication ports exposed
to the network.

The suite_services.py script supports the private (default) and public
environment argument:

private: you are deploying the stack in a safe environment, all AI-Suite
ports are accessible
public: the stack is deployed in a public environment, all AI-Suite ports
except 80 and 443 are closed

suite_services.py ... --environment argument:

Argument	Scope
`private`	Private network
`public`	Public network

Example command:

The AI-Suite stack initialized with...

python suite_services.py --profile n8n opencode cpp-cpu --environment private

is equal to being initialized with:

python suite_services.py --profile gpu-nvidia

The log command argument

The suite_services.py script enables stream (console) logging and setting the
logging level. File logging is always enabled at DEBUG and is not affected
by this argument. The default console logging level is INFO.

environment argument:

suite_services.py ... --log argument:

Argument	Scope
`OFF`	Console logging is disabled
`DEBUG`	Debug logging level
`INFO`	Standard output logging level
`WARNING`	Warning logging level
`ERROR`	Error logging level
`CRITICAL`	Critical logging level

Example command:

python suite_services.py --profile n8n opencode cpp-cpu --operation update --log DEBUG

Auto-configuration, HTTPS Reverse Proxy and Access Management

By default, AI-Suite will automatically configure Caddy (Default) or
Nginx HTTPS reverse proxy and Authelia 2FA (Two Factor Authentication)
IAM (Identity and Access Management) on install or update.
Additionally, auto_configure will generate secrets, the .env file and Docker compose
file updates for AI-Suite modules, including Supabase and OpenClaw.

You can disable this behaviour using
the no-auto-config or manual-configuration profile arguments.

suite_services.py --profile No auto-configure or manual configuration argument:

Argument	Behaviour
`no-auto-config`	Override auto-configure AI-Suite settings
`manual-configuration`	Override auto-configure AI-Suite settings

Example command:

python suite_services.py --profile ai-all no-auto-config --operation update

Deploying to the Cloud

Prerequisite

Linux machine (preferably Unbuntu) with Nano, Git, and Docker installed

Extra steps

Before running the above commands to pull the repo and install everything:

[!WARNING]
ufw does not shield ports published by Docker, because the iptables rules
configured by Docker are analyzed before those configured by ufw. There is a
solution to change this behavior, but that is out of scope for this project.
Just make sure that all traffic runs through the Caddy service via port 443.
Port 80 should only be used to redirect to port 443.

Run the commands as root to open up the necessary ports:
```
ufw enable
ufw allow 80 && ufw allow 443
ufw reload
```
Run the suite_services.py script with the environment argument public
to indicate you are going to run the package in a public environment. The
script will make sure that all ports, except for 80 and 443, are closed
down, e.g.
```
python3 suite_services.py --profile gpu-nvidia --environment public
```
Set up A records for your DNS provider to point your subdomains you'll set
up in the .env file for Caddy to the IP address of your cloud instance.

For example, A record to point n8n to [cloud instance IP] for n8n.yourdomain.com

[!NOTE]
If you are using a cloud machine without the "docker compose" command
available by default such as a Ubuntu GPU instance on DigitalOcean, run these
commands before running suite_services.py:

Docker Compose setup commands

DOCKER_COMPOSE_VERSION=$(curl -s https://api.github.com/repos/docker/compose/releases/latest | grep 'tag_name' | cut -d\\" -f4)
sudo curl -L "https://github.com/docker/compose/releases/download/${DOCKER_COMPOSE_VERSION}/docker-compose-linux-x86_64" -o /usr/local/bin/docker-compose
sudo chmod +x /usr/local/bin/docker-compose
sudo mkdir -p /usr/local/lib/docker/cli-plugins
sudo ln -s /usr/local/bin/docker-compose /usr/local/lib/docker/cli-plugins/docker-compose

⚡️ Quick start and usage

All components of the self-hosted AI-Suite, except if running LLAMA from your
host, is installed through suite_services.py and managed through a Docker Compose
file pre-configured with network and disk so there isn’t much else you need to
install. After completing the installation steps above, follow the steps below
to get started. First, start with n8n.

Use the following settings to confirm or upate n8n Credentials.

Local Ollama service: base URL http://ollama:11434/ (n8n config), http://localhost:11434/
(browser)
Local LLaMA.cpp service: base URL http://llamacpp:8040/ (n8n config), http://localhost:8040/
(browser)
Local QdrantApi database: base URL http://qdrant:6333/ (n8n config), http://localhost:6333/
(browser)
Postgres account: use POSTGRES_HOST, POSTGRES_USER, and POSTGRES_PASSWORD
from your .env file.
Google Drive: This credential is optional. Follow this guide from n8n.

Full list of AI-Suite service endpoints:

Service	Container	Docker connect	Host connect
`n8n`	n8n:5678	http://host.docker.internal:5678	http://localhost:5678
`openclaw-gateway`	openclaw-gateway:18789		http://localhost:18789
`openclaw-cli`	openclaw-cli
`Open WebUI`	open-webui:8080/	http://host.docker.internal:8080/	http://localhost:8080/
`Opencode`	opencode		./opencode/run_opencode_docker.py
`Flowise`	flowise:3001/	http://host.docker.internal:3001/	http://localhost:3001/
`Open webUI MCPO`	open-webui-mcpo:8090/	http://host.docker.internal:8090/	http://localhost:8090/
`Ollama`	ollama:11434/	http://host.docker.internal:11434/	http://localhost:11434/
`LLaMA.cpp`	llamacpp:8040	http://host.docker.internal:8040	http://localhost:8040
`QDrant`	qdrant:6333/dashboard	http://host.docker.internal:6333/dashboard	http://localhost:6333/dashboard
`Subabase`	supabase-kong:8000	http://host.docker.internal:8000	http://localhost:8000
`Postgres`	postgres:5432	http://host.docker.internal:5432/	http://localhost:5432/
`MCP Gateway`	mcp-gateway:8060/	http://host.docker.internal:8090/	http://localhost:8060/
`Open webUI Filesystem`	open-webui-filesystem:8091/docs	http://host.docker.internal:8091/docs	http://localhost:8091/docs
`Redis`	redis:6379/	http://host.docker.internal:6379/	http://localhost:6379/
`MinIO`	minio:9001/	http://host.docker.internal:9001/	http://localhost:9001/
`Langfuse Web`	langfuse-web:3000/	http://host.docker.internal:3000/	http://localhost:3000/
`Langfuse Worker`	langfuse-worker:3030/	http://host.docker.internal:3030/	http://localhost:3030/
`Logflare`	supabase-analytics:4000/dashboard	http://host.docker.internal:4000/dashboard	http://localhost:4000/dashboard
`ClickHouse`	clickhouse:8123/	http://host.docker.internal:8123/	http://localhost:8123/
`SearXNG`	searxng:8081/	http://host.docker.internal:8081/	http://localhost:8081/
`Neo4j`	neo4j:7473/	http://host.docker.internal:7473/	http://localhost:7473/
`Caddy`	caddy:443/	http://host.docker.internal:443/	http://localhost:443/

[!IMPORTANT]
For Supabase, POSTGRES_HOST is 'db' since that is the name of the
service running Supabase.

[!NOTE]
If you are running LLAMA on your Host, for the credential Local Ollama
service, set the base URL to http://host.docker.internal:11434/ and set
Local QdrantApi database to http://host.docker.internal:6333/. For a
LLaMA.cpp service, you can create a Local LLaMA service node using the
connection credential http://host.docker.internal:8040/ or simply point
the Local Ollama service to this credential.

Don't use localhost for the redirect URI, instead, use another domain.
It will still work!
Alternatively, you can set up local file triggers.

Open http://localhost:5678/ in your browser to initialize and set up n8n.
You’ll only have to set your admin login credentials once. You are NOT creating
an account with n8n in the setup here, it is only a local account for your
instance!
- Go to http://localhost:5678/home/credentials to configure credentials.
- Click on Local QdrantApi database and set the base URL as specified above.
- Click on Local Ollama/LLaMA service and set the base URL as specified above.
- Click on Create credential, enter Postgres in the search field and
  follow the subsequent dialogs to setup the Postgres account as specified
  above.
Open the Demo workflow and
confirm the credentials for Local Ollama/LLaMA service is properly configured.
Select Test workflow to confirm the workflow is properly configured.
If this is the first time you’re running the workflow, you may need to wait
until Ollama finishes downloading the specified model. You can inspect the
docker console logs to check on the progress.
Toggle the Demo workflow as active and treat the RAG AI Agent workflows.
Configure additional n8n workflows as desired:
V1 Local RAG AI Agent

V2 Qdrant RAG AI Agent

V3 Local Agentic RAG AI Agent

V4 Local_Get_Postgres_Tables
Next, configure Open WebUI. Open http://localhost:8080/ in your browser
to initialize and set up Open WebUI. You’ll only have to set your admin login
credentials once. You are NOT creating an account with Open WebUI in the setup
here, it is only a local account for your instance!
Go to Workspace → Functions to setup the n8n Pipes (Pipeline) function.
This function will enable integration with n8n as an entry in your model dropdown
list.
- Click on New Function
- Enter n8n Pipeline at Function Name and Function ID will auto-populate
  with n8n_Pipeline
- Enter An optimized streaming-enabled pipeline for interacting with n8n workflows
  in Description
- Copy the n8n_Pipeline function code at n8n.py
  (or the downloaded instance at ./open-webui/functions/owndev/pipelines/n8n/n8n.py)
  and paste it into the edit dialog.
Copy the webhook URL from the n8n_Pipeline function set in step 6.
Click on the gear icon and set the n8n_url to the webhook URL you copied
in a previous step.
Toggle the function on and now it will be available in your model dropdown
in the top left.

To access n8n, visit http://localhost:5678/ from your browser.

To access OpenClaw, run suite_services.py --operation clawdock-status

To access Open WebUI, visit http://localhost:3000/ from your browser.

To access OpenCode, run ./opencode/run_opencode_docker.py from a new terminal.

Additional Configuration

[!NOTE]
AI-Suite is designed to help you get started with self-hosted AI
workflows and agents. While it is not fully optimized for production environments,
it combines robust components that work well together for personal porjects.
Of course, you can further customize it to meet your specific needs.

To keep everything local, use the Ollama/LLaMA.cpp node for your language
model and Qdrant as your vector store.

OpenClaw

The OpenClaw Docker setup uses three config files on the host. The container
never stores secrets — everything is bind-mounted from local files.

[!NOTE]
The OpenClaw <project> directory for AI-Suite is ./openclaw while the
User home directory ~/ may be /mnt/c/Users/<You>/ on Windows WSL.

OpenClaw Docker files

File	Purpose
`Dockerfile`	Builds the `openclaw:local` image (Node 22, pnpm, non-root `node` user)
`docker-compose.yml`	Defines `openclaw-gateway` and `openclaw-cli` services, bind-mounts, ports
`scripts/docker/setup.sh`	First-time setup — builds image, creates `.env` from `.env.example`
`.env.example`	Template for `<project>/.env` with all supported vars and docs
`docker-compose.extra.yml`	Optional overrides — auto-loaded by ClawDock helpers if present

OpenClaw Config files

File	Purpose	Examples
`<project>/.env`	Docker infra — image, ports, gateway token	`OPENCLAW_GATEWAY_TOKEN`, `OPENCLAW_IMAGE`, `OPENCLAW_GATEWAY_PORT`, `OPENCLAW_AUTH_PROFILE_SECRET_DIR`
`~/.openclaw/.env`	Secrets — API keys and bot tokens	`OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `TELEGRAM_BOT_TOKEN`
`~/.openclaw/openclaw.json`	Behavior config — models, channels, policies	Model selection, WhatsApp allowlists, agent settings

[!CAUTION]
Do NOT put API keys or bot tokens in openclaw.json. Use ~/.openclaw/.env
for all secrets.

Initial OpenClaw setup

./scripts/docker/setup.sh handles the first-time Docker configuration:

Builds the openclaw:local image from Dockerfile
Creates <project>/.env from .env.example with a generated gateway token
Creates the auth-profile secret key directory
Sets up ~/.openclaw directories if they don't exist

./scripts/docker/setup.sh

After installation setup, you may add your API keys:

vim ~/.openclaw/.env

See .env.example for all supported keys.

The Dockerfile supports two optional build args:

OPENCLAW_DOCKER_APT_PACKAGES — extra apt packages to install (e.g. ffmpeg)
OPENCLAW_INSTALL_BROWSER=1 — pre-install Chromium for browser automation (adds
~300MB, but skips the 60-90s Playwright install on each container start)

How OpenClaw works in docker

docker-compose.yml bind-mounts both config and workspace from the host:

volumes:
  - ${OPENCLAW_CONFIG_DIR}:/home/node/.openclaw
  - ${OPENCLAW_WORKSPACE_DIR}:/home/node/.openclaw/workspace
  - ${OPENCLAW_AUTH_PROFILE_SECRET_DIR}:/home/node/.config/openclaw

This means:

~/.openclaw/.env is available inside the container at /home/node/.openclaw/.env
— OpenClaw loads it automatically as the global env fallback
~/.openclaw/openclaw.json is available at /home/node/.openclaw/openclaw.json
— the gateway watches it and hot-reloads most changes
~/.openclaw-auth-profile-secrets is available at /home/node/.config/openclaw
— OpenClaw stores the auth-profile encryption key there
Downloadable plugin packages and install records live under the mounted OpenClaw
home
No need to add API keys to docker-compose.yml or configure anything inside the
container
Keys survive clawdock-update, clawdock-rebuild, and clawdock-clean because
they live on the host

The project .env feeds Docker Compose directly (gateway token, image name, ports).
The ~/.openclaw/.env feeds the OpenClaw process inside the container.

OpenClaw setup environment variables

Optional .env environment variables:

Variable	Purpose
`OPENCLAW_IMAGE`	Use a remote image instead of building locally
`OPENCLAW_IMAGE_APT_PACKAGES`	Install extra apt packages during build (space-separated)
`OPENCLAW_IMAGE_PIP_PACKAGES`	Install extra Python packages during build (space-separated)
`OPENCLAW_EXTENSIONS`	Pre-install plugin dependencies at build time (space-separated names)
`OPENCLAW_EXTRA_MOUNTS`	Extra host bind mounts (comma-separated `source:target[:opts]`)
`OPENCLAW_HOME_VOLUME`	Persist `/home/node` in a named Docker volume
`OPENCLAW_SANDBOX`	Opt in to sandbox bootstrap (`1`, `true`, `yes`, `on`)
`OPENCLAW_SKIP_ONBOARDING`	Skip the interactive onboarding step (`1`, `true`, `yes`, `on`)
`OPENCLAW_DOCKER_SOCKET`	Override Docker socket path
`OPENCLAW_DISABLE_BONJOUR`	Disable Bonjour/mDNS advertising (defaults to `1` for Docker)
`OPENCLAW_DISABLE_BUNDLED_SOURCE_OVERLAYS`	Disable bundled plugin source bind-mount overlays
`OTEL_EXPORTER_OTLP_ENDPOINT`	Shared OTLP/HTTP collector endpoint for OpenTelemetry export
`OTEL_EXPORTER_OTLP_*_ENDPOINT`	Signal-specific OTLP endpoints for traces, metrics, or logs
`OTEL_EXPORTER_OTLP_PROTOCOL`	OTLP protocol override. Only `http/protobuf` is supported today
`OTEL_SERVICE_NAME`	Service name used for OpenTelemetry resources
`OTEL_SEMCONV_STABILITY_OPT_IN`	Opt in to latest experimental GenAI semantic attributes
`OPENCLAW_OTEL_PRELOADED`	Skip starting a second OpenTelemetry SDK when one is preloaded

OpenClaw dependencies

For dependencies available from Debian packages, use
OPENCLAW_IMAGE_APT_PACKAGES during image build. The legacy
OPENCLAW_DOCKER_APT_PACKAGES name is still accepted.
For Python dependencies, use OPENCLAW_IMAGE_PIP_PACKAGES. This runs
python3 -m pip install --break-system-packages during the image build, so pin
package versions and use only package indexes you trust.

OpenClaw example `~/.openclaw/.env`

OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
TELEGRAM_BOT_TOKEN=123456:ABCDEF...

OpenClaw example `<project>/.env`

OPENCLAW_CONFIG_DIR=/Users/you/.openclaw
OPENCLAW_WORKSPACE_DIR=/Users/you/.openclaw/workspace
OPENCLAW_GATEWAY_PORT=18789
OPENCLAW_BRIDGE_PORT=18790
OPENCLAW_GATEWAY_BIND=lan
OPENCLAW_GATEWAY_TOKEN=<generated-by-docker-setup>
OPENCLAW_AUTH_PROFILE_SECRET_DIR=/Users/you/.openclaw-auth-profile-secrets
OPENCLAW_IMAGE=openclaw:local

OpenClaw env var precedence

OpenClaw loads env vars in this order (highest wins, never overrides existing):

Process environment — docker-compose.yml environment: block (gateway
token, session keys)
.env in CWD — project root .env (Docker infra vars)
~/.openclaw/.env — global secrets (API keys, bot tokens)
openclaw.json env block — inline vars, applied only if still missing
Shell env import — optional login-shell scrape (OPENCLAW_LOAD_SHELL_ENV=1)

Update OpenClaw

[!IMPORTANT]
The openclaw update command does not work inside Docker.
The container runs as a non-root user with a source-built image, so npm i -g
will fail with an EACCES output message.
Use suite_services.py --operation clawdock-update instead — it pulls, rebuilds,
and restarts from the host.

PROJECTS_PATH environment variable

You can use the PROJECTS_PATH environment variable to allow n8n,
OpenCode, and Open WebUI Filesystem access to your project files.
During the installation process, if the key is not already present (or has no
value) in your .env file, the key and value are written to the working
environment variables with the value set to ~/projects. You can override this
behaviour by manually setting your desired path for this key in the .env file.

PROJECTS_PATH forms a volume bind mount to container paths for the functional
modules described above:

Module	Container	Bind Mount
n8n	n8n	`/home/node/projects`
OpenCode	opencode	`/root/projects`
Open WebUI Tool Filesystem	open-webui-filesystem	`/nonexistent/tmp`

n8n

With n8n, you have access to over 400 integrations and a suite of basic
and advanced AI nodes such as:
AI Agent,
Text classifier,
and Information Extractor
nodes.

MCP Client
- Configure MCP Client credentials.
  - In Nodes panel, search for MCP.
  - Select MCP Client.
  - Set MCP Endpoint URL: http://host.docker.internal:8060.
MCP Client (node)
- Install community nodes - You may need to restart container.
  - Go to Settings → Community nodes
  - Use npm Package Name: n8n-nodes-mcp.
  - Install node.
- Configure MCP Client (node) credentials.
  - In Nodes panel, search for MCP.
  - Select MCP Client (node).
  - In the node settings, select Connection Type: HTTP Streamable.
  - Create new credentials of type MCP Client (HTTP Streamable) API.
  - Set HTTP Streamable URL: http://host.docker.internal:3001/stream.
  - Add any required headers for authentication.

Open WebUI

MCPO
- Your MCP tool is available at http://host.docker.internal:8090.
- Test it live at http://host.docker.internal:8090/docs.
- Using the config file ./open-webui/mcpo/config.json,
  set additional configuration settings as desired.
- As we are using the config file config.json, each tool will be accessible
  under its own unique route, e.g. http://host.docker.internal:8090/MCP_DOCKER.

Locally available functions

Pipes:

n8n

./open-webui/functions/owndev/pipelines/n8n/

Anthropic

./open-webui/functions/open-webui/functions/pipes/anthropic/

Open AI

./open-webui/functions/open-webui/functions/pipes/openai/

Filters:

Various filters
```
./open-webui/functions/owndev/filters/
```

Agent hotswap

./open-webui/functions/open-webui/functions/filters/agent_hotswap/

Context clip

./open-webui/functions/open-webui/functions/filters/context_clip/

Dynamic vision router

./open-webui/functions/open-webui/functions/filters/dynamic_vision_router/

Max turns

./open-webui/functions/open-webui/functions/filters/max_turns/

Moderation

./open-webui/functions/open-webui/functions/filters/moderation/

Summarizer

./open-webui/functions/open-webui/functions/filters/summarizer/

Manual Configuration.
- Navigate to the locally available functions folder containing your desired
  function .py file.
- Copy the complete code from the function file (e.g. main.py)
- Add as a new Function in OpenWebUI → Admin Panel → Functions
- Configure function-specific settings as needed - follow function README for
  details.
- Enable the Function (also be sure to enable to Agent Swapper Icon in chat)

Filesystem (Server Tool)
- Your Filesystem server is available at http://host.docker.internal:8091/docs.
Pipelines
- Connect to Open WebUI.
  - Navigate to the Settings → Connections → OpenAI API section in Open WebUI.
  - Set the API URL to http:\\host.docker.internal:9099 and the API key
    to 0p3n-w3bu!. Your pipelines should now be active.
- Manage Configurations.
  - In the admin panel, go to Admin Settings → Pipelines tab.
  - Select your desired pipeline and modify the valve values directly from WebUI.

Open Code

run_opencode_docker.py
- Copy ./opencode/run_opencode_docker.py to or run it from your current work
  project.
opencode.jsonc
- Using the config file at ./opencode/opencode.jsonc
- Set additional configuration settings as desired.
PROJECT_PATH environment variable
- Set the PROJECT_PATH env variable to your working project directory before
  running OpenCode if you wish to set the work path to your current project but
  you will NOT launch OpenCode from the root of your working project.
  If the PROJECT_PATH var is not defined, the currend working directory from
  which OpenCode was launched is assumed.
project_path argument
- You can also pass a project_path argument to ./opencode/run_opencode_docker.py
  with -p, --project_path so an example command would be:
```
python run_opencode_docker.py --project_path 'opencode'
```

[!NOTE]
It is recommended that your working project directory be within and relative to
the path set for PROJECTS_PATH in the AI-Suite .env file - see
PROJECTS_PATH environment variable section described above.

Important: The format of the PROJECT_PATH entry must be the portion of
your project path that is relative to the entry specified in PROJECTS_PATH.
For example, if the full path to your project is ~/projects/ai-suite/opencode
, your PROJECT_PATH entry must be ai-suite/opencode, if your PROJECTS_PATH
entry is ~/projects.

When set, PROJECT_PATH is appended to the OpenCode container bind mounted
path /root/projects and the resulting path is set as work_dir to form the
OpenCode Docker exec command's workdir=work_dir keyword argument.

Ollama or LLaMA.cpp - running on host

LLAMA_PATH environment variable
- If Ollama is installed in a custom location or you are using LLaMA.cpp,
  Add LLAMA_PATH with its absolute path (including the file name) to your
  .env file.
OLLAMA_SERVER_ARGS environment variable
- Add OLLAMA_SERVER_ARGS with additional Ollama server process start arguments
  to your .env file.
LLAMACPP_MODELS_DIR environment variable
- If you are using LLaMA.cpp with models that were not downloaded with
  that instance of LLaMA.cpp, add LLAMACPP_MODELS_DIR with said models path
  to your .env file.
LLAMACPP_SERVER_ARGS environment variable
- Add LLAMACPP_SERVER_ARGS with additional LLaMA.cpp server process start arguments
  to your .env file.

Upgrading

To update AI-Suite images to their latest versions (n8n, Open WebUI, etc.),
run the update operation command argument optionally preceded by the specified
profile arguments (functional modules).
Alternatively, you run the install_ operation argument to perform an update without
the confirmation prompt. Using install, AI-Suite will assume you are proceeding
as if performing a new installation - i.e. no previous installation exists.

suite_services.py [--profile arguments] --operation argument:

Argument	Operation
`update`	Update - for specified containers, stop, pull images, and restart
`install`	Install - proceed as if performing a new installation

[!CAUTION]
Installation updates can impact the AI-Suite integrity. Consider backing
up your volumes to enable rollback or restoration. Performing an install
will prune both named and anonymous volumes. Volumes are not disturbed when
performing an update.

AI-Suite data volume mounts:

Data Volume Mount	Mount Path	Container
`n8n_node_data`	`/home/node/.n8n`	n8n
`neo4j_data`	`/data`	neo4j
`neo4j_config_data`	`/config`	neo4j
`ollama_data`	`/root/.ollama`	ollama
`opencode_data`	`/root/.config/opencode`	opencode
`open_webui_data`	`/app/backend/data`	open-webui
`open_webui_pipelines_data`	`/app/pipelines`	open-webui-pipelines
`postgres_data`	`/var/lib/postgresql/data`	postgres
`qdrant_data`	`/qdrant/storage`	qdrqnt
`redis_valkey_data`	`/data`	redis
`langfuse_clickhouse_data`	`/var/lib/clickhouse`	clickhouse
`langfuse_minio_data`	`/data`	minio
`llamacpp_data`	`/root/.cache`	llamacpp
`caddy_data`	`/data`	caddy
`caddy_config_data`	`/config`	caddy
`db-config`	`/etc/postgresql-custom`	supabase-db

Argument	Operation
`backup-data`	Backup Data - backup volume mount data to backup file
`restore-data`	Restore Data - restore volume mount data from backup file

Example command for volume mount data backup:

python suite_services.py --operation backup-data

[!NOTE]
The suite_services.py update operation argument will stop, pull the
image and restart containers for the specified --profile arguments.
However, to update the entire suite, simply omit the profile arguments.

If no profile arguments are specified, container images for all functional
modules plus Docker LLAMA (Ollama/LLaMA.cpp) will be pulled but only
functional module containers (n8n, Open WebUI, OpenCode etc.) will be
started. Docker LLAMA containers will not be started unless they are
explicitly specified as a profile argument.

Example command for full update:

python suite_services.py --operation update

Example command for full (new) install with Docker Ollama running on CPU:

python suite_services.py --profile ai-all cpu --operation install

Manual steps to upgrade

Stop services for running containers

# Before starting the update, stop services for running containers
docker compose -p ai-suite -f docker-compose.yml --profile <arguments> down --volumes

Update images built locally (Supabase, Open WebUI Filesystem)

# First, pull the Supabase GitHub repository
cd ai-suite/supabase
git pull

Perform the Supabase Docker Compose build

# Next, perform the Supabase Docker Compose build
# Note: If in public environment, add '-f ../docker-compose.override.public.yml'
docker compose -p ai-suite -f docker/docker-compose.yml up -d --build --remove-orphans

Pull the Open WebUI Tools Fileserver repository

# Next, pull the Open WebUI Tools Fileserver
cd ../open-webui/tools
git pull

Perform the, Fileserver Docker Compose build

# Next, perform the, Fileserver Docker Compose build
# Note: If in public environment, add '-f ../../../../docker-compose.override.public.yml'
docker compose -p ai-suite -f servers/filesystem/compose.yaml up -d --build --remove-orphans

Return to AI-Suite root directory

# Return to AI-Suite root directory
cd ../../

Pull latest versions of container images

# Pull latest versions of container images for specified profile arguments
docker compose -p ai-suite -f docker-compose.yml --profile <arguments> pull

Start services again for specified profile arguments

# Start services again for specified profile arguments
# Note: If in public environment, replace 'docker-compose.override.private.yml' with 'docker-compose.override.public.yml'
docker compose -p ai-suite -f docker-compose.yml -f docker-compose.override.private.yml --profile <arguments> up -d --build --remove-orphans

Replace profile <arguments> with ai-all to update all container images or
with your desired functional modules, e.g. n8n, opencode etc, plus your CPU/GPU
argument [cpu | gpu-nvidia | gpu-amd] if you are running Ollama in Docker.
See the profile arguments table above for all arguments.

Accessing local files

Some AI-Suite functional modules require access to a project workspace, a
shared data folder and/or its configuration file located on the Docker host.
These resources are mounted from the host to the module container the using a
Docker Compose volume bind mount.

AI-Suite Docker Compose bind mounts

<container>:
   - <host path>:<container path>[:<read/write access>]

n8n creates a shared folder located at /data/shared - use this path in
nodes that interact with the host filesystem. Additional folders include the
n8n-files folder located at /home/node/.n8n-files, the projects folder
located at /home/node/projects and the data folder located at /data.
The host root path is ./n8n/data.

n8n:
   - ./n8n/local-files:/home/node/.n8n-files
   - ./n8n/data:/data
   - ${PROJECTS_PATH:-./n8n/local-files}:/home/node/projects

n8n-import:
   - ./n8n/data:/data

OpenClaw OpenClaw configuration files.

openclaw-gateway:
  - $HOME/.openclaw:/home/node/.openclaw
  - $HOME/.openclaw/workspace:/home/node/.openclaw/workspace
  - $HOME/.openclaw-auth-profile-secrets:/home/node/.config/openclaw
openclaw-cli:
  - $HOME/.openclaw:/home/node/.openclaw
  - $HOME/.openclaw/workspace:/home/node/.openclaw/workspace
  - $HOME/.openclaw-auth-profile-secrets:/home/node/.config/openclaw

Open WebUI MCPO OpenAPI configuration file.

open-webui-mcpo:
   - ./open-webui/mcpo/config.json:/app/config.json

Open WebUI Filesystem local project files access.

open-webui-filesystem:
   - ${PROJECTS_PATH:-../shared}:/nonexistent/tmp

Open WebUI Pipelines shared files access.

open-webui-pipelines:
   - ./open-webui/piplines:/root/.pipelines

OpenCode configuration file and local project files access.

opencode:
   - ./opencode/opencode.jsonc:/root/.config/opencode/opencode.jsonc
   - ${PROJECTS_PATH:-./opencode}:/root/projects

Flowise shared files access.

flowise:
   - ./flowise:/root/.flowise

SearXNG shared files access.

searxng:
   - ./searxng:/etc/searxng:rw

Caddy configuration file and addond folder access.

caddy:
   - ./caddy/Caddyfile:/etc/caddy/Caddyfile:ro
   - ./caddy/addons:/etc/caddy/addons:ro

n8n Nodes that interact with the local filesystem

Repository Structure

Relevant AI-Suite repository folder and document resources

.
├── LICENSE                                  # AI-Suite license
├── README.md                                # AI-Suite readme
├── access/
│   ├── authelia/
│   │   ├── db/
│   │   │   └── schema-authelia.sh           # Authelia schema template
│   │   └── configuration.yml                # Authelia configuration
│   ├── caddy/                               # Caddy proxy server
│   │   ├── addons/
│   │   │   └── cors.conf                    # Caddy cors configuration
│   │   └── Caddyfile                        # Caddy configuration
│   ├── nginx/                               # Nginx configuration
│   │   └── addons/
│   │       ├── authelia-authrequest.conf    # Nginx Authelia auth request configuration
│   │       ├── authelia-location.conf       # Nginx Authelia location configuration
│   │       ├── common_proxy_headers.conf    # Nginx proxy headers configuration
│   │       ├── cors.conf                    # Nginx cors configuration
│   │       └── proxy.conf                   # Nginx proxy configuration
│   └── auto_config.sh                       # Proxy setup and access management script
├── assets/                                  # README.md gif image
├── flowise/                                 # Ready-to-import Flowise workflows
│   └── uploads/
├── langfuse/                                # Langfuse resources
│   └── clickhouse/
│       └── logs/                            # Clickhouse logs
├── llama.cpp/                               # LLM inference engine (if running LLaMA.cpp from host)
├── n8n/                                     # N8N resources
│   ├── data/
│   │   ├── credentials/                     # Ready-to-import n8n credentials
│   │   └── workflows/                       # Ready-to-import n8n workflows
│   └── local-files/
├── neo4j/                                   # Neo4j resources
│   ├── logs/
│   └── plugins/
├── open-webui/                              # Open WebUI resources
│   ├── functions/
│   │   ├── open-webui/
│   │   │   └── functions/                   # Open WebUI functions
│   │   └── owndev/
│   │       ├── docs/
│   │       ├── filters                      # Open WebUI filters
│   │       └── pipelines/                   # Open WebUI pipes
│   ├── mcpo/                                # Open WebUI MCPO
│   │   └── config.json                      # Open WebUI MCPO configuration
│   ├── piplines/                            # Open WebUI pipelines
│   └── tools/                               # Open WebUI tools
│       └── servers/                         # Open WebUI tool servers
│           └── filesystem/                  # Open WebUI filesystem tool
├── openclaw/                                # OpenClaw resources
│   ├── scripts/
│   │   ├── clawdock/
│   │   │   └── clawdock-helpers.sh          # OpenClaw clawdock commands setup script
│   │   └── docker/
│   │       ├── sandbox/
│   │       │   ├── Dockerfile               # OpenClaw sandbox Docker configuration
│   │       │   ├── Dockerfile.browser       # OpenClaw sandbox Docker browser configuration
│   │       │   └── Dockerfile.common        # OpenClaw sandbox Docker common configuration
│   │       └── setup.sh                     # OpenClaw Docker setup and install script
│   ├── .env.example
│   ├── docker-compose.sandbox.yml           # OpenClaw Docker Compose sandbox configuration
│   ├── docker-compose.yml                   # OpenClaw Docker Compose configuration
│   └── Dockerfile                           # OpenClaw Docker configuration
├── opencode/
│   ├── opencode.jsonc                       # OpenCode configuration
│   └── run_opencode_docker.py               # OpenCode launch script
├── searxng/
│   └── settings-base.yml                    # SearXNG configuration
├── state/                                   # AI-Suite state
├── supabase/                                # Supabase resources
│   └── docker/
│       ├── volumes/
│       │   └── db/
│       │       └── schema-authelia.sh       # Supabase Authelia schema install
│       ├── env.example                      # Supabase template for environment variables
│       └── docker-compose.yml               # Supabase Docker Compose configuration
├── .env.example                             # AI-Suite environment variables template
├── .openclaw.example.json                   # AI-Suite OpenClaw configuration template
├── docker-compose.override.private.yml      # AI-Suite Docker Compose local configuration
├── docker-compose.override.public.yml       # AI-Suite Docker Compose production configuration
├── docker-compose.yml                       # AI-Suite Docker Compose configuration
└── suite_services.py                        # AI-Suite Installation and service management script

Troubleshooting

Here are solutions to common issues you might encounter:

Supabase Issues

Supabase Pooler Restarting: If the supabase-pooler container keeps
restarting itself, follow the instructions in this GitHub issue.
Supabase Analytics Startup Failure: If the supabase-analytics container
fails to start after changing your Postgres password, delete the folder supabase/docker/volumes/db/data.
If using Docker Desktop: Go into the Docker settings and make sure
"Expose daemon on tcp://localhost:2375 without TLS" is turned on
Supabase Service Unavailable - Make sure you don't have an "@" character
in your Postgres password! If the connection to the kong container is working
(the container logs say it is receiving requests from n8n) but n8n says it
cannot connect, this is generally the problem from what the community has
shared. Other characters might not be allowed too, the @ symbol is just the
one I know for sure!
SearXNG Restarting: If the SearXNG container keeps restarting, run the
command "chmod 755 searxng" within the ai-suite folder so SearXNG has the
permissions it needs to create the uwsgi.ini file.
Files not Found in Supabase Folder - If you get any errors around files
missing in the supabase/ folder like .env, docker/docker-compose.yml, etc. This
most likely means you had a "bad" pull of the Supabase GitHub repository when
you ran the suite_services.py script. Delete the supabase/ folder within the
Local AI Package folder entirely and try again.

GPU Support Issues

Windows GPU Support: If you're having trouble running Ollama with GPU
support on Windows with Docker Desktop:
1. Open Docker Desktop settings
2. Ensure 'Enable WSL2 backend' is enabled
3. See the Docker GPU documentation
  for more details
Linux GPU Support: If you're having trouble running Ollama with GPU
support on Linux, follow the Ollama Docker instructions.

🛍️ More AI templates

For more AI workflow ideas, visit the official n8n AI template
gallery. From each workflow,
select the Use workflow button to automatically import the workflow into
your local n8n instance.

👓 Recommended reading

Useful content for deeper understanding.

AI concepts

OpenClaw Highlights

Local-first Gateway — single control plane
for sessions, channels, tools, and events.
Multi-channel inbox — WhatsApp, Telegram,
Slack, Discord, Google Chat, Signal, iMessage, IRC, Microsoft Teams, Matrix, Feishu,
LINE, Mattermost, Nextcloud Talk, Nostr, Synology Chat, Tlon, Twitch, Zalo, Zalo
Personal, WeChat, QQ, WebChat, macOS, iOS/Android.
Multi-agent routing — route
inbound channels/accounts/peers to isolated agents (workspaces + per-agent sessions).
Voice Wake +
Talk Mode — wake words on macOS/iOS and
continuous voice on Android (ElevenLabs + system TTS fallback).
Live Canvas — agent-driven
visual workspace with A2UI.
First-class tools — browser, canvas, nodes,
cron, sessions, and Discord/Slack actions.
Companion apps — macOS menu bar
app + iOS/Android nodes.
Onboarding +
skills — onboarding-driven setup with
bundled/managed/workspace skills.

📜 License

This project (portions of which were adapted from content produced by the n8n
team, then Cole Medin, links at the top of the README) is licensed under the
Apache License 2.0 - see the LICENSE file for details.

AI-Suite

What’s included

Prerequisites

Installation

Step 1: Clone the repository and set environment variables

Step 2: Run suite_services.py

The profile command arguments

For Docker LLAMA with Nvidia GPU users

For Docker LLAMA with AMD GPU users

For LLAMA on Mac running Apple Silicon users

For LLAMA running on the Host users

For everyone else (...using CPU)

The operation command argument

The environment command argument

The log command argument

Auto-configuration, HTTPS Reverse Proxy and Access Management

Deploying to the Cloud

Prerequisite

Extra steps

⚡️ Quick start and usage

Additional Configuration

OpenClaw

OpenClaw Docker files

OpenClaw Config files

Initial OpenClaw setup

How OpenClaw works in docker

OpenClaw setup environment variables

OpenClaw dependencies

OpenClaw example ~/.openclaw/.env

OpenClaw example <project>/.env

OpenClaw env var precedence

Update OpenClaw

PROJECTS_PATH environment variable

n8n

Open WebUI

Open Code

Ollama or LLaMA.cpp - running on host

Upgrading

Manual steps to upgrade

Accessing local files

n8n Nodes that interact with the local filesystem

Repository Structure

Troubleshooting

Supabase Issues

GPU Support Issues

🛍️ More AI templates

👓 Recommended reading

AI concepts

OpenClaw Highlights

📜 License

Yorumlar (0)

OpenClaw example `~/.openclaw/.env`

OpenClaw example `<project>/.env`