CBass - Self-Hosted AI Stack

CBass is a self-hosted AI Docker Compose orchestration platform that combines AI and database services into a cohesive local infrastructure stack with automatic HTTPS via Caddy.

Fork of coleam00/local-ai-packaged, which itself builds on n8n-io/self-hosted-ai-starter-kit

Live Deployment: cbass.space

Features

n8n - Visual workflow automation with 400+ integrations
Open WebUI - ChatGPT-style interface for local LLMs
Supabase - PostgreSQL database with pgvector for embeddings
Ollama - Local LLM inference (Llama, Qwen, etc.)
Flowise - No-code AI agent builder
Qdrant - High-performance vector database
Neo4j - Knowledge graph database
Langfuse - LLM observability and tracing
SearXNG - Privacy-respecting meta search engine
Kali Linux - Browser-based security desktop (KasmWeb)
Caddy - Automatic HTTPS with Let's Encrypt

Architecture

graph TB
    subgraph Internet
        User[User Browser]
    end

    subgraph Reverse Proxy
        Caddy[Caddy :80/:443]
    end

    subgraph Frontend Layer
        Dashboard[Dashboard :3000]
        OpenWebUI[Open WebUI :8080]
        Flowise[Flowise :3001]
    end

    subgraph Workflow & AI Layer
        N8N[n8n :5678]
        Ollama[Ollama :11434]
        Langfuse[Langfuse :3000]
    end

    subgraph Data Layer
        Supabase[Supabase Kong :8000]
        Qdrant[Qdrant :6333]
        Neo4j[Neo4j :7474]
    end

    subgraph Utilities
        SearXNG[SearXNG :8080]
        Kali[Kali :6901]
        Updater[Updater :9000]
    end

    User --> Caddy
    Caddy --> Dashboard
    Caddy --> OpenWebUI
    Caddy --> N8N
    Caddy --> Flowise
    Caddy --> Supabase
    Caddy --> Langfuse
    Caddy --> Neo4j
    Caddy --> SearXNG
    Caddy --> Kali

    OpenWebUI --> N8N
    N8N --> Ollama
    N8N --> Supabase
    N8N --> Qdrant
    N8N --> Neo4j
    Flowise --> Ollama
    Flowise --> Supabase

Request Flow

The primary user flow routes chat messages from Open WebUI through n8n for AI agent processing:

sequenceDiagram
    participant User
    participant Caddy
    participant OpenWebUI
    participant n8n_pipe
    participant n8n
    participant Ollama
    participant Supabase
    participant Qdrant

    User->>Caddy: HTTPS Request
    Caddy->>OpenWebUI: Proxy to :8080
    OpenWebUI->>n8n_pipe: Chat Message
    n8n_pipe->>n8n: POST /webhook
    n8n->>Supabase: Query Context
    n8n->>Qdrant: Vector Search
    n8n->>Ollama: LLM Inference
    Ollama-->>n8n: Response
    n8n-->>n8n_pipe: JSON Response
    n8n_pipe-->>OpenWebUI: Display
    OpenWebUI-->>User: Chat Response

Service Inventory

Core Services

Service	Container	Internal Port	Purpose
Dashboard	dashboard	3000	Next.js AI Command Center
n8n	n8n	5678	Workflow automation & AI agents
Open WebUI	open-webui	8080	Chat interface for LLMs
Flowise	flowise	3001	Visual no-code AI builder
Ollama	ollama	11434	Local LLM inference

Data Layer

Service	Container	Internal Port	Purpose
Supabase Kong	kong	8000	API Gateway
Supabase DB	db	5432	PostgreSQL + pgvector
Qdrant	qdrant	6333, 6334	Vector database (HTTP/gRPC)
Neo4j	neo4j	7474, 7687	Graph database (Browser/Bolt)

Observability

Service	Container	Internal Port	Purpose
Langfuse Web	langfuse-web	3000	LLM tracing UI
Langfuse Worker	langfuse-worker	3030	Background processing
ClickHouse	clickhouse	8123, 9000	Analytics database
MinIO	minio	9000, 9001	S3-compatible storage

Utilities

Service	Container	Internal Port	Purpose
SearXNG	searxng	8080	Meta search engine
Redis	redis	6379	Cache (Valkey)
Kali	kali	6901	Browser-based Kali desktop
Updater	updater	9000	Webhook container updates
Caddy	caddy	80, 443	Reverse proxy with auto-TLS

Supabase Components

Service	Container	Internal Port	Purpose
Kong	kong	8000, 8443	API Gateway
GoTrue	auth	9999	Authentication
PostgREST	rest	3000	REST API
Realtime	realtime	4000	WebSocket subscriptions
Storage	storage	5000	File storage
Studio	studio	3000	Dashboard UI
Supavisor	pooler	6543	Connection pooler

Port Reference

Port	Service	Protocol	Notes
80	Caddy	HTTP	Redirects to 443
443	Caddy	HTTPS	Main entry point
3000	Dashboard / Langfuse / Studio	HTTP	Multiple services share port internally
3001	Flowise	HTTP	Visual AI builder
5432	PostgreSQL	TCP	Database
5678	n8n	HTTP	Workflow engine
6333	Qdrant	HTTP	Vector DB REST API
6334	Qdrant	gRPC	Vector DB gRPC
6379	Redis/Valkey	TCP	Cache
6543	Supavisor	TCP	Connection pooler
6901	Kali	HTTPS	VNC/noVNC (KasmWeb)
7474	Neo4j	HTTP	Browser UI
7687	Neo4j	Bolt	Graph queries
8000	Kong	HTTP	Supabase API Gateway
8080	Open WebUI / SearXNG	HTTP	Multiple services share port internally
8123	ClickHouse	HTTP	Analytics queries
9000	MinIO / Updater / ClickHouse	HTTP	Multiple services share port internally
9001	MinIO Console	HTTP	S3 management UI
11434	Ollama	HTTP	LLM API

Subdomain Routing (Production)

Subdomain	Target Service	Internal Port
`cbass.space`	dashboard	3000
`www.cbass.space`	(redirect to cbass.space)	-
`n8n.cbass.space`	n8n	5678
`openwebui.cbass.space`	open-webui	8080
`flowise.cbass.space`	flowise	3001
`supabase.cbass.space`	kong	8000
`langfuse.cbass.space`	langfuse-web	3000
`neo4j.cbass.space`	neo4j	7474
`searxng.cbass.space`	searxng	8080
`kali.cbass.space`	kali	6901

VPS vs Local Repository

Component	VPS (cbass.space)	Local Repository	Notes
`dashboard/`	Running	Source code	Next.js app built in container
`scripts/`	Running	Source code	Webhook automation
Kali service	Running	In docker-compose.yml	KasmWeb desktop
Updater service	Running	In docker-compose.yml	Container updates
`.env`	Production secrets	Use `env.example`	Never committed
`supabase/`	Cloned at runtime	Auto-cloned on first run	Sparse checkout
`node_modules/`	Built in container	gitignored
`searxng/`	Generated config	Auto-generated	Created on first run

Quick Start

Prerequisites

Python 3.x
Docker Desktop or Docker + Docker Compose
Git

Installation

# Clone the repository
git clone https://github.com/mdc159/cbass.git
cd cbass

# Create environment file
cp env.example .env

# Edit .env with your secrets (see Environment Variables section)

# Start with NVIDIA GPU
python start_services.py --profile gpu-nvidia --environment private --open-dashboard

# Or start with CPU only
python start_services.py --profile cpu --environment private --open-dashboard

Access Services

After startup, access services at:

Service	URL
n8n	http://localhost:5678
Open WebUI	http://localhost:8080
Flowise	http://localhost:3001
Supabase Studio	http://localhost:8000
Neo4j Browser	http://localhost:7474
Langfuse	http://localhost:3000

Documentation

Start with docs/README.md for the full documentation index, including deployment, services, architecture, and operations guides.

Claude Code users can use /onboard or /cbass-context to load project context quickly (see docs/README.md and CLAUDE.md).

Environment Variables

Required Secrets

# n8n - generate with: openssl rand -hex 32
N8N_ENCRYPTION_KEY=
N8N_USER_MANAGEMENT_JWT_SECRET=

# Supabase - see: https://supabase.com/docs/guides/self-hosting/docker
POSTGRES_PASSWORD=          # No '@' character!
JWT_SECRET=
PG_META_CRYPTO_KEY=        # Reuse JWT_SECRET value if desired
ANON_KEY=
SERVICE_ROLE_KEY=
DASHBOARD_USERNAME=
DASHBOARD_PASSWORD=
POOLER_TENANT_ID=
NEXT_PUBLIC_SUPABASE_URL=  # e.g. http://localhost:8000
NEXT_PUBLIC_SUPABASE_ANON_KEY=  # same value as ANON_KEY

# Neo4j
NEO4J_AUTH=neo4j/your_password

# Langfuse
CLICKHOUSE_PASSWORD=
MINIO_ROOT_PASSWORD=
LANGFUSE_SALT=
NEXTAUTH_SECRET=
ENCRYPTION_KEY=             # generate with: openssl rand -hex 32

Production Hostnames

# Uncomment and set for production deployment
N8N_HOSTNAME=n8n.yourdomain.com
WEBUI_HOSTNAME=openwebui.yourdomain.com
FLOWISE_HOSTNAME=flowise.yourdomain.com
SUPABASE_HOSTNAME=supabase.yourdomain.com
LANGFUSE_HOSTNAME=langfuse.yourdomain.com
NEO4J_HOSTNAME=neo4j.yourdomain.com
SEARXNG_HOSTNAME=searxng.yourdomain.com
KALI_HOSTNAME=kali.yourdomain.com
LETSENCRYPT_EMAIL=your@email.com

Optional Configuration

# Flowise authentication
FLOWISE_USERNAME=
FLOWISE_PASSWORD=

# SearXNG workers
SEARXNG_UWSGI_WORKERS=4
SEARXNG_UWSGI_THREADS=4

# Google OAuth for Supabase
ENABLE_GOOGLE_SIGNUP=true
GOOGLE_CLIENT_ID=
GOOGLE_CLIENT_SECRET=

Dashboard Supabase Target

The dashboard auth/API target is controlled by NEXT_PUBLIC_SUPABASE_URL.

Local development: NEXT_PUBLIC_SUPABASE_URL=http://localhost:8000
Remote deployment: NEXT_PUBLIC_SUPABASE_URL=https://supabase.yourdomain.com

If this points to a remote URL, http://localhost:3002 will authenticate against that remote Supabase, not your local container.

Directory Structure

CBass/
├── start_services.py           # Main entry point
├── docker-compose.yml          # Service definitions
├── docker-compose.override.private.yml  # Dev mode (ports exposed)
├── docker-compose.override.public.yml   # Prod mode (only 80/443)
├── Caddyfile                   # Reverse proxy routing
├── n8n_pipe.py                 # Open WebUI → n8n bridge
├── env.example                 # Environment template
│
├── dashboard/                  # Next.js AI Command Center
│   ├── app/                   # App router pages
│   ├── components/            # UI components (shadcn/ui)
│   ├── public/                # Static assets
│   └── Dockerfile             # Container build
│
├── scripts/                    # Webhook automation
│   ├── hooks.json             # Webhook configuration
│   └── update-container.sh    # Container update script
│
├── n8n/backup/                 # Pre-built RAG workflows
│   ├── workflows/
│   └── credentials/
│
├── flowise/                    # Flowise tools & chatflows
├── neo4j/                      # Neo4j data/logs/plugins
├── searxng/                    # SearXNG config (generated)
├── supabase/                   # Auto-cloned on first run
└── shared/                     # Shared files accessible from n8n

Deployment Profiles

GPU Profiles

Profile	Command	Use Case
`gpu-nvidia`	`--profile gpu-nvidia`	NVIDIA GPU with CUDA
`gpu-amd`	`--profile gpu-amd`	AMD GPU with ROCm (Linux)
`cpu`	`--profile cpu`	CPU-only inference
`none`	`--profile none`	External LLM API only

Environment Modes

Mode	Command	Behavior
`private`	`--environment private`	All ports exposed locally (default)
`public`	`--environment public`	Only ports 80/443 via Caddy

Startup Convenience Flags

Flag	Default	Behavior
`--open-dashboard`	off	Opens dashboard in your default browser after startup
`--dashboard-url <url>`	`http://localhost:3002`	URL used when `--open-dashboard` is enabled

Examples

# Local development with NVIDIA GPU
python start_services.py --profile gpu-nvidia --environment private --open-dashboard

# Production deployment
python start_services.py --profile gpu-nvidia --environment public

# CPU-only development
python start_services.py --profile cpu

# Use external API (no local Ollama)
python start_services.py --profile none

# Open a custom dashboard URL after startup
python start_services.py --profile none --environment private --open-dashboard --dashboard-url http://localhost:3002

Common Operations

# Check service status
docker compose -p localai ps

# View logs for a service
docker compose -p localai logs -f n8n
docker compose -p localai logs -f open-webui

# Restart a specific service
docker compose -p localai restart n8n

# Stop all services
docker compose -p localai down

# Update all containers
docker compose -p localai -f docker-compose.yml --profile <your-profile> pull
python start_services.py --profile <your-profile>

Pre-built Workflows

n8n RAG Agents

Located in n8n/backup/workflows/:

V1_Local_RAG_AI_Agent.json - Basic RAG with local LLM
V2_Local_Supabase_RAG_AI_Agent.json - RAG using Supabase vectors
V3_Local_Agentic_RAG_AI_Agent.json - Advanced agentic RAG

Flowise Tools

Located in flowise/:

Google Docs creation
PostgreSQL queries
Slack integration
Web search tools

Open WebUI + n8n Integration

The n8n_pipe.py script bridges Open WebUI to n8n webhooks:

Start services and create an n8n account at http://localhost:5678
Import a workflow from n8n/backup/workflows/
Activate the workflow and copy the Production webhook URL
In Open WebUI, go to Workspace → Functions → Add Function
Paste the code from n8n_pipe.py
Click the gear icon and set n8n_url to your webhook URL
Enable the function - it appears in the model dropdown

Also available at: openwebui.com/f/coleam/n8n_pipe

Cloud Deployment

Prerequisites

Linux server (Ubuntu recommended)
Docker and Docker Compose installed
Domain with DNS access

Steps

Open firewall ports:

ufw enable
ufw allow 80 && ufw allow 443
ufw reload

Clone and configure:

git clone https://github.com/mdc159/cbass.git
cd cbass
cp env.example .env
# Edit .env with production hostnames and secrets

Set up DNS A records pointing your subdomains to your server IP

Start in public mode:

python3 start_services.py --profile gpu-nvidia --environment public

Troubleshooting

Supabase Issues

Pooler Restarting: See GitHub issue #30210
Analytics Startup Failure: After changing Postgres password, delete supabase/docker/volumes/db/data
Service Unavailable: Ensure no @ character in POSTGRES_PASSWORD - this breaks URI parsing
Files Not Found: Delete the supabase/ folder and re-run start_services.py to re-clone

SearXNG Issues

Container Restarting: Set permissions: chmod 755 searxng

GPU Issues

Windows GPU: Enable WSL 2 backend in Docker Desktop. See Docker GPU docs
Linux GPU: Follow Ollama Docker instructions

Docker Desktop

Enable "Expose daemon on tcp://localhost:2375 without TLS" in settings

Credential Configuration

When setting up n8n credentials:

Service	Host	Port	Notes
Ollama	`ollama`	11434	`http://ollama:11434`
PostgreSQL	`db`	5432	Use container name, not localhost
Qdrant	`qdrant`	6333	`http://qdrant:6333`
Neo4j	`neo4j`	7687	Bolt protocol

Mac Users Running Ollama Locally

If running Ollama outside Docker on Mac:

Set OLLAMA_HOST=host.docker.internal:11434 in n8n service environment
Update n8n Ollama credential to use http://host.docker.internal:11434/

Resources

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Attribution

Original: n8n-io/self-hosted-ai-starter-kit
Enhanced by: coleam00/local-ai-packaged
This fork: mdc159/cbass

Name		Name	Last commit message	Last commit date
Latest commit History 52 Commits
.claude		.claude
.github		.github
dashboard		dashboard
docs		docs
flowise		flowise
issues		issues
mcp/flowise-enhanced		mcp/flowise-enhanced
n8n-tool-workflows		n8n-tool-workflows
references		references
scripts		scripts
searxng		searxng
vendor		vendor
.env.ip-only		.env.ip-only
.gitattributes		.gitattributes
.gitignore		.gitignore
.gitmodules		.gitmodules
.mcp.json.example		.mcp.json.example
CHANGELOG.md		CHANGELOG.md
CLAUDE.md		CLAUDE.md
Caddyfile		Caddyfile
LICENSE		LICENSE
README.md		README.md
docker-compose.override.private.supabase.yml		docker-compose.override.private.supabase.yml
docker-compose.override.private.yml		docker-compose.override.private.yml
docker-compose.override.public.supabase.yml		docker-compose.override.public.supabase.yml
docker-compose.override.public.yml		docker-compose.override.public.yml
docker-compose.yml		docker-compose.yml
env.example		env.example
n8n_pipe.py		n8n_pipe.py
start_services.py		start_services.py
wrap_flowise.ps1		wrap_flowise.ps1

Folders and files

Latest commit

History

Repository files navigation

CBass - Self-Hosted AI Stack

Features

Architecture

Request Flow

Service Inventory

Core Services

Data Layer

Observability

Utilities

Supabase Components

Port Reference

Subdomain Routing (Production)

VPS vs Local Repository

Quick Start

Prerequisites

Installation

Access Services

Documentation

Environment Variables

Required Secrets

Production Hostnames

Optional Configuration

Dashboard Supabase Target

Directory Structure

Deployment Profiles

GPU Profiles

Environment Modes

Startup Convenience Flags

Examples

Common Operations

Pre-built Workflows

n8n RAG Agents

Flowise Tools

Open WebUI + n8n Integration

Cloud Deployment

Prerequisites

Steps

Troubleshooting

Supabase Issues

SearXNG Issues

GPU Issues

Docker Desktop

Credential Configuration

Mac Users Running Ollama Locally

Resources

License

Attribution

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages