AIStack

Unified AI Infrastructure Stack — One-command Docker Compose for LLM routing, vector storage, private search, and browser automation.

---

Overview

AIStack is a production-ready self-hosted AI infrastructure stack that brings together the best open-source tools for running AI agents locally. Deploy with a single command — no GPU required.

What's Included

Service	Purpose	Port	Status
Qdrant	Vector database for semantic search & embeddings	:6333	✅
SearXNG	Private metasearch engine (100+ providers, no tracking)	:8080	✅
Camofox	Stealth browser automation for web agents	:9377	✅
Honcho	Agent memory management (PostgreSQL + Redis)	:8081	✅
Obsidian	Browser-based note vault for agents & humans	:8083	✅
RAG Dashboard	Document Q&A with retrieval-augmented generation	:8123	✅

LLM layer: Ollama is not shipped in the compose file. Point the RAG Dashboard at any Ollama endpoint via OLLAMA_URL (defaults to the host's :11434).

Quick Start

git clone https://github.com/OneByJorah/AIStack.git
cd AIStack
cp .env.example .env
docker compose up -d

Access Services

Service	URL	Credentials
Qdrant Dashboard	http://localhost:6333/dashboard	—
SearXNG	http://localhost:8080	—
Camofox API	http://localhost:9377	CAMOFOX_API_KEY (if set)
Honcho API	http://localhost:8081	HONCHO_TOKEN (if set)
Obsidian	http://localhost:8083	—
RAG Dashboard	http://localhost:8123	admin / admin (change via RAG_ADMIN_PASSWORD)

Features

• One-Command Deploy — docker compose up -d and you're ready
• LLM Flexibility — Point services at any Ollama/OpenAI-compatible endpoint via env vars
• Vector Storage — Qdrant for embeddings and semantic search (ANN indexing)
• Private Web Search — SearXNG aggregates 100+ search providers without tracking
• Browser Automation — Camofox stealth browser for undetectable web interaction
• Agent Memory — Honcho provides persistent context for AI agents
• Notes Vault — Obsidian accessible from the browser at :8083
• CPU-Only — Runs entirely without GPU, works on any hardware
• Production Ready — Health checks, monitoring, and structured logging

Architecture

┌──────────────────────────────────────────────────────────┐
│                       AIStack                            │
│                                                          │
│  ┌───────────┐  ┌───────────┐  ┌───────────┐            │
│  │  SearXNG  │  │  Qdrant   │  │  Camofox  │            │
│  │  :8080    │  │  :6333    │  │  :9377    │            │
│  └─────┬─────┘  └─────┬─────┘  └─────┬─────┘            │
│        │               │               │                  │
│  ┌─────▼─────┐  ┌─────▼─────┐  ┌─────▼─────┐            │
│  │  Honcho   │  │ Obsidian  │  │    RAG    │            │
│  │  :8081    │  │  :8083    │  │ Dashboard │            │
│  └───────────┘  └───────────┘  └───────────┘            │
│                                                          │
│  Optional: Caddy gateway (caddy/Caddyfile, not wired by  │
│  default) · external LLM endpoint via OLLAMA_URL         │
└──────────────────────────────────────────────────────────┘

Tech Stack

Category	Technology
Orchestration	Docker Compose
LLM	Ollama (llama.cpp)
Vector DB	Qdrant
Search	SearXNG
Browser	Camofox (Playwright-based)
Memory	Honcho + PostgreSQL (pgvector) + Redis
Dashboard	FastAPI + ChromaDB + Sentence Transformers

Configuration

See .env.example for all configuration options.

Variable	Default	Description
SERVER_IP	—	Mesh-VPN IP of the host (used by healthcheck)
OBSIDIAN_VAULT_PATH	./obsidian-vault	Host path for the Obsidian vault
HONCHO_TOKEN	—	Honcho API auth token
HONCHO_DB_PASSWORD	placeholder — set it!	PostgreSQL password for Honcho
SEARXNG_SECRET_KEY	dev fallback — set it!	SearXNG instance secret
CAMOFOX_API_KEY / CAMOFOX_ADMIN_KEY	—	Camofox auth keys
OLLAMA_URL	http://host.docker.internal:11434	LLM endpoint for RAG Dashboard
RAG_MODEL	rag-assistant	Ollama model for RAG queries
RAG_ADMIN_PASSWORD	admin	RAG Dashboard admin password

Screenshots

Click to expand

Stack Dashboard

Project Structure

AIStack/
├── docker-compose.yml     # Main orchestration
├── .env.example           # Environment template
├── searxng/
│   └── settings.yml       # SearXNG configuration
├── rag-dashboard/         # RAG web application
│   ├── app.py             # FastAPI backend
│   └── Dockerfile
├── browser-search/        # Camofox/CloakBrowser skill + scripts
├── caddy/Caddyfile        # Optional reverse proxy (not wired by default)
├── litellm/config.yaml    # Optional LLM routing config (not wired by default)
├── scripts/               # Utility scripts
│   ├── bootstrap.sh       # Initial setup
│   ├── healthcheck.sh     # Service health checks
│   ├── init-honcho.sh     # Honcho initialization
│   └── init-obsidian.sh   # Vault initialization
├── docs/                  # Documentation
└── README.md

Use Cases

• AI Agent Backend — Run autonomous agents with local LLMs, search, and memory
• Private RAG — Document Q&A without sending data to third parties
• Development — Test AI features locally before cloud deployment
• Homelab — Self-hosted AI infrastructure for your home server
• Education — Learn about AI agent architectures hands-on

Contributing

Contributions are welcome! See CONTRIBUTING.md and CODE_OF_CONDUCT.md.

Security

Report vulnerabilities to info@jorahone.com — see SECURITY.md.

Known exposure notes (by design for mesh-VPN/LAN use — do not expose to the public internet):

• All published ports bind 0.0.0.0; Qdrant (:6333), Obsidian (:8083) and SearXNG (:8080) have no built-in auth.
• Honcho ships with AUTH_USE_AUTH: "false"; set a HONCHO_TOKEN and enable auth before any shared-network deployment.
• The RAG Dashboard defaults to admin / admin; set RAG_ADMIN_PASSWORD on first run.
• Prefer binding ports to 127.0.0.1: or fronting them with an authenticated proxy in production.

License

MIT © Jhonattan L. Jimenez (OneByJorah)

---

OneByJorah · jorahone.com · Hermes Setup · Changelog