StackDeploy Dashboard

Self-hosted all-in-one API platform — deploy SearXNG, Qdrant, Honcho, Ollama, Camofox, and Obsidian behind a single gateway with auto-discoverable APIs, Tailscale mesh, and optional Cloudflare Tunnel for public HTTPS. Choose between local LLM inference (Ollama) or cloud API (OpenRouter) for Honcho's AI features.

Quick Start

Zero-config auto-deploy (recommended)

git clone https://github.com/OneByJorah/StackDeploy-Dashboard.git
cd StackDeploy-Dashboard
sudo ./bootstrap.sh --auto

This single command generates secure passwords, deploys the full stack (gateway, SearXNG, Qdrant, Honcho, Obsidian, Camofox, Ollama), pulls a 1B LLM model (llama3.2:1b), and auto-configures Honcho to use it — zero prompts, zero manual editing.

Interactive setup

./setup.sh              # Interactive credential prompts
sudo ./bootstrap.sh     # Deploy the stack

Open http://localhost:9090 for the onboarding dashboard or http://localhost:9090/api/v1/discover for the agent auto-discovery API.

Features

• Onboarding API — Agents hit /api/v1/discover to auto-configure to all local services
• One-command auto-deploy — sudo ./bootstrap.sh --auto generates everything: secure passwords, Ollama config, model pull, Honcho setup. Zero interaction needed
• Interactive Setup — ./setup.sh prompts for passwords, generates .env, no manual editing
• Local LLM (Ollama) — Run Honcho entirely offline with local model inference (opt-in via --with-local-llm or --auto)
• Cloud or Local — Choose between local Ollama or cloud OpenRouter during setup; .env.honcho configured automatically
• Tailscale Mesh — Each service gets its own Tailscale identity for secure mesh networking
• Cloudflare Tunnel — Optional public HTTPS access via Cloudflare Tunnel (no open firewall ports)
• Auto-Discovery — Gateway aggregates health and connection info for all backend services

Architecture

graph TB
    subgraph Public
        A[Cloudflare Tunnel]
    end
    subgraph Tailscale_Mesh
        B[Gateway :9090]
        C[SearXNG :8080]
        D[Qdrant :6333]
        E[Honcho :8000]
    end
    subgraph Local_LLM
        I[Ollama :11434]
    end
    subgraph Local_Network
        F[Camofox :9377]
        G[Obsidian :8083]
        H[CloakBrowser :9222]
    end

    A -->|HTTPS| B
    B -->|Discover API| C
    B -->|Discover API| D
    B -->|Discover API| E
    B -->|Discover API| F
    B -->|Discover API| G
    B -->|Discover API| H
    E ---->|LLM inference| I

    style B fill:#1a1a2e,stroke:#FFB300,color:#fff
    style C fill:#1a1a2e,stroke:#FFB300,color:#fff
    style D fill:#1a1a2e,stroke:#FFB300,color:#fff
    style E fill:#1a1a2e,stroke:#FFB300,color:#fff
    style I fill:#1a1a2e,stroke:#10b981,color:#fff

Loading

StackDeploy Dashboard is the control-plane island in the JorahOne archipelago — the single ingress through which agents discover and connect to every service.

Setup

First-time install

./setup.sh

This interactive script will prompt for:

1. Admin credentials — username/password for the gateway dashboard
2. Tailscale auth key — optional, for mesh networking (get one from https://login.tailscale.com/admin/settings/keys)
3. Cloudflare Tunnel token — optional, for public HTTPS access (create a tunnel at https://one.dash.cloudflare.com/)
4. LLM Provider — choose between:
   • (L)ocal Ollama — runs entirely on-device, no API key needed. You pick a default model (default: llama3.2)
   • (C)loud API (OpenRouter) — uses OpenRouter or any OpenAI-compatible API. You provide an API key
5. Service passwords — Honcho DB, Camofox API keys

All credentials are auto-generated if left blank. The .env.honcho file is generated automatically based on your LLM provider choice — no manual editing needed.

Auto-deploy (zero config)

# One-command deploy — generates .env, deploys stack, pulls Ollama model
sudo ./bootstrap.sh --auto

# With a different model
sudo ./bootstrap.sh --auto --model llama3.2:1b

# With custom model + Tailscale + Cloudflare
sudo ./bootstrap.sh --auto --with-tailscale --with-public

--auto does everything: generates a secure .env with random passwords, generates .env.honcho pointing to local Ollama, deploys all services, pulls the default model (llama3.2:1b), and restarts Honcho to activate it. The admin password is printed at the end.

Manual deploy

# Basic deploy (cloud API)
sudo ./bootstrap.sh

# With local LLM (Ollama) — auto-pulls model
sudo ./bootstrap.sh --with-local-llm

# With a specific Ollama model
sudo ./bootstrap.sh --with-local-llm --model qwen2.5:0.5b

# With Tailscale mesh
sudo ./bootstrap.sh --with-tailscale

# With Tailscale + Cloudflare Tunnel (public HTTPS)
sudo ./bootstrap.sh --with-tailscale --with-public

# With everything
sudo ./bootstrap.sh --with-tailscale --with-public --with-local-llm

# Skip setup prompt (use existing .env)
sudo ./bootstrap.sh --skip-setup

Local LLM (Ollama) Quick Start

When using --with-local-llm or --auto, Honcho is automatically configured to route all AI requests (memory summarization, reasoning, embeddings) to the local Ollama instance. The default 1B model is pulled automatically. To manage models:

# Pull additional models
docker exec ollama ollama pull llama3.2

# List available models
docker exec ollama ollama list

# Remove a model
docker exec ollama ollama rm qwen2.5:0.5b

The Ollama API is also discoverable via the gateway at http://ollama:11434 for other services to use. The Honcho config applies the same model across all features (deriver, summarization, dialectic, dream, embeddings) via per-feature env var overrides.

Agent Onboarding

Agents (Hermes, Claude Code, custom scripts) can auto-configure to the API stack by hitting the discover endpoint:

# Get all available services with connection details
curl -u admin:your-password http://localhost:9090/api/v1/discover

Response includes each service's internal URL, health status, and description. The onboarding page at http://localhost:9090/onboard shows a human-friendly dashboard.

Endpoints

Endpoint	Description
/ or /onboard	Human-friendly onboarding dashboard
/api/v1/discover	Agent auto-discovery JSON (auth required)
/api/v1/health	Aggregated health status of all services

Services

Service	Internal URL	Description
SearXNG	http://searxng:8080	Private meta-search engine
Qdrant	http://qdrant:6333	Vector database for semantic memory
Honcho	http://honcho:8000	AI memory & session management
Ollama	http://ollama:11434	Local LLM inference (opt-in)
Camofox	http://camofox-browser:9377	Browser automation
Obsidian	http://obsidian:8080	Notes & knowledge management
CloakBrowser	http://cloak-browser:9222	Protected browser

Contributing

1. Fork the repo
2. Create a branch: fix/your-fix or feature/your-feature
3. Open a PR against main
4. Response time: I read PRs within 48 hours

License

MIT — JorahOne LLC