🐉 DragonPi — Cybersecurity Toolkit for Raspberry Pi

⚠️ Vibe Coded Disclaimer: This project was built entirely with AI assistance (OpenCode Go + DeepSeek) as an experiment to see what could be accomplished. Use at your own risk. It is 100% vibe coded — no manual coding, no guarantees, no warranties. If it breaks, roll it back. If it works, you're welcome.

A complete, ready-to-deploy cybersecurity platform for Raspberry Pi 4/5. Web dashboard, AI chatbot, 40+ tools, one-click launchers, automated pentesting — everything a security professional needs.

Web dashboard — live system stats, 50+ pentest tools, AI chatbot, threat intelligence, and automated pentesting pipeline

📡 Quick Install

curl -sSL https://raw.githubusercontent.com/RufflezAU/dragonpi/main/install.sh | sudo bash

After install, open http://dragonpi.local in your browser.

💡 Recommended: Configure OpenCode API Key

The chatbot uses deepseek-v4-flash-free by default (no key needed). For better results — especially pentest self-healing and complex analysis — add a Pro API key for deepseek-v4-pro. The system auto-falls back to free if Pro is unavailable.

mkdir -p ~/.opencode
cat > ~/.opencode/config.json << 'EOF'
{
  "providers": {
    "opencode-go": {
      "api_key": "YOUR_API_KEY_HERE"
    }
  }
}
EOF
sudo systemctl restart chatbot

Get a key at opencode.ai. The free tier works without any key.

🌐 Optional: Remote Access via Tailscale

The installer includes Tailscale by default. Skip it with TAILSCALE=0:

curl -sSL https://raw.githubusercontent.com/RufflezAU/dragonpi/main/install.sh | TAILSCALE=0 sudo bash

After install, connect your Pi to your tailnet:

sudo tailscale up

Then install Tailscale on your laptop (macOS App Store, or curl -fsSL https://tailscale.com/install.sh | sh). Log in with the same account and you can access your Pi from anywhere:

http://<tailscale-ip>          # Dashboard
http://<tailscale-ip>/pentest   # Pentest console
ssh dragonpi@<tailscale-ip>     # SSH

Find your Pi's Tailscale IP with tailscale status. Pentest tools automatically use the physical network interface — Tailscale VPN traffic is never scanned.

🖥 Features

Feature	Description
Web Dashboard	Custom Flask dashboard engine with 50+ tools in 11 categories, live system stats, threat intel
AI Chatbot	OpenCode-powered AI assistant — chat to run tools, scan networks, manage the system. Auto-falls back to free model if Pro unavailable
Automated Pentesting	24+ tool pentest pipeline (external + internal). Per-tool checkboxes, DAST vulnerability scanning, compliance scoring (ISO 27001, NIST CSF, ASD E8, PCI DSS, HIPAA, CIS)
AI Self-Healing	When tools fail or return zero results, the Pi's AI diagnoses and attempts a fix — then retries automatically
AI Post-Pentest Review	After each pentest, AI reviews all tool output and writes an ai-review.txt with improvement suggestions
Premium Reports	Branded HTML + PDF with executive summary, compliance graphs (risk gauge, cadence bars, severity charts), CVSS scoring, remediation matrix, root cause analysis
Evidence Compilation	All raw tool output compiled into evidence.txt for manual review — every command, every result
Chain Scanning	Re-scan discovered subdomains, hosts, and URLs from the results page — select targets and launch batch scans
Job Queue	Run up to 2 pentests in parallel. Additional jobs auto-queue and start when a slot frees
Web Terminal	ttyd — full bash terminal in your browser, proxied through nginx
One-Click Launchers	Click any tool → terminal opens with 44 beginner-friendly guides
System Monitor	Live CPU/RAM/Disk/Temp with colored progress bars
RSS News Ticker	Scrolling cybersecurity headlines from 9 sources
Threat Intelligence	Live CVE feed, AlienVault OTX pulses, attack map
Podcast Player	Built-in player for cybersecurity podcasts (Risky Business, Darknet Diaries, Smashing Security)
Remote Access	Optional Tailscale VPN — access your Pi securely from anywhere, no port forwarding needed
Cockpit	System management web UI on port 9090

🧠 AI Chat Interface

The DragonPI AI Chat you're talking to is built on OpenCode Go (opencode.ai) — an AI coding assistant CLI. DragonPI wraps it with a custom Flask web UI (multi-session chat, tool integration, system control) and extends it with automated pentesting, threat intelligence, and dashboard features. The chat runs locally on your Pi — no cloud dependency beyond the AI model API.

🛠 Tools Included (50+)

Category	Tools
Vulnerability Scanning	Nmap (50+ CVE scripts), Nikto, SQLmap, SearchSploit, Nuclei (8000+ templates), OWASP ZAP, Dalfox
Network Scanning	Nmap, Masscan, Netdiscover, Naabu, Amass, Subfinder, TheHarvester, Httpx
Network Analysis	tcpdump, tshark, Netcat, Socat
Password Attacks	Hashcat, John, Hydra, Crunch, CeWL
Windows/AD	Enum4linux, SMB Client, Impacket, Certipy, BloodHound, Coercer
Web Application	Gobuster, ffuf, Wfuzz, Dirb, WhatWeb, Wafw00f
MITM & Tunneling	Mitmproxy, Ettercap, Chisel, SSHuttle, Proxychains
Forensics	Autopsy (Web UI), Binwalk, Foremost, ExifTool, Steghide, dc3dd, Sleuthkit
Wireless	Aircrack-ng, Wifite, Reaver, PixieWPS, airbase-ng, Horst, hcxdumptool
SSL/TLS	testssl.sh, sslscan

🤖 AI Chatbot — Model Failover

The chatbot supports automatic model fallback if the primary AI model is unavailable (API key expired, no credits, quota exceeded, etc.).

How it works

1. The chatbot tries models in priority order (defined by the OPENCODE_MODELS environment variable)
2. If the primary model fails — especially with auth/credit errors — it automatically falls back to the secondary model
3. The response includes a _model field indicating which model answered

Default model priority

Priority	Model	Type
1st	opencode-go/deepseek-v4-pro	Primary — Go subscription
2nd	opencode/deepseek-v4-flash-free	Fallback — free tier

Configuration

Set OPENCODE_MODELS in the systemd service file as a comma-separated list (no spaces):

Environment=OPENCODE_MODELS=opencode-go/deepseek-v4-pro,opencode/deepseek-v4-flash-free

🏗 Architecture

dragonpi.local (port 80)
├── Custom Dashboard    → / (Flask :5000, nginx proxy)
├── DragonAI Chatbot   → /chat → Flask :5000 → OpenCode CLI (multi-model)
├── Pentest Console    → /pentest → Flask → 24+ tool pipeline (per-tool toggles)
├── Podcast Player     → /podcast → Flask → RSS feeds (client-side audio)
├── Web Terminal       → /terminal/ → nginx → ttyd :7681
├── Cockpit            → :9090
├── RSS Feed           → /rss → Flask proxy (9 sources)
├── Threat APIs        → /api/threat/* → CVE, OTX, live threats
└── Tool Launcher      → /term/TOOL → redirect → terminal

📁 File Structure

dragonpi/
├── install.sh                  # One-line installer
├── chatbot/
│   ├── app.py                  # Flask backend (dashboard, chat, podcast, pentest routes)
│   ├── pentest.py              # 24+ tool pentest engine + self-healing + AI review
│   └── templates/
│       ├── dashboard.html      # Custom dashboard UI
│       ├── index.html          # Chat UI
│       ├── podcast.html        # Podcast player
│       ├── pentest.html        # Pentest console UI
│       └── report_template.html # Premium HTML report template
├── dashboard/
│   └── config.yml              # Dashboard configuration (services, links, categories)
├── nginx/
│   └── dragonpi.conf           # Nginx reverse proxy config
├── services/
│   ├── chatbot.service         # DragonAI systemd unit (with model env vars)
│   ├── ttyd.service            # Web terminal systemd unit
│   └── autopsy.service         # Autopsy forensic UI systemd unit
├── scripts/
│   ├── dragonpi-shell          # Terminal wrapper (auto-runs tools from dashboard)
│   └── guides/                 # Beginner-friendly guide scripts for each tool
├── AGENTS.md                   # Project memory (gitignored — local only)
├── secrets.md                  # Credentials (gitignored — never committed)
└── .gitignore

🤖 AI Self-Healing & Post-Pentest Review

Self-Healing

When tools fail or return zero results unexpectedly, the Pi's OpenCode AI:

1. Diagnoses the failure (missing package, wrong CLI flag, broken config)
2. Fixes with a safe, non-interactive bash command
3. Retries the tool automatically
4. Logs the diagnosis and result

Triggers: nmap 0 hosts, nikto 0 findings, nuclei 0 vulns, ZAP 0 alerts (both modes).

Post-Pentest AI Review

After each pentest finishes:

1. All raw tool output is compiled into evidence.txt
2. The AI reviews it for anomalies, missed findings, tool configuration issues
3. An ai-review.txt is saved with improvement suggestions
4. Both files appear in the report downloads

Per-Tool Toggles

Every tool in the pipeline has its own checkbox — enable/disable individual scanners. External mode: 18 tools. Internal mode: 12 tools. All checked by default.

Evidence File

evidence.txt contains every raw tool command output + the full pentest log. Download it from the reports list or via GET /api/pentest/log/<job_id>.

🔧 Development

git clone https://github.com/RufflezAU/dragonpi.git
cd dragonpi
sudo ./install.sh

To update an existing install:

cd /opt/dragonpi
git pull
sudo ./install.sh

📋 Requirements

• Raspberry Pi 4 or 5 (4GB+ RAM recommended)
• Raspberry Pi OS (Debian 12/13, 64-bit)
• 32GB+ SD card
• OpenCode account (free tier works — subscription unlocks premium models)

🔐 Security Notes

• No secrets in git: password files, API keys, .env files, and *secret* patterns are all gitignored
• secrets.md and AGENTS.md are local-only by design (listed in .gitignore)
• The chatbot authenticates via the OpenCode CLI — no API keys stored in the codebase
• All dashboard config is served from /opt/dragonpi/dashboard/config.yml

📜 License

MIT — Use responsibly. Only test systems you own or have permission to test.