COMMANDER
Workstation WorkhorseLODY
Raspberry Pi 4B (4GB)MAINY
Dell Linux Server (i3)CLOUD REASONING
Google DeepMind APIUnified Pipeline Architecture
Per system design, models cannot be selected manually. All queries are automatically orchestrated through our high-efficiency Two-Tier Cognitive Workflow: queries hit an ultra-fast Small Language Model (SLM) first to enrich context and classify intent, before fan-out to heavy code synthesis models or cloud reasoning.
OPERATOR INPUT
Raw natural language query or task received via Web, Terminal, or API.
Input IngestionSTAGE 1: REFLEX SLM
Lody (RPi 4B) runs qwen2.5:0.5b to classify intent, extract technical specs, and expand the prompt.
STAGE 2: HEAVY EXECUTION
Mainy (Dell PC) runs qwen2.5-coder:1.5b for full technical synthesis and code generation.
OUTPUT & TELEMETRY
Real-time SSE token stream delivered with TTFT, TPS, and token consumption metrics.
Unified StreamInstalled Cluster Models
| MODEL | PARAMETERS | HOST NODE | PIPELINE ROLE | STATUS | SELECTOR |
|---|---|---|---|---|---|
| qwen2.5:0.5b | 0.49 Billion | Lody (RPi 4B) | Stage 1: Intent Analysis & Prompt Enhancement | ACTIVE | 🔒 Autonomous |
| qwen2.5-coder:1.5b | 1.54 Billion | Mainy (Dell PC) | Stage 2: Technical & Code Synthesis Engine | ACTIVE | 🔒 Autonomous |
| qwen2.5:1.5b | 1.54 Billion | Mainy (Dell PC) | Stage 2: General Conversational Reasoning | ACTIVE | 🔒 Autonomous |
| gemini-2.0-flash | Cloud MoE | Google Cloud | Deep Multi-Modal & Massive Context Reasoning | STANDBY | 🔒 Autonomous |
Prompt Dispatcher
Type any request. Watch Stage 1 SLM enhance your prompt in real-time before Mainy generates the final output.
Real-Time Cluster Stream
// Output will stream here in real-time...Cluster API Tokens
Use this Bearer token to connect mobile clients, external servers, or cron automations to your Jarvis Core.
Quickstart Code Snippets
curl -N -X POST http://127.0.0.1:8000/api/chat \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-jarvis-cluster-daksh-1416-prod" \
-d '{"prompt": "Write a python port scanner"}'Is this setup local-only right now?
YES — By default, your cluster IP addresses (192.168.1.107, 192.168.1.108, 192.168.1.110) are private RFC-1918 LAN addresses assigned by your home Wi-Fi router.
🏠 Current State: Home LAN Only
- Works when your laptop/phone is connected to the same Wi-Fi router.
- Fails if you switch to 5G cellular, mobile hotspot, or office/college Wi-Fi.
- Zero exposure to the public internet (secure, but location-bound).
🚀 Solution: Zero-Trust Global Tunnel
- Deploy a lightweight Cloudflare Tunnel (
cloudflared) on your PC or Dell server. - Gives you an encrypted, global HTTPS endpoint (e.g.
https://jarvis.daksh.pages.dev). - Connect from any router, phone, or country with zero port forwarding required!
Quick 1-Command Cloudflare Tunnel (Free & Instant)
Run this command on your Commander workstation to expose Jarvis to the world securely:
# Instant secure global tunnel without configuration:
npx cloudflared tunnel --url http://127.0.0.1:8000Cloudflare will print a free https://<random>.trycloudflare.com URL. Paste that URL into the ENDPOINT box at the top right of this dashboard to control Jarvis from anywhere on earth!