ConsultDevin
ConsultDevin node.consultdevin.local · v0.9.3
Sovereign Local Runtime · node.consultdevin.local · v0.9.3

On-Premise Node Operator Dashboard

Private, local-first computing gateway installed directly on family estates and commercial facilities. Dual NVIDIA RTX 6000 Ada GPU priority with automatic AtlasCloud serverless AI mesh fallback for saturated or mobile operations. Zero external telemetry leakage and 100% cryptographic data custody.

Physical Hardware Fabric

Dual RTX Multi-GPU Telemetry · 96GB VRAM Pool

Combined Pool: 96 GB High-Bandwidth ECC Zero-Egress Shield: Active
GPU 0 · RTX 6000 Ada
48 GB VRAM
Temp: 52°C Power: 185W Fan: 42%
Workload: DeepSeek-R1 (Q4_K_M)
GPU 1 · RTX 6000 Ada
48 GB VRAM
Temp: 49°C Power: 140W Fan: 38%
Workload: vLLM Batched Serving
Clean 240V Power Circuit
242.4 V / 60 Hz
THD: < 1.5% Surge: Protected
✓ LiFePO4 Inverter Transfer Ready
DNS Telemetry Sinkhole
0 Drops / Sinkholed
Unbound: Running Cloud Egress: 0.00 B/s
✓ Air-Gapped Privacy Guaranteed
v0.9.3 Hybrid Engine Compute Failover Mesh

AtlasCloud Hybrid Compute Router · TTFT Telemetry

Local On-Premise Cluster PRIMARY
85 ms TTFT
Hardware: Dual RTX 6000 Ada (96GB)
Data Custody: 100% Zero-Egress
Egress Cost: $0.00 / token
Optimal for family estates, private medical records, and local industrial sensors.
AtlasCloud Serverless Mesh FALLBACK
220 ms TTFT
Endpoint: api.atlascloud.ai/v1
Models: DeepSeek-R1, Llama 3.3 70B
Trigger: Saturation or Headless Node
Zero-cold-start cloud mesh providing endless elasticity without hardware constraints.
Legacy Centralized Cloud DEPRECATED
1450 ms TTFT
Bandwidth: Heavy Egress Overhead
Data Custody: External Model Retention
Rate Limits: Strict Enterprise Tiers
Avoided by ConsultDevin LLC sovereign architecture policies.
• Automatic Failover Protocol: Monitors local VRAM saturation > 92% or local hardware absence; seamlessly hot-routes inference to https://api.atlascloud.ai/v1.
→ Test in API Console
Service Orchestration

Active Local Ports & Containers

Reverse Proxy: Caddy 2.8 Edge · mDNS: node.consultdevin.local
PORT 11434 ONLINE

Ollama Local LLM Engine

Native multi-model runtime serving quantizations of DeepSeek-R1, Llama 3.3, and Mistral directly from system VRAM.

Endpoint: http://localhost:11434
PORT 3000 ONLINE

OpenWebUI Private Chat Portal

Multi-user private web interface for family members and commercial staff with zero cloud account requirements.

Endpoint: http://localhost:3000
PORT 8000 ONLINE

vLLM Inference Server

PagedAttention high-throughput OpenAI-compatible API engine for back-office webhooks, MCP tools, and 24/7 AI concierges.

Endpoint: http://localhost:8000/v1
PORT 8188 ONLINE

ComfyUI Media & VFX Synthesis

Node-based visual effects generator for audio visualizers, custom POD artwork, and brand marketing collateral.

Endpoint: http://localhost:8188
PORT 8123 ONLINE

Home Assistant IoT Core

Local Zigbee/Matter orchestration monitoring LiFePO4 battery levels, well pump power, and solar generation.

Endpoint: http://localhost:8123
DDS DOMAIN 0 STANDBY

ROS 2 Mobile Robotics Bridge

Robot Operating System 2 local bridge communicating with AMRs, grounds equipment, and optical camera feeds.

Status: ✓ Physical E-Stop Armed