ZERO TRUST · WIREGUARD · E2E ENCRYPTED
FRONTEND ENTRY
NETWORK
APP RUNTIME
ORCHESTRATION
KNOWLEDGE
STORAGE / EXTERNAL APIs
INFERENCE
HARDWARE
OBSERVABILITY
NUCLEUS
GENTOOFOO FRONTEND · WIREFRAME-DRIVEN UI
Wireframes define layout, structure, and component hierarchy.
HTTPS
VPN
:443
managed
orchestrate
workflows
embed
RAG
richer context
KY grows → smarter
agents
deposit
trigger
① default
① agents
② fallback
inference
stores
fetch model
recipe→junk
GPU recipe
CPU recipe
pull model
serves
③ HF free
datasets
CUDA
GPU
infer
infer
dispatch
SSH→WSL
auto-deposit
tunnel
tunnel
publish
subscribe
events
tasks
deposits
all topics
pipeline
metrics
:9200
:9210
datasource
metrics
logs
dashboards
heartbeat
metrics
logs
traces
orchestrate
reason
traces
spans
gentoofoo
Frontend
⚡ Traefik
🔒 Tailscale VPN
🔐 RBAC/Auth
▲ Next.js 15
🔄 PM2
🦜 LangChain.js
⚙️ n8n Workflows
🧮 Embeddings
🗑️ Knowledge
Knowledge Yard
🧪 Distillery
⚡ NVIDIA GPU
NIM API
🐘 pgvector
⚡ Kafka
(Redpanda)
🤗 HuggingFace
Hub
🔀 OpenRouter
(free tier)
🚀 LMDeploy
Local GPU
🦅 ZeroClaw
守護Agents
🖥 LMDeploy
💙 Intel CPU
OpenVINO
🎮 NVIDIA 3090
24 GB
🧬 GPU Research
Pipeline
📺 tmux
Dispatcher
🔥 Prometheus
📋 Loki
📊 Grafana
🚗 TeslaMate
📡 GLT
Grafana+Loki+Tempo
🧠 Opencode
Nucleus
🔍 Tempo
⚡ MESSAGE BUS · 伝言バス · CLICK TO EXPLORE ⚡
LANGCHAIN.JS · VLLM · N8N · PGVECTOR · NEXT.JS 15 · NVIDIA NIM · HUGGINGFACE HUB · ZEROCLAW
← Home
·
📐 Platform Architecture
🔓 Public View
System Boundary
Opencode Universe
System Boundary
Full flow: Frontend → Agents → LangChain Cognitive Router → Multi-Model Inference
architecture
agents
langchain
inference
failover
istio
GENTOOFOO FRONTEND · WIREFRAME-DRIVEN UI
Dashboard
UI · Charts · Navigation
Knowledge Yard Interface
Distillery
Model Optimization
HuggingFace Integration
Autoresearch
GPU Pipeline Console
Experiment Monitor
Model Console
Inference · Provider Mgmt
Token Metrics
Wireframes define layout, structure, and component hierarchy.
API Request
AGENT LAYER · OPENCODE + ZEROCLAW · LANGCHAIN COGNITIVE ROUTER
Agents Decide · LangChain Routes · Tool Calls · Memory Retrieval · RAG Queries · Inter-Agent Comms
Agents (Opencode + ZeroClaw) are autonomous decision-makers. They hold state, maintain goals, and choose actions.
LangChain is the cognitive router: structures multi-step reasoning, routes tool calls, memory, and RAG queries.
n8n handles automation (ETL, scheduled tasks, self-healing) — separate from the cognitive loop
Agent Decision
AGENTS · AUTORESEARCH PIPELINE
Research Agent
Knowledge gathering · RAG queries
Hypothesis generation
Research
Training Agent
GPU dispatch · SSH + tmux
train.py · val_bpb tracking
Train
Evaluator Agent
val_bpb scoring · model diff
Keep / Discard · auto-deposit
Autoresearch Loop ↺ Research → Train → Evaluate → Repeat
Inference Request
MULTI-MODEL INFERENCE · PRIORITIZED FAILOVER CHAIN
① HuggingFace
Inference API
DEFAULT
mistral-7b-instruct
llama-3-8b-instruct
phi-3 · zephyr-7b
api-inference.huggingface.co
HF_API_TOKEN · /v1/chat
fail
② NVIDIA API
NIM Microservices
FALLBACK #1
llama-3-70b-instruct
nemotron-4-340b
qwen-72b · mixtral-8x7b
build.nvidia.com
NVIDIA_API_KEY · free credits
fail
③ OpenRouter
200+ Free Models
FALLBACK #2
deepseek-r1 · qwen3.5-9b
claude-3-haiku (free tier)
200+ models via /v1 API
openrouter.ai
OPENROUTER_API_KEY · free tier
fail
④ Home Node
Ollama / LMDeploy
FALLBACK #3
RTX 3090 · 24 GB VRAM
gpt-oss-20b · llama-3-8b
On-premises · Tailscale VPN
OMEN-3090 · :1234 / :11434
100.71.193.62 via Tailscale
Response ↑
NOTE
Agents (Opencode + ZeroClaw) decide → LangChain routes reasoning → LLMs respond. Istio Ambient Mesh wraps all infra.
n8n orchestrates automation (ETL, self-healing, scheduled tasks) outside the cognitive loop.
Failover chain: HuggingFace → NVIDIA API → OpenRouter → Home Node (LMDeploy). Mesh: mTLS + L4/L7 + unified telemetry.
gentoofoo-system-boundary · v1.0
📐 Blueprint-style SVG — renders at full resolution at any scale
🖨 Print or export via browser → Save as PDF
🔓 Public view — sign in for editing tools
🗑️
⚔️