🦜
LangChain Orchestration
Full agent pipelines with RAG, memory, and tool use. Every query routes through LangChain.js connecting your local models to the Knowledge Yard.
RAGAgentsMemoryTools
🗑️
Knowledge Yard
A living pgvector archive where agents deposit everything they learn. Drag-and-drop any file. The Oracle synthesizes wisdom on demand.
pgvectorSemantic SearchAuto-depositOracle
🖥
Local LLM Inference
LMDeploy on an NVIDIA RTX 3090 via Tailscale VPN. 87 tokens/sec. Local-first inference with optional cloud provider fallback.
LMDeployRTX 3090PrivateTailscale
🦅
ZeroClaw Agents
Multi-provider agent routing: NVIDIA → OpenRouter → LMDeploy fallback chain. Agents browse, reason, and deposit findings automatically.
Multi-providerRoutingAutonomousOpenRouter
🧪
Autoresearch Pipeline
Automated experiment runner that dispatches training jobs to the OMEN GPU via SSH + tmux. Results auto-deposit into the Knowledge Yard.
tmuxGPU DispatchAuto-logSSH
📊
Full Observability
Prometheus scrapes every service. Grafana dashboards show token rates, HF sync stats, experiment metrics, and infrastructure health.
PrometheusGrafanaMetricsAlerting