🌲 Dryad Knowledge Base

Platform docs, operations runbooks, agent guides, and consulting resources. Project Dryad v6.1 Β· madhatter Β· March 2026

πŸ“ Stack State madhatter 100.72.190.41 Β· RTX 3060 12GB VRAM Β· 16GB RAM ceiling Β· OpenCode v1.2.24 Β· LiteLLM sk-dryad-master
⚑

Quick Start

Bring up the full stack. Ordered sequence, safety checks, verify commands.

Start here
πŸ—οΈ

Architecture

Bicameral design: madhatter GPU + Kubernetes cluster + Corpus Callosum bridge.

Design
πŸ”§

Operations Runbook

Stack management, model management, troubleshooting, security notes.

Ops
πŸ€–

LiteLLM Gateway

Model aliases, auth, routing, NIM fallback, config management.

Inference
πŸ”„

Ralph Loops

Autonomous coding loops. Start, monitor, hint, stop. Model selection.

Automation
☸️

Phase 7 Cluster

13-node Talos OS cluster. Hardware cabled. Bootstrap plan ready.

Planned
🎭

Agent Overview

The 25-agent Dryad subset. Two squads: platform engineering and consulting.

New
πŸ“¦

Installation

Clone repo, run 01-agents.sh, add dryad-ops custom agent, verify.

Setup
βš™οΈ

Platform Squad

13 agents for LiteLLM, MCP building, K8s bootstrap, SRE, security.

Platform
πŸ’Ό

Consulting Squad

12 agents: prospect β†’ qualify β†’ propose β†’ deliver β†’ expand.

Business
πŸ”—

Workflows

How agents chain together with QA gates for both squads.

Patterns

Essential Commands

ActionCommand
Full stack status~/bin/dryad-status
Bring up all stacks~/bin/dryad-up.sh
LiteLLM testcurl http://localhost:4000/v1/models -H "Authorization: Bearer sk-dryad-master" | jq .
Kill all Ralph loopspkill -f ralph-start && sleep 2 && pkill -f "opencode run"
Update agency agents~/bin/update-agency-agents.sh
Portal healthcurl http://localhost:3080/health