Ml Brainclone

ml-brainclone

A scaffold for building a personal AI second brain using **Claude Code** and **Obsidian**.

Larry is an AI-powered personal knowledge system that combines an Obsidian vault with Claude Code as the primary interface, plus specialized agents for image generation, audio/TTS, and privacy enforcement. Everything runs locally on your machine.

What You Get

ml-brainclone/
├── README.md                     <- You are here
├── ARCHITECTURE.md               <- System design, tri-modal architecture, freedom router
├── SETUP.md                      <- Step-by-step installation guide
├── CLAUDE-template.md            <- Project instructions template (becomes CLAUDE.md)
├── docs/                          <- incl. model-tiering.md, output-lifecycle.md
│   ├── architecture-overview.md  <- Agent architecture + data flows
│   ├── larry-setup.md            <- Larry (Claude Code) configuration
│   ├── barry-setup.md            <- Barry (image agent) setup
│   ├── harry-setup.md            <- Harry (audio agent) setup
│   ├── garry-setup.md            <- Garry (spatial agent) setup + Trellis 2 + Blender
│   ├── mempalace-setup.md        <- MemPalace (semantic memory) setup + GPU config
│   ├── memory-system.md          <- Persistent memory architecture
│   ├── parry-setup.md            <- Parry (gatekeeper agent) setup + commands
│   ├── tarry-setup.md            <- Tarry (temporal agent) setup + queue + Task Scheduler
│   ├── karry-setup.md            <- Karry (location agent) setup + geo-fences + MCP tools
│   ├── farry-setup.md            <- Farry (video agent) setup
│   ├── brains-bus-setup.md       <- SQLite event bus between agents, guarded by Parry
│   ├── task-dispatch.md          <- Inter-agent work queue: dispatch from any channel
│   ├── proactivity.md            <- Larry acts, doesn't just report: scanner + dispatcher + nightly triggers
│   ├── privacy-architecture.md   <- Privacy layers + auto-tagging + tone learning
│   ├── eval-smoketest.md