#llm

A lightweight, self-hosted Personal AI Daemon written in Go. Features Hybrid RAG (SQLite + Vectors), Long-term Memory, and Voice I/O. Optimized for low-end VPS and Raspberry Pi.

Go library of shared primitives for building MCP servers on the official modelcontextprotocol/go-sdk

Bundled LLM tokenizer vocabularies for @plurnk/plurnk-mimetypes' tokenizer seam — exact token counting for window math

Build-time-vendored snapshot of model metadata (context window + pricing) from models.dev — a fallback for plurnk when no live provider data is available.

SSH workflow helper for LLM agents

Durable, local-first memory for AI coding agents over MCP — zero-dependency, curated & semantically de-duped, you own the data (SQLite + Markdown). Works with Claude Code, Codex & any MCP host.

Persistent memory plugin for OpenCode. Obsidian-style knowledge base that survives across sessions