Wondermotor AI — Local AI Tools

Wondermotor AI

Open tools for running local AI

tq

Available

Manager for TurboQuant+ by TheTom — run local LLMs with auto-configured KV cache compression

Install — macOS (Apple Silicon) & Windows (CUDA)

$ curl -fsSL https://wondermotor.com/install.sh | bash

3.8x KV Compression

turbo4 compresses V cache to 4.25 bits — 8x more context in the same RAM.

Auto-Config

Detects GPU, RAM, quant type. Calculates optimal cache settings automatically.

OpenAI-Compatible

/v1/chat/completions. Works with any client — curl, Pi, LM Studio.

Idle Auto-Stop

Stops after 5 min idle. Like Ollama — no wasted resources.

Without vs With tq

	Without	With tq
8GB RAM, 3B model	4,096 context	32,768 context
KV cache memory	f16 (full size)	3.8× compressed
Config	Manual flags	Auto-detected
Idle behavior	Always running	Auto-stops

Quick Start

Install: curl -fsSL https://wondermotor.com/install.sh | bash
List models: tq list
Search & download: tq search "qwen2.5 coder 7b" then tq download <model>
Serve: tq serve 1 — TurboQuant auto-configured.

Commands

Command	Description
`tq list`	List local GGUF models
`tq search <query>`	Search HuggingFace
`tq download <model>`	Download with SHA256 verify
`tq serve 1`	Launch with auto TQ config
`tq serve 1 --dry-run`	Preview command
`tq status`	Check running server
`tq stop`	Stop the server
`tq logs`	View server logs
`tq install`	Install TurboQuant+ binary
`tq doctor`	Verify setup
`tq config show`	Show/edit config

More tools coming soon

Wondermotor AI is building more open tools for local AI. Stay tuned.