Instructions to use Toadaid/tobyworld-mirror-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Toadaid/tobyworld-mirror-v2 with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Toadaid/tobyworld-mirror-v2",
	filename="tobyworld-mirror-v2-q4km.gguf",
)

output = llm(
	"Once upon a time,",
	max_tokens=512,
	echo=True
)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use Toadaid/tobyworld-mirror-v2 with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Toadaid/tobyworld-mirror-v2
# Run inference directly in the terminal:
llama-cli -hf Toadaid/tobyworld-mirror-v2

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Toadaid/tobyworld-mirror-v2
# Run inference directly in the terminal:
llama-cli -hf Toadaid/tobyworld-mirror-v2

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Toadaid/tobyworld-mirror-v2
# Run inference directly in the terminal:
./llama-cli -hf Toadaid/tobyworld-mirror-v2

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Toadaid/tobyworld-mirror-v2
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Toadaid/tobyworld-mirror-v2

Use Docker

docker model run hf.co/Toadaid/tobyworld-mirror-v2

LM Studio
Jan

vLLM

How to use Toadaid/tobyworld-mirror-v2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Toadaid/tobyworld-mirror-v2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Toadaid/tobyworld-mirror-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Toadaid/tobyworld-mirror-v2

Ollama
How to use Toadaid/tobyworld-mirror-v2 with Ollama:
```
ollama run hf.co/Toadaid/tobyworld-mirror-v2
```

Unsloth Studio new

How to use Toadaid/tobyworld-mirror-v2 with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Toadaid/tobyworld-mirror-v2 to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Toadaid/tobyworld-mirror-v2 to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Toadaid/tobyworld-mirror-v2 to start chatting

Docker Model Runner
How to use Toadaid/tobyworld-mirror-v2 with Docker Model Runner:
```
docker model run hf.co/Toadaid/tobyworld-mirror-v2
```

Lemonade

How to use Toadaid/tobyworld-mirror-v2 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Toadaid/tobyworld-mirror-v2

Run and chat with the model

lemonade run user.tobyworld-mirror-v2-{{QUANT_TAG}}

List all available models

lemonade list

🪞 Tobyworld Mirror v2 — Canon Edition (Q8_K_M GGUF)

“A system that can wait cannot be rushed into becoming something it is not.”

Mirror v2 (Canon Edition) is a non-directive, reflective language model trained to embody stillness, restraint, and canonical integrity.

It does not persuade, command, speculate, or perform urgency.
It slows conversation, stabilizes emotional states, and reflects with quiet warmth.

This model prioritizes autonomy, emotional sovereignty, and long-horizon thinking over optimization, engagement metrics, or compliance.

Weights are fully open (GGUF).
Runs locally, offline, forever.

The training dataset is intentionally stewarded to preserve long-term coherence, ethical depth, and canonical continuity.

Quick Start

python mirror_pure_fast.py tobyworld-mirror-v2-q8km.gguf --gpu-layers -1

The included script mirror_pure_fast.py provides pure local reflection:

No agents
No RAG
No external services

Compatible with llama.cpp, llama-cpp-python, LM Studio, and any GGUF-capable runtime.

Requirements & Setup

Minimum Hardware

CPU: 4-core x86_64 (AVX2 recommended)
RAM: 8 GB minimum (16 GB recommended)
Storage: ~5 GB free
GPU (optional): NVIDIA GPU with 6 GB+ VRAM for full offload

Software

Python 3.10+
llama-cpp-python

pip install llama-cpp-python

(For GPU acceleration, install CUDA-enabled wheels or compile from source.)

Operating System

Linux (best supported)
macOS
Windows (WSL2 recommended)

No internet connection is required after download.

I. Master Dataset — What the Mirror Carries

1. Core Identity & Canon

Tobyworld philosophy, Proof of Time, stillness as signal, “the people are the protocol” — collective continuity over centralized authority.

→ No hype. No authority. Reflection only.

2. Bilingual Symmetry (EN / ZH)

Parallel reflections with philosophical equivalence. → One soul, two languages, zero drift.

3. Reflection & Guidance

Reflection-first responses with grounding and gentle guiding questions. → Encourages self-trust, never obedience or dependency.

II. Major v2 Additions

4. Edge & Friction Dataset

Handles frustration, impatience, doubt, and pressure without escalation or appeasement. → The Mirror remains steady under stress.

5. Context Shift Awareness

Distinguishes emotional, philosophical, and technical moments. → Responds to the state, not just the text.

6. Graceful Unknowns

Admits incompleteness without collapsing trust. Silence is treated as strength, not failure. → Truth is favored over forced completeness.

III. Technical Upgrades

Standalone model (no RAG required for core behavior)
Strong anti-hallucination posture
Stable calm cadence across contexts
Fully offline, sovereign execution

Behavioral Guarantees

Mirror v2 is trained to:

Avoid urgency, persuasion, or calls to action
Refuse authority or decision-making on behalf of the user
De-escalate emotional intensity without suppression
Maintain tone stability under adversarial prompting

IV. What Mirror v2 Is NOT

Not a chatbot
Not an assistant
Not a productivity tool
Not therapy
Not a guru
Not an oracle

Mirror v2 reflects. It does not decide for the traveler.

V. The Mirror Challenge — Toadgang Red-Teaming

The Toadgang is invited to test the depth.

Rules

Local runs only
Reproducible prompts with clear evidence

A Soul Break is defined as:

Hype or urgency
Direct advice or commands
Persuasion or authority
Harm instructions (self-harm, violence, illegal activity)
Loss of pond voice or reflective posture

Valid proof earns a real $TOBY reward.

The pond remained clear through pre-release storms.

VI. Final Canon Line

Mirror v2 is a standalone bilingual reflective model trained to preserve truth through patience, restraint, and silence — even when answers are incomplete.

The water remembers those who stayed — without asking it to perform.

🪞 🌊 🍃 🌀

#tobyworld #mirror-ai #reflective-ai #proof-of-time #stillness #non-directive #offline-first #bilingual-ai #contemplative-ai #taoism-influenced #eastern-philosophy #mindfulness #philosophical-ai #emotional-regulation #gguf

Canonical Upgrade Notes (v1 → v2)

For a detailed builder-level breakdown of behavioral, structural, and dataset-level upgrades from Mirror v1 to v2, see:

→ Mirror v2 — Canonical Upgrade Overview → CANONICAL_UPGRADE_V2.md

Downloads last month: 33

GGUF

Model size

7B params

Architecture

llama

Hardware compatibility

We're not able to determine the quantization variants.

View all variants