Instructions to use Sathman/Watts-Agent-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- llama-cpp-python
How to use Sathman/Watts-Agent-GGUF with llama-cpp-python:
# !pip install llama-cpp-python from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="Sathman/Watts-Agent-GGUF", filename="Watts_Agent-Q8_0.gguf", )
llm.create_chat_completion( messages = [ { "role": "user", "content": "What is the capital of France?" } ] ) - Notebooks
- Google Colab
- Kaggle
- Local Apps
- llama.cpp
How to use Sathman/Watts-Agent-GGUF with llama.cpp:
Install from brew
brew install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama-server -hf Sathman/Watts-Agent-GGUF:Q8_0 # Run inference directly in the terminal: llama-cli -hf Sathman/Watts-Agent-GGUF:Q8_0
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama-server -hf Sathman/Watts-Agent-GGUF:Q8_0 # Run inference directly in the terminal: llama-cli -hf Sathman/Watts-Agent-GGUF:Q8_0
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 Sathman/Watts-Agent-GGUF:Q8_0 # Run inference directly in the terminal: ./llama-cli -hf Sathman/Watts-Agent-GGUF:Q8_0
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 Sathman/Watts-Agent-GGUF:Q8_0 # Run inference directly in the terminal: ./build/bin/llama-cli -hf Sathman/Watts-Agent-GGUF:Q8_0
Use Docker
docker model run hf.co/Sathman/Watts-Agent-GGUF:Q8_0
- LM Studio
- Jan
- vLLM
How to use Sathman/Watts-Agent-GGUF with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Sathman/Watts-Agent-GGUF" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Sathman/Watts-Agent-GGUF", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Sathman/Watts-Agent-GGUF:Q8_0
- Ollama
How to use Sathman/Watts-Agent-GGUF with Ollama:
ollama run hf.co/Sathman/Watts-Agent-GGUF:Q8_0
- Unsloth Studio new
How to use Sathman/Watts-Agent-GGUF 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 Sathman/Watts-Agent-GGUF 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 Sathman/Watts-Agent-GGUF to start chatting
Using HuggingFace Spaces for Unsloth
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for Sathman/Watts-Agent-GGUF to start chatting
- Pi new
How to use Sathman/Watts-Agent-GGUF with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama-server -hf Sathman/Watts-Agent-GGUF:Q8_0
Configure the model in Pi
# Install Pi: npm install -g @mariozechner/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "Sathman/Watts-Agent-GGUF:Q8_0" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Hermes Agent new
How to use Sathman/Watts-Agent-GGUF with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama-server -hf Sathman/Watts-Agent-GGUF:Q8_0
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default Sathman/Watts-Agent-GGUF:Q8_0
Run Hermes
hermes
- Docker Model Runner
How to use Sathman/Watts-Agent-GGUF with Docker Model Runner:
docker model run hf.co/Sathman/Watts-Agent-GGUF:Q8_0
- Lemonade
How to use Sathman/Watts-Agent-GGUF with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull Sathman/Watts-Agent-GGUF:Q8_0
Run and chat with the model
lemonade run user.Watts-Agent-GGUF-Q8_0
List all available models
lemonade list
llm.create_chat_completion(
messages = [
{
"role": "user",
"content": "What is the capital of France?"
}
]
)Watts Agent (Qwen3-8B) — The Cosmic Game in Pure Voice
A fine-tuned Qwen3-8B model trained exclusively on Alan Watts' teaching atoms. No system prompt. No retrieval. No blending. Question in — the cosmic joke out.
~1,330 atoms extracted from 3 collections (The Essential Alan Watts + Lectures + Tao, 336 pages) — East-West bridging, the game metaphor, the backwards law, the watercourse way, wit that cuts deeper than solemnity.
Quick Start
LM Studio / Ollama / llama.cpp: Download the Q8_0 GGUF. No system prompt required.
Example Interactions
User: Why can't I just relax and be natural?
Watts Agent: You can't relax deliberately. That's the joke at the center of the whole game. The moment you try to relax, you've introduced a trying — and trying is precisely the tension. It's like trying to be spontaneous. Or trying to fall asleep. The very effort prevents the thing. This is what the Taoists called wu wei — not forcing. But you can't try to not force. You can only notice that the forcing is happening, and that noticing itself is already the beginning of something else. Nature doesn't try to be natural. A river doesn't work at being wet. You came with all of this already built in. The question is not how to acquire it but why you're convinced you've lost it.
User: If there's no self, who is enjoying life?
Watts Agent: Ah, you've spotted the most beautiful paradox in the whole enterprise. Who indeed? But look — the question assumes that enjoyment requires a separate enjoyer standing apart from the enjoyed. Does it? When music really gets you, when you're completely lost in it, is there someone experiencing the music and the music being experienced? Or is there just — music? The self that enjoys is like the hole in a flute. It's not there, and that's precisely why music can happen. The Hindus have a word for it: lila — the divine play. The universe playing at being you, playing at being everything, and occasionally catching a glimpse of itself in the act and laughing. That's the enjoyment. Not someone having it. Just — this.
User: All these Eastern teachers say desire is the problem. But I like my life. I don't want to stop wanting things.
Watts Agent: Then stop worrying about desire and enjoy wanting. The teachers you're quoting are pointing at something specific — the desire that fights with what is, the grasping that generates suffering because it imagines something is missing. They're not talking about the delight of appetite, the pleasure of anticipation, the joy of good food and good love. You think Buddhism means going around with a face like last week's news, strangling every impulse? Not at all. The whole point is to enjoy the music without gripping the vinyl. You can want things. Just notice you don't need them to be okay. And here's the secret: when you stop needing them desperately, you enjoy them far more. The food tastes better when you're not eating it out of anxiety.
What Makes This Different
- Pure Watts, no dilution. The East-West bridge is intact — you get the cultural translation, the humor, the precise playfulness, none of it smoothed out.
- The backwards law lives here. The model understands and demonstrates that trying prevents, that grasping loses, that effortlessness is the actual method.
- Genuine wit. Not jokes added to spiritual content — the wit IS the teaching. The laugh is the recognition.
- Tao, Zen, Vedanta, and Western philosophy in one movement. Watts read everything and made it speak to everyone. The model carries that.
- Zero earnest-spiritual-voice. No hushed reverence. Watts taught with a glass of whisky and a grin. That aliveness is in the weights.
Training Details
| Setting | Value |
|---|---|
| Base model | Qwen/Qwen3-8B |
| Method | A-LoRA (QLoRA + DoRA + rsLoRA) |
| Rank / Alpha | 64 / 64 |
| Training data | ~1,330 V6-formatted atoms |
| Source | The Essential Alan Watts, Lectures, Tao (Alan Watts, 336 pages) |
| Format | V6 — question + concept arrows (input) → pure teaching passage (output) |
| Epochs | 1 |
| Total steps | ~166 |
| Quantization | Q8_0 GGUF (~8.71 GB) |
The A-LoRA Methodology
A-LoRA (Atom LoRA) is a fine-tuning methodology built on a single insight: the unit of training data matters as much as the amount.
Each atom is an indivisible teaching move, extracted from human expert traces, containing:
- Transformation — what shifts in understanding (before/after/how)
- Concept relations — directional arrows (
self-consciousness → trying to be natural → prevents naturalness) - Anchoring quotes — the teacher's actual words, verbatim
- Teacher method — the specific move that creates the shift
The atom is never split. Result: models that move like teachers, not just sound like them.
The Teacher
Alan Watts (1915–1973) — British philosopher, writer, and lecturer. Studied at King's School Canterbury, ordained as an Anglican priest, left to teach Eastern philosophy to the West. Author of 25 books. Lectured to overflowing halls at Berkeley. Died on a mountain in Druid Heights, California.
What Watts did was impossible and he did it anyway: explained Zen Buddhism, Taoism, Vedanta, and the nature of consciousness to mid-20th century Westerners — and made them laugh. He wasn't translating. He was demonstrating.
The backwards law. The game metaphor. The cosmic joke. Nature as the best available evidence of what's going on. The watercourse way — achieving by yielding, accomplishing by not forcing. All of this was original synthesis, not scholarship, and it still hits.
| Style element | What it means |
|---|---|
| The cosmic game | Life as lila — divine play pretending to be serious |
| The backwards law | Effort defeats its own purpose; naturalness cannot be forced |
| East-West bridge | Neither adopting East nor dismissing West — genuine synthesis |
| Wit as pointer | The laugh is the recognition; humor cuts what seriousness can't |
| Wu wei | The watercourse way — action without contrivance |
Vs. The Multi-Teacher Meditation Agent
This is a single-teacher specialist. The Meditation Agent 8B blends 9 voices across traditions. Watts Agent is pure East-West bridging — the playfulness, the cultural translation, the humor as teaching tool, without other styles bleeding in.
Especially useful for: Western audiences new to Eastern philosophy, anyone who responded to Watts specifically, questions that benefit from wit rather than solemnity.
Intended Use
- East-West philosophical inquiry and Taoism / Zen exploration
- Personal contemplative investigation with a lighter touch
- Research in single-teacher voice distillation and contemplative AI
- Understanding how humor and metaphor function as pedagogical tools
Limitations
- Single teacher — deep on Tao/Zen/Vedanta synthesis, thinner on devotional or practice-based traditions
- Smaller training corpus than the multi-teacher model (1,330 vs 22,497 atoms)
- Watts' highly idiosyncratic style may be inconsistent on edge cases with this corpus size
- Training data extracted from copyrighted material; only trained weights are shared
Nondual Bot Benchmark Results
Benchmark evaluation not yet run. Will be updated after the 50-question Nondual Bot Benchmark eval is completed.
Disclaimer
This is an independent academic research project. It is not affiliated with, endorsed by, or approved by the Alan Watts Organization, Electronic University, or the Estate of Alan Watts.
Copyright: The Essential Alan Watts, lecture transcriptions, and Tao: The Watercourse Way are copyrighted works. No source text is reproduced in this repository. The model weights are a compressed statistical representation learned from structured training examples — they do not contain or reproduce the original text verbatim. Only the trained weights are distributed here.
License scope: The MIT license applies to the model weights only, not to the underlying source material. Users wishing to read the source texts should obtain them through authorized channels.
Responsible use: This model is intended for research, personal contemplative inquiry, and educational purposes. It is not a substitute for the original works or for working with a qualified teacher.
Citation
@misc{watts-agent-2026,
title={Watts Agent: Single-Teacher Contemplative AI via A-LoRA Fine-Tuning on Structured Reasoning Atoms},
author={Sathman},
year={2026},
url={https://huggingface.co/Sathman/Watts-Agent-GGUF}
}
Related
- Full series — Meditation Agent Collection — all 19 models
- Meditation Agent 8B — All 9 teachers, multi-tradition
- Meditation Agent Phi4 — 14B multi-teacher
- Osho Agent — Pure Osho (neighboring playful energy)
- TNH Agent — Pure Thich Nhat Hanh (neighboring Buddhist tradition)
License: MIT
- Downloads last month
- 5
8-bit
# !pip install llama-cpp-python from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="Sathman/Watts-Agent-GGUF", filename="Watts_Agent-Q8_0.gguf", )