Daoism-Qwen3.5-9B-GGUF

Daoism-Qwen3.5-9B-GGUF

GGUF quantized version of lius-cc/Daoism-Qwen3.5-9B — the world's first open-source Taoist LLM.

Released by 鼎稔道學館 Dingren Daoxue Lab (2026-05-16), Apache 2.0.

---

🏛️ 關於釋出者

劉厝派鼎新門不是 AI 開發公司，而是一間道教教育機構。 由台灣正一道道長創辦，本業為道教知識的整理、傳承與教學。

「我們是道教教育機構，不是 AI 開發人員。但我們把最完整的能力與資料，交給開發人員。」

完整身份說明見 fp16 model card。

---

📁 Available files

---

🚀 Quick Start

Ollama (one-liner)

ollama pull hf.co/lius-cc/Daoism-Qwen3.5-9B-GGUF:Q4_K_M
ollama run hf.co/lius-cc/Daoism-Qwen3.5-9B-GGUF:Q4_K_M

LM Studio (GUI, no code)

1. Install LM Studio
2. Search lius-cc/Daoism-Qwen3.5-9B-GGUF
3. Pick Q5_K_M for best quality
4. Load → chat

llama.cpp (CLI)

./llama-cli -hf lius-cc/Daoism-Qwen3.5-9B-GGUF \
  -p "你是道教智慧 AI。問：補財庫的科儀流程是什麼？答：" \
  -n 1024 --temp 0.6

llama-cpp-python (Python, OpenAI 相容 API)

CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python[server]

from llama_cpp import Llama

llm = Llama.from_pretrained(
    repo_id="lius-cc/Daoism-Qwen3.5-9B-GGUF",
    filename="daoism-qwen3-5-9b.Q5_K_M.gguf",
    n_ctx=4096,
    n_gpu_layers=-1,  # full GPU offload
)
out = llm.create_chat_completion(
    messages=[
        {"role": "system", "content": "你是道教智慧 AI。"},
        {"role": "user", "content": "擲筊三聖筊代表什麼？"},
    ],
    max_tokens=1024,
    temperature=0.6,
)
print(out["choices"][0]["message"]["content"])

Serve as OpenAI-compatible API

python -m llama_cpp.server \
  --model daoism-qwen3-5-9b.Q5_K_M.gguf \
  --n_gpu_layers -1 --n_ctx 4096 \
  --host 0.0.0.0 --port 8000

curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"daoism","messages":[{"role":"user","content":"你好"}]}'

---

💻 System Requirements

Apple Silicon (M-series): M2/M3 16 GB+ runs Q4_K_M smoothly via LM Studio or Ollama.

---

🔍 RAG Integration (recommended)

For best depth, combine with our companion dataset or public RAG API:

import requests
from llama_cpp import Llama

llm = Llama.from_pretrained(
    repo_id="lius-cc/Daoism-Qwen3.5-9B-GGUF",
    filename="daoism-qwen3-5-9b.Q5_K_M.gguf",
    n_ctx=8192, n_gpu_layers=-1,
)

def ask(question):
    # RAG retrieve from lius.cc
    hits = requests.post(
        "https://lius.cc/api/llm-rag",
        json={"q": question, "n": 5},
        timeout=10,
    ).json().get("hits", [])
    ctx = "\n\n---\n\n".join(
        f"### {h['name']}\n{h['summary']}\n\n{h['content'][:1500]}"
        for h in hits
    )
    out = llm.create_chat_completion(
        messages=[
            {"role": "system", "content": f"你是道教 AI。請以下列館藏資料為基礎深度回答：\n\n{ctx}"},
            {"role": "user", "content": question},
        ],
        max_tokens=4096,
        temperature=0.6,
    )
    return out["choices"][0]["message"]["content"]

print(ask("城隍信仰與道教有什麼關係？"))

---

📜 License

Apache 2.0 — 商用可，修改可，再發佈可。請保留原始作者標示。

---

📚 Related Resources

• 🤖 fp16 source model: lius-cc/Daoism-Qwen3.5-9B
• 📚 Companion dataset: lius-cc/daoism-knowledge-rag (95,919 entries)
• 🔌 Public RAG API: POST https://lius.cc/api/llm-rag
• 📖 Deployment Guide: lius.cc/llm

---

📜 Citation

@misc{daoism-qwen3-9b-2026,
  author = {Liu, Chi-Ying and Dingren Daoxue Lab},
  title  = {Daoism-Qwen3.5-9B: An Open-Source Taoist Knowledge Language Model},
  year   = {2026},
  month  = {May},
  publisher = {HuggingFace},
  url    = {https://huggingface.co/lius-cc/Daoism-Qwen3.5-9B}
}

---

🙏 Acknowledgements

• Quantization: llama.cpp by Georgi Gerganov et al.
• Base model: Alibaba Qwen team
• Compute: vast.ai community spot GPUs

---

「真正的開源精神，是把完整的能力交到開發者手中。」 — 鼎稔道學館 · 劉啟穎