Instructions to use sadiqoon/sadiq-cite-8b-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use sadiqoon/sadiq-cite-8b-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M

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 sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M

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 sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M

Use Docker

docker model run hf.co/sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use sadiqoon/sadiq-cite-8b-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "sadiqoon/sadiq-cite-8b-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": "sadiqoon/sadiq-cite-8b-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M

Ollama
How to use sadiqoon/sadiq-cite-8b-GGUF with Ollama:
```
ollama run hf.co/sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M
```
Unsloth Desktop

How to use sadiqoon/sadiq-cite-8b-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use sadiqoon/sadiq-cite-8b-GGUF with Docker Model Runner:
```
docker model run hf.co/sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M
```

Lemonade

How to use sadiqoon/sadiq-cite-8b-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.sadiq-cite-8b-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use sadiqoon/sadiq-cite-8b-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M

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 sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use sadiqoon/sadiq-cite-8b-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "sadiqoon/sadiq-cite-8b-GGUF:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Sadiq-Cite 8B — GGUF

Sadiq-Cite is a retrieval-grounded citation model for Islamic jurisprudence in the Ja'fari (Imami) tradition. Given a question and a set of retrieved passages, it finds the rulings that bear on the question, quotes them verbatim between «…», and lists their book and page. It does not issue verdicts, does not reason its way to a ruling, and does not apply the law to the asker's situation — and when the passages contain nothing relevant, it says so instead of improvising.

It is the answer stage of a pipeline, not a chatbot: run it behind LM Studio's server, inside AnythingLLM, in an n8n flow, or from any OpenAI-compatible client, with your own corpus supplying the passages. Built by Sadiqoon Technologies on Qwen/Qwen3-8B.

نموذج استشهاد مؤسَّس على الاسترجاع في الفقه الإماميّ. يُعطى السؤال والمقاطع المسترجعة، فيلتقط النصوص التي تخصّ المسألة وينقلها حرفيًّا بين «» ويعزوها بالكتاب والصفحة. لا يُفتي ولا يستنبط ولا ينزّل الحكم على حال السائل، وإذا لم يكن في المقاطع ما يخصّ السؤال صرّح بذلك ولم يخترع. يُشغَّل خلف خادم محلّي داخل منظومة استرجاع، لا كواجهة محادثة.

Why a citation model rather than an answering one

A jurisprudential answer has two halves: what the sources say, and what follows for you. The first is verifiable — the quote is either in the passage or it is not. The second is where a language model, however large, quietly invents. Sadiq-Cite is trained to do only the first half, so that everything it produces can be checked by a machine.

That check ships with it: sadiq_fiqh.verify() matches every «quote» in the output against the passages you supplied and returns a per-quote verdict, so your pipeline can drop an unverified quote before a reader ever sees it.

Evaluation

282 held-out questions with their retrieved passages, plus 100 refusal probes — real questions paired with passages deliberately drawn from unrelated chapters. All metrics are computed programmatically; no LLM judge was used.


Answers that quote at least one passage	99.6%
Quotes traceable to the supplied passages	87.5% (84.5% exact after orthographic normalization, 3.0% near-verbatim)
Quotes not found in the passages	10.8%
Answers where every quote verified	79.4%
Answers giving a book-and-page citation	97.2%
Citations matching a supplied passage	82.5%
Correct refusals on the 100 probes	100%
False refusals on genuine questions	0.4% (1 of 282)

Two things are worth reading carefully. Refusal discrimination is exact: on a hundred questions given deliberately irrelevant context it refused every time, and on 282 answerable ones it refused once. Groundedness, on the other hand, is high but not perfect — roughly one quote in nine is reconstructed from what the model learned rather than copied from what it was given. That is the reason the verifier is part of the release rather than an afterthought: run it, and the residual becomes a rejected quote instead of a false citation.

Robustness to prompt format

Different RAG tools inject context differently, so the same 60 questions were re-run in four injection styles:

Injection style	Answers with quotes	Quote verbatim rate	Citation matches a passage
Native (the shape below)	100%	88.8%	67.8%
AnythingLLM-style context block	98.3%	85.5%	79.7%
XML `<doc source= page=>` tags	100%	87.7%	66.7%
Bare passages, no book or page given	100%	83.0%	0%

Quoting holds up across all four — the model does not need our exact template to find and copy the right text. Citation is a different matter: if you do not pass the book and page with each passage, the model will produce a citation anyway, and it will be fabricated. Always include source metadata in the passage header, or strip the citations from the output.

Training

3,949 question–passage–answer examples in which the answer quotes and cites only, plus 500 refusal examples. LoRA (r=32) on all attention and MLP projections, two epochs, merged into the base weights.

Files

File	Quant	Size	Runs on
`sadiq-cite-8b-Q4_K_M.gguf`	Q4_K_M	~4.7 GB	Recommended. A laptop with 8 GB of RAM
`sadiq-cite-8b-Q5_K_M.gguf`	Q5_K_M	~5.7 GB	16 GB of RAM; a little closer to full precision
`sadiq-cite-8b-Q8_0.gguf`	Q8_0	~8.7 GB	Effectively lossless
`sadiq-cite-8b-F16.gguf`	F16	~16 GB	Reference

The original weights are in sadiqoon/sadiq-cite-8b.

Running it

LM Studio as a server

Load the model, start the local server, then point any OpenAI-compatible client at it:

from sadiq_fiqh import Sadiq, Passage, verify        # pip install sadiq-fiqh
s = Sadiq(base_url="http://localhost:1234/v1", model="sadiq-cite-8b")

passages = [Passage(text="لا يجب الخمس في ما لا يدخل في مسمّى التكسّب…", book="…", page=120)]
answer = s.cite("هل يجب الخمس في الذهب الملبوس؟", passages)
print(verify(answer, passages))   # {'quotes': [...], 'all_verbatim': True}

AnythingLLM

Point the workspace's LLM at your local server (LM Studio or Ollama) and select this model; use Fiqh-Embed as the embedder — AnythingLLM supports LM Studio and Ollama embedders natively, so the whole stack stays local. Paste the system prompt below into the workspace's prompt setting.

Ollama

ollama pull sadiqoon/sadiq-cite

Or build it yourself from the GGUF in this repo: ollama create sadiq-cite -f Modelfile.

Ollama applies Qwen3 thinking-mode parsing by default, which routes the answer into the thinking field and leaves content empty. This model does not reason before answering, so turn it off in the request:

curl http://localhost:11434/api/chat -d '{
  "model": "sadiqoon/sadiq-cite",
  "think": false,
  "stream": false,
  "messages": [{"role": "user", "content": "..."}]
}'

With "think": false the citation block arrives in content as expected.

n8n and custom pipelines

Retrieve with Fiqh-Embed, re-rank with Fiqh-Rerank, then call this model. sadiq_fiqh.format_prompt() builds the message in the shape the model expects; verify() gates the result.

System prompt

أنت «صادق»، مساعد بحث في فقه أهل البيت (عليهم السلام). مهمتك أن تعثر في المقاطع المرفقة على النصوص الشرعية التي تخصّ سؤال السائل وتنقلها حرفيًّا بين علامتي «» مرقّمةً، ثم تسرد مصادرها بالكتاب ورقم الصفحة. لا تفتِ ولا تستنبط ولا تطبّق الحكم على حالة السائل. إن لم تجد في المقاطع ما يخصّ السؤال فقل ذلك صراحةً، وإن كان السؤال ناقصًا فاطلب التوضيح.

Expected input shape

السؤال:
<the question>

المقاطع المسترجعة:
[1] (<book>، ص <page>)
<passage text>

[2] (<book>، ص <page>)
<passage text>

Greedy decoding (temperature=0) is recommended.

The rest of the stack

Fiqh-Embed · Fiqh-Rerank · Fiqh-Classifier · Fiqh-Glossary Linker

Citation

@misc{sadiqoon2026sadiqcite,
  title  = {Sadiq-Cite: Retrieval-Grounded Quotation and Citation for Imami Jurisprudence},
  author = {Sadiqoon Technologies},
  year   = {2026},
  url    = {https://huggingface.co/sadiqoon/sadiq-cite-8b}
}

License & Contact

MIT (the base model is Qwen3-8B, Apache 2.0). Built and maintained by Sadiqoon Technologies Ltd, London. Questions and collaboration: info@sadiqoon.uk

Downloads last month: 76

GGUF

Model size

8B params

Architecture

qwen3

Hardware compatibility

4-bit

5-bit

8-bit

16-bit

Model tree for sadiqoon/sadiq-cite-8b-GGUF

Base model

Qwen/Qwen3-8B-Base

Finetuned

Qwen/Qwen3-8B

Finetuned

sadiqoon/sadiq-cite-8b

Quantized

(1)

this model