Tiranyx/sidix-lora

SIDIX LoRA — Free & Open Source AI Agent

SIDIX adalah AI agent open source yang berjalan 100% lokal — tidak ada biaya per-query, tidak ada data yang dikirim ke server eksternal.

Model ini adalah LoRA adapter yang di-fine-tune di atas Qwen2.5-7B-Instruct menggunakan QLoRA (4-bit quantization, Kaggle T4 GPU).

🔗 GitHub: github.com/fahmiwol/sidix
🌐 Live Demo (Free): app.sidixlab.com
📄 License: MIT

---

✨ Apa yang Membedakan SIDIX?

Fitur	SIDIX	ChatGPT / Gemini
Biaya inference	Gratis	Per-token billing
Data user	Tetap di server Anda	Dikirim ke cloud mereka
Open source	Ya, MIT	Tidak
Self-hosting	Ya	Tidak
Fallback ke LLM lain	Tidak pernah	N/A
Epistemik labeling	[FACT]/[OPINION]/[UNKNOWN]	Tidak ada

---

🏗️ Arsitektur

SIDIX dibangun di atas IHOS (Islamic Holistic Ontological System) — framework epistemologi yang memetakan konsep keilmuan Islam klasik ke arsitektur AI modern:

• Sanad → Citation chain di setiap output ([FACT] / [OPINION] / [UNKNOWN])
• Muhasabah → Self-refinement loop (Niyah → Amal → Muhasabah, CQF ≥ 7.0)
• Maqashid → 5 objective filter gates (kehidupan, akal, iman, keturunan, harta)
• Ijtihad → ReAct agentic reasoning loop (35 active tools)

---

🚀 Quick Usage

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
import torch

# Load base model with 4-bit quantization
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_compute_dtype=torch.float16,
    bnb_4bit_quant_type="nf4",
)

base_model = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen2.5-7B-Instruct",
    quantization_config=bnb_config,
    device_map="auto",
)

# Load SIDIX LoRA adapter
model = PeftModel.from_pretrained(base_model, "sidixlab/sidix-lora")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-7B-Instruct")

# Inference
messages = [
    {"role": "system", "content": "Kamu adalah SIDIX, AI agent yang jujur dan bersumber."},
    {"role": "user", "content": "Jelaskan konsep sanad dalam konteks AI."},
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer([text], return_tensors="pt").to(model.device)

with torch.no_grad():
    output = model.generate(**inputs, max_new_tokens=512, temperature=0.7)

print(tokenizer.decode(output[0][inputs.input_ids.shape[1]:], skip_special_tokens=True))

Via Ollama (Recommended for Self-Hosting)

# Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh

# Run SIDIX
ollama run sidixlab/sidix-lora

---

📊 Training Details

Parameter	Value
Base model	Qwen/Qwen2.5-7B-Instruct
Method	QLoRA (4-bit NF4)
LoRA rank	64
LoRA alpha	128
Target modules	q_proj, v_proj, k_proj, o_proj
Training GPU	Kaggle T4 (15GB)
Training data	SIDIX corpus (~1,182 documents, Indonesian/English/Arabic)
Domains	Islamic epistemology, creative writing, coding, brand strategy, content planning
Languages	Indonesian (primary), English, Arabic

---

🎯 5 Personas

SIDIX menyesuaikan gaya, kedalaman, dan framing berdasarkan konteks:

Persona	Spesialisasi
MIGHAN	Sintesis riset, long-form, epistemologi Islam
TOARD	Data, logika, analisis terstruktur, code review
FACH	Technical deep-dive, system design, implementasi
HAYFAR	Pengajaran, kurikulum, penjelasan ramah pemula
INAN	Tugas harian, kreatif, percakapan umum

---

🛠️ 35 Active Tools

SIDIX bukan sekadar chatbot — ia adalah agent dengan 35 tools aktif:

Knowledge: search_corpus · read_chunk · concept_graph
Web:       web_fetch · web_search · pdf_extract  
Code:      code_sandbox · code_analyze · code_validate · project_map
Creative:  generate_copy · brand_kit · plan_campaign · generate_ads
Image:     text_to_image (SDXL self-hosted)
Meta:      self_inspect · orchestration_plan · muhasabah_refine
Growth:    prompt_optimizer · roadmap_* · workspace_*

---

🔒 Privacy & Security

• ✅ Zero data ke server eksternal — semua inference lokal
• ✅ No vendor API key required — tidak ada Groq, Gemini, OpenAI
• ✅ 4-label epistemic tagging — halusinasi dilabel, tidak disembunyikan
• ✅ MIT License — bebas digunakan, dimodifikasi, didistribusikan

---

🤝 Contribute

• Telegram Bot: @sidixlab_bot — kirim pesan → masuk corpus queue
• GitHub PR: tambah research note ke brain/public/research_notes/
• Code: fork github.com/fahmiwol/sidix

---

📜 Citation

@software{sidix2026,
  title={SIDIX: Free & Open Source AI Agent Built on Islamic Epistemology},
  author={Mighan Lab},
  year={2026},
  url={https://github.com/tiranyx/sidix},
  license={MIT}
}

---

Built by Tiranyx · sidixlab.com · "We don't build AI that replaces human judgment. We build AI that makes human judgment more informed."