enver / ayncoding-gemma2-2b

🏛️ AynCoding-Gemma2: Sovereign Epistemic Classical Arabic Logic (Manṭiq) 2.6B

AynCoding-Gemma2 ( ayncoding-gemma2-2b ) is a sovereign, epistemic code synthesis and architectural refactoring model based on Google's flagship Gemma 2 2B architecture ( google/gemma-2-2b-it ), aligned and conditioned with Classical Arabic Logic ( Manṭiq ) and Morphological Root Lexicography ( Ishtiqāq ) .

Built upon Google DeepMind's distilled Gemma 2 architecture, AynCoding-Gemma2 marries cutting-edge model distillation with 1,200 years of classical Arabic epistemological rigor. It natively engages a structured Chain-of-Thought reasoning block ( <ayn_mantiq> ... </ayn_mantiq> ) prior to code generation, analyzing software specifications through real essential definitions ( Al-Ḥadd bi al-Dhātiyyāt ), eliminating logical fallacies ( Dafʿ al-Dawr , Dafʿ al-Tasalsul , ʿAdam al-Tanāquḍ ), and mapping domain concepts to authentic tri-consonantal roots.

🌟 The 5 Classical Arabic Epistemic Pillars

AynCoding-Gemma2 is strictly governed by the foundational canons of classical Arabic scholarship:

1. Abū Ḥāmid al-Ghazālī ( Miʿyār al-ʿIlm fī Fann al-Manṭiq & Miḥakk al-Naẓar ):
   • Al-Ḥadd bi al-Dhātiyyāt (Real Definition by Essence) : Entities are defined by their constitutive essential attributes, never accidental symptoms.
   • Dafʿ al-Dawr (Elimination of Circularity) : Zero circular dependencies, cyclic callbacks, or self-referential ungrounded types.
   • Dafʿ al-Tasalsul (Elimination of Infinite Regress) : Strictly bounded recursion, guaranteed loop termination, and timeout bounds.
   • ʿAdam al-Tanāquḍ (Law of Non-Contradiction) : Mutually exclusive states are unrepresentable simultaneously.
2. Al-Khalīl ibn Aḥmad al-Farāhīdī ( Kitāb al-ʿAyn ) & Ibn Manẓūr ( Lisān al-ʿArab ):
   • Tri-Consonantal Root Decomposition & Combinatorial Safety : Complex systems are decomposed into orthogonal, irreducible primitives anchored in authentic semantic roots ( قفل , حفظ , نقل , عقد , حسب , رتب , سلم , حكم ).
   • Exhaustive State-Space Coverage : Zero unhandled match arms and zero silent exception swallowing ( except Exception: pass is prohibited).
3. Al-Rāghib al-Iṣfahānī ( Al-Mufradāt fī Gharīb al-Qurʾān ):
   • Ontological Domain Modeling & Teleology ( Ghāyah ) : Absolute ban on amorphous, vague identifiers ( data , temp , val , mgr , helper ). Every symbol reflects its distinct teleological purpose.
4. Al-Zamakhsharī ( Asās al-Balāghah ):
   • Abstraction Integrity & Rhetorical Eloquence ( Ḥaqīqah vs Majāz ) : Delineating physical machine reality (memory, CPU caches, IO) from software abstractions. Zero leaky abstractions.
5. Sībawayh ( Al-Kitāb ):
   • Syntactic Governance ( Al-ʿĀmil wa al-Maʿmūl ) : Strict caller-callee hierarchy, rigorous static typing, and guaranteed AST integrity.

🧠 Cognitive Chain-of-Thought Protocol ( <ayn_mantiq> )

Before emitting production code, AynCoding-Gemma2 executes its epistemic reasoning:

<ayn_mantiq>
🏛️ AYN-ENGINE EPISTEMIC LOGIC & MORPHOLOGY REASONING:
- Classical Root & Morphology (الجذر والتصريف): [Root e.g. ق-ف-ل, linguistic significance]
- Real Definition & Essence (الحد بالذاتيات - معيار العلم للغزالي): [Essential attributes and invariants]
- Epistemic Fallacy Invariants (دفع الدور والتسلسل ونفي التناقض): [Circularity, regress, and contradiction guards]
- Lexicographical Teleology (الغاية وبلاغة التجريد - المفردات والأساس): [Pure purpose, zero vague abstractions]
- Syntactic Governance (العامل والمعمول - كتاب سيبويه): [Strict type contracts, caller-callee hierarchy]
</ayn_mantiq>

🧪 Benchmark & Dynamic Evaluation

In live benchmark evaluations using the 5-Pillar Static Epistemic Auditor :

• Sliding Window Rate Limiter:
  • Overall Epistemic Score: 87.8% (Grade: B+)
  • AST Syntax Integrity: 100% Valid ✅
  • Banned Placeholders: 0 (Zero-Loss Complete) ✅
  • Asās al-Balāghah (Eloquence): 10.0 / 10
  • Kitāb al-ʿAyn (Decomposition): 10.0 / 10
  • Sībawayh (Governance): 10.0 / 10
  • Dynamic Execution: Automatically evicted oldest items in $O(1)$ under real multi-threaded execution!

🚀 Quickstart: Running with Ollama

1. Download the GGUF model and Modelfile:

   huggingface-cli download enver/ayncoding-gemma2-2b ayncoding-gemma2-2b.gguf Modelfile --local-dir ./ayncoding-gemma2
   cd ayncoding-gemma2
   

2. Register with Ollama:

   ollama create ayncoding-gemma2 -f Modelfile
   

3. Run Inference:

   ollama run ayncoding-gemma2 "Write a thread-safe sliding window rate limiter in Python"
   

💻 Python ( llama-cpp-python ) Usage

from llama_cpp import Llama

llm = Llama(
    model_path="ayncoding-gemma2-2b.gguf",
    n_ctx=8192,
    n_threads=16  # NUMA optimization
)

output = llm.create_chat_completion(
    messages=[
        {"role": "user", "content": "Explain how Ghazalian logic eliminates circularity in software architecture."}
    ],
    temperature=0.2,
    max_tokens=4096
)

print(output["choices"][0]["message"]["content"])

🛠️ Included Tools in Repository

• Modelfile : Sovereign Ollama definition with the 5 Classical Arabic Pillars and hyperparameter tuning.
• dataset/ayn_mantiq_epistemic_dataset_50.jsonl : 50 AST-validated training samples across 10 classical domains with <ayn_mantiq> CoT reasoning.
• tools/core/mantiq_engine.py : Classical Logic fallacy detection engine ( Dawr , Tasalsul , Tanāquḍ ).
• tools/core/mantiq_purifier.py : Automated code sanitizer purging amorphous variables and silent bare except: clauses.
• tools/bin/ayncode : Sovereign CLI developer tool for code synthesis, auditing, and purification.

📜 Heritage & Citations

• Google DeepMind, Gemma 2: Improving Open Language Models at a Practical Size .
• Abū Ḥāmid al-Ghazālī, Miʿyār al-ʿIlm fī Fann al-Manṭiq .
• Al-Khalīl ibn Aḥmad al-Farāhīdī, Kitāb al-ʿAyn .
• Al-Rāghib al-Iṣfahānī, Al-Mufradāt fī Gharīb al-Qurʾān .
• Ibn Manẓūr, Lisān al-ʿArab .
• Sībawayh, Al-Kitāb .