AI Code Review Model - Javascript

This is a fine-tuned code review model specialized for Javascript code analysis.

Model Details

Base Model: Qwen/Qwen2.5-Coder-1.5B-Instruct
Training Method: LoRA fine-tuning with MLX
Format: GGUF (Q4_K_M quantization)
Target Language: Javascript
Purpose: Automated code review for CI/CD pipelines

Usage

Docker (Recommended)

docker pull ghcr.io/iq2i/ai-code-review:javascript-latest
docker run --rm -v $(pwd):/workspace ghcr.io/iq2i/ai-code-review:javascript-latest /workspace/src

llama.cpp

# Download the model
wget https://huggingface.co/loicsapone/ai-code-review-javascript/resolve/main/model-Q4_K_M.gguf

# Run inference
./llama-cli -m model-Q4_K_M.gguf -p "Review this code: ..."

Python (llama-cpp-python)

from llama_cpp import Llama

llm = Llama(model_path="model-Q4_K_M.gguf")
output = llm("Review this code: ...", max_tokens=512)
print(output)

Output Format

The model outputs JSON structured code reviews:

{
  "summary": "Brief overview of code quality",
  "score": 8,
  "issues": [
    {
      "type": "bug",
      "severity": "medium",
      "line": 42,
      "description": "Potential null pointer",
      "suggestion": "Add null check"
    }
  ],
  "positive_points": [
    "Good error handling",
    "Clear variable names"
  ]
}

Training

This model was trained on curated Javascript code review examples using:

MLX framework for Apple Silicon acceleration
LoRA adapters (r=8, alpha=16)
Custom dataset of real-world code issues

For training details, see the GitHub repository.

Limitations

Optimized for Javascript syntax and best practices
May not catch all edge cases or security vulnerabilities
Should be used as a supplementary tool, not a replacement for human review

License

Apache 2.0

Citation

@software{ai_code_review_javascript,
  title = {AI Code Review Model for Javascript},
  author = {IQ2i Team},
  year = {2025},
  url = {https://github.com/iq2i/ai-code-review}
}