peer-ai-reviewer

Peer-AI Code Reviewer

AI-powered code security reviewer fine-tuned on security vulnerability patterns.

Model Description

Peer-AI is a Qwen2.5-Coder-1.5B model fine-tuned with LoRA for code security review tasks. It can identify common security vulnerabilities including:

• CWE-78: Command Injection
• CWE-89: SQL Injection
• CWE-22: Path Traversal
• CWE-79: Cross-Site Scripting (XSS)
• CWE-94: Code Injection
• CWE-120: Buffer Overflow
• CWE-190: Integer Overflow
• CWE-416: Use After Free
• CWE-502: Deserialization
• CWE-798: Hardcoded Credentials

Languages Supported

• Python
• Go
• C/C++
• Rust
• JavaScript/TypeScript

Usage

With Ollama

# Create model
ollama create peer-ai -f Modelfile

# Review code
curl -s http://localhost:11434/api/generate -d '{
  "model": "peer-ai",
  "prompt": "Review the following python code...",
  "stream": false
}'

With CLI

pip install peer-ai
git diff | peer-ai review -
peer-ai review src/main.py

Training

• Base Model: Qwen/Qwen2.5-Coder-1.5B-Instruct
• Method: QLoRA (r=16, alpha=32)
• Dataset: 223 security vulnerability examples
• Epochs: 3
• Final Loss: 0.21

Output Format

{
  "line": 2,
  "severity": "high",
  "category": "security",
  "rule": "CWE-78",
  "title": "Command injection vulnerability",
  "message": "User input passed to shell command.",
  "suggestion": "Use subprocess with shell=False."
}

License

Apache 2.0

Links

• GitHub
• Training Code