A Secure Code Review Collaborator

DreamtestAI scans your codebase with modern LLMs and generates vulnerability reports in Markdown, JSON, or CSV. Built for security analysts, auditors, and developers who want automated secure code review.

Source: github.com/promptagainstthemachine/DreamtestAI

Features

Supports OpenAI, Claude, Gemini, and Ollama-compatible local models
Parses files in a directory and sends code with line numbers to the model
Markdown-styled findings in the terminal
Progress bar during analysis
Structured output to JSON and CSV
Highlights file names, line numbers, and vulnerable lines
Simple CLI with rich formatting

Installation

git clone https://github.com/promptagainstthemachine/DreamtestAI.git
cd DreamtestAI
pip install -r requirements.txt

Usage

python main.py --folder ./your/codebase

Works well with local models such as qwen2.5-coder:latest via Ollama.

Optional arguments

Flag	Description
`--model`	Model name (e.g. `openai/gpt-4`, `claude-3-opus`, `gemini-1.5-pro`)
`--folder`	Path to the source folder to analyze
`--output-json`	Save results to `vulnerabilities.json`
`--output-csv`	Save results to `vulnerabilities.csv`
`--debug`	Enable debug logging

Examples

# Local / default model
python main.py --folder ./src

# OpenAI with exports
python main.py --folder ./src --model openai/gpt-4 --output-json --output-csv

Output formats

Terminal (Markdown)

Findings print in a readable Markdown format in the terminal.

JSON

[
  {
    "file": "login.js",
    "issue": "Insecure Redirect",
    "line": "88",
    "code": "window.location.href = userInput;"
  }
]

CSV

File Name	Issue	Line Number	Line Text
login.js	Insecure Redirect	88	window.location.href = userInput;

Supported models

Provider	Examples
Local / Ollama	`qwen2.5-coder:latest`, `llama3`, `mistral`
OpenAI	`gpt-4`, `gpt-3.5-turbo`
Claude	`claude-3-opus`, `claude-2.1`
Gemini	`gemini-1.5-pro`

License

MIT — see the repository LICENSE file.