LMN
LMN
Back to projects
Python

lineup

Multi-agent AI code review CLI powered by Google Gemini and ADK.

PythonGoogle GeminiADKMulti-AgentRich
Source Code

Lineup

Lineup is a multi-agent AI code review CLI powered by Google Gemini and the Agent Development Kit (ADK). It orchestrates a team of specialized AI agents (Bug Detector, Security, Performance, and Style Reviewers) to deeply analyze your codebase or uncommitted changes and provide actionable, high-confidence feedback.

Features

  • Multi-Agent Orchestration: Specialized agents handle different aspects of a code review (Security, Bugs, Performance, Style) independently and report back to a central orchestrator.
  • Git Diff & Local File Support: Review an entire file/directory or just your uncommitted git diff.
  • Rich Terminal UI: Beautiful, interactive terminal output with live markdown updates and styled review tables.
  • Markdown Export: Export the full AI thought process and review findings into a markdown report.
  • Robust Retry Logic: Built-in exponential backoff and retry UI for handling API rate limits and temporary outages.

Prerequisites

  • Python 3.10+
  • uv (recommended) for package management
  • A Google Gemini API key

Installation

Clone the repository and install the dependencies:

git clone https://github.com/LikhinMN/lineup.git
cd lineup
uv sync

Configuration

Copy the .env.example file to .env:

cp .env.example .env

Add your Google Gemini API key to the .env file:

GEMINI_API_KEY=your_gemini_api_key_here

Usage

You can run the CLI using uv run.

Review a specific file or directory

uv run cli.py review path/to/your/code

Review uncommitted git changes

uv run cli.py review --diff

Export the report to a file

uv run cli.py review src/ --export report.md

Architecture

The system uses a hierarchical agent architecture defined via google.adk:

  1. Orchestrator Agent: Manages the review workflow and delegates analysis.
  2. Specialized Agents:
    • Bug Detector: Looks for logic errors, race conditions, memory leaks, etc.
    • Security: Identifies vulnerabilities like SQLi, XSS, hardcoded secrets.
    • Performance: Finds inefficient algorithms, blocking I/O, N+1 queries.
    • Style: Ensures code readability, solid principles, and maintainability.
  3. Reader Agent: A utility agent for reading local files safely.

License

MIT