Pika Review 🦊

Enterprise-grade AI Architectural Sentinel & Compliance Engine

Pika Review is a high-performance CLI tool designed to perform surgical code reviews using AI. It focuses on deep architectural debt, security vulnerabilities, and project-specific compliance that traditional linters miss.

GitHub: HackX-IN/pika-review

🚀 Key Features

• 🧠 Architecture Rules Engine: Enforce project-specific standards via .pika-rules.md.
• 🦙 Native Local Ollama Support: Run 100% private, free, offline scans using local LLMs (e.g. Qwen, Llama).
• 🎛️ Local Model Selector: Interactively choose and configure local models from the command line.
• 🛡️ Git Commit Safeguard Hook: Prevent pushing high-severity compliance debt and vulnerabilities automatically.
• 📊 Health Dashboard: Track architectural health trends over time with pika-review stats.
• 🎨 Premium Interactive Reports: Immersive, dark-mode HTML reports for deep triage.
• 🔍 Smart Context Scanning: Uses -U10 context windows for higher AI reasoning accuracy.
• 🛡️ "Pika-Ignore" support: Suppress specific lines using // pika-ignore comments.
• ⚡ Parallel Orchestration: Scans multiple files concurrently with p-limit.
• 🌍 Provider Agnostic: Works with OpenAI, Claude, Grok, local Ollama, or any OpenAI-compatible endpoint.

📦 Installation

# Via Bun (Recommended)
bun add -g pika-review

# Via NPM
npm install -g pika-review

🛠️ Setup

1. Initialize your configuration:

   pika-review init

2. Configure your AI provider in ~/.pika-review.yaml:

   ai:
     apiKey: "your-api-token"
     model: "gpt-4o" # or your preferred model
     baseURL: "https://api.openai.com/v1"

🏗️ Enterprise Features

1. Architecture Rules Engine

Create a .pika-rules.md file in your repository root to guide the AI with project-specific context:

# Architectural Rules

- Use Functional Components with Hooks, never Class Components.
- All service calls must go through `src/api/client.ts`.
- Database queries are restricted to the Repository layer.

2. Pika-Ignore

Suppress false positives or acknowledged risks directly in your code:

const secret = "12345"; // pika-ignore (intentional for testing)

3. Health Stats

Monitor your codebase's architectural debt over time:

pika-review stats

🔍 Usage

| Command | Description | | :---------------------- | :----------------------------------------------------------- | | pika-review scan | Scan staged git changes (Default) | | pika-review scan -i | Interactive file selection mode | | pika-review view | Open the latest interactive report | | pika-review stats | View architectural health trends and scan dashboard | | pika-review models | Interactively select and configure local Ollama models | | pika-review hook <act>| Install (install) or uninstall (uninstall) Git safeguards | | pika-review rules -g | Auto-generate architectural .pika-rules.md guidelines | | pika-review scan --ci | Fail pipeline on Critical/High issues |

🦙 Local Ollama & Offline Setup

Pika Review fully supports local, 100% private, offline code reviews via Ollama.

1. Configure Ollama Provider

Initialize your configuration:

pika-review init

Open ~/.pika-review.yaml and configure the Ollama provider:

ai:
  provider: "ollama"
  model: "qwen2.5-coder:7b" # Or your pulled model
  baseURL: "http://localhost:11434/v1"

2. Interactive Local Setup

To avoid editing files manually, you can manage everything directly from the command line:

# Discover local pulled models and switch active model instantly
pika-review models

# Automatically bootstrap customized architectural rules for your tech stack
pika-review rules --generate

# Register git safeguard hooks to run scans automatically before commits
pika-review hook install

🤝 Contributing

We welcome contributions from the community to help make Pika Review the ultimate architectural code reviewer!

Please read our Contributing Guide for details on how to set up local development, run compiler checks, write custom CLI commands, and submit high-quality Pull Requests.

🛡️ Privacy & Security

Pika Review processes your local git diffs and transmits them directly to your configured AI provider via SSL. No code is stored or cached by the Pika Review engine.

📄 License

MIT © Pika Review Contributors