geminikit

A comprehensive Gemini CLI workspace kit configured as a Bun monorepo. This package provides a pre-configured environment with advanced AI skills and tools, ready to be dropped into any project.

Features

• Skill System: Modular skills for specialized tasks.
  • Chrome DevTools: Browser automation, performance analysis, and debugging via Puppeteer.
  • Code Review: Protocols for rigorous technical feedback reception, review requests, and verification gates.
  • Debugging: Systematic framework (Investigation, Pattern Analysis, Hypothesis, Implementation) with root cause tracing.
  • Directus Manager: Manage Directus content, schema, files, and flows via the Directus Model Context Protocol (MCP).
  • Frontend Design: Aesthetic direction, strict frontend implementation rules, and high-quality image sourcing.
  • Media Processing: FFmpeg and ImageMagick wrappers for optimized video/audio conversion and batch image processing.
  • Planning: Structured impact analysis, architectural design, and detailed implementation planning.
  • Presentation: Professional Slidev presentation generation using a "Think-Plan-Act" workflow.
  • Problem Solving: Strategic techniques (Simplification Cascades, Inversion, etc.) to overcome blocks.
  • Recording Analysis: Adaptive video/screen recording analysis using OmniParser (GPU) or PaddleOCR (CPU).
  • Research: Deep technical research, documentation analysis, and report synthesis.
  • Sequential Thinking: Structured, reflective thought process for decomposing complex problems.
• Bun Monorepo: Optimized for speed and modern JavaScript development.
• TypeScript: Full type safety across all scripts and tools.
• GeminiKit CLI (gk): Dedicated CLI to manage the environment and logs.

Prerequisites & GPU Support (CUDA)

The recording-analysis skill and certain Python-based tools require specific setup for GPU acceleration.

CUDA Setup Guide

It appears you have an NVIDIA GPU, but PyTorch cannot access it. This usually means the CUDA Toolkit is missing or mismatched.

Quick Fixes

Windows

1. Install the latest NVIDIA Drivers.
2. Install PyTorch with CUDA support:

   pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

Linux

1. Install NVIDIA drivers and CUDA toolkit:

   sudo apt install nvidia-cuda-toolkit

2. Reinstall PyTorch (see above).

Installation

To use this kit in your project:

bun add geminikit
# or
npm install geminikit
# or
pnpm add geminikit

After installation, run the setup command to configure the .gemini environment:

npx gk setup

Directus MCP Setup

To use the Directus Manager skill, you must configure the MCP server in your .gemini/settings.json file and provide authentication credentials.

1. Update .gemini/settings.json

Add the following configuration to the mcpServers object in your .gemini/settings.json:

"directus": {
  "command": "npx",
  "args": [
    "-y",
    "dotenv-cli",
    "--",
    "npx",
    "@directus/content-mcp@latest"
  ],
  "env": {
    "DIRECTUS_URL": "http://your-directus-url:8055"
  }
}

2. Authentication

The Directus MCP server requires authentication. You must set ONE of the following sets of environment variables in your system or project context (e.g., via .env if using dotenv-cli or system variables):

• Option A (Recommended): User Credentials

  • DIRECTUS_USER_EMAIL: Your Directus user email.
  • DIRECTUS_USER_PASSWORD: Your Directus user password.

• Option B: Access Token

  • DIRECTUS_TOKEN: A static access token for a Directus user.

Ensure DIRECTUS_URL is also set in the env section of the config or as an environment variable.

Manual Installation & Troubleshooting

Manual Installation

If npx gk setup encounters issues, you might need to install Bun.sh and the Gemini CLI manually.

Install Bun.sh

Follow the official Bun.sh installation guide for your operating system:

• Windows:

  powershell -c "irm bun.sh/install.ps1 | iex"

  Alternatively, you can use npm:

  npm install -g bun
  # or
  pnpm install -g bun

• macOS / Linux:

  curl -fsSL https://bun.sh/install | bash

After installation, verify bun is installed by running:

bun --version

Install Gemini CLI

npm install -g @google/generative-ai-cli
# or
pnpm install -g @google/generative-ai-cli

Verify gemini is installed by running:

gemini --version

Troubleshooting: Telemetry .gemini folder

The npx gk setup command is designed to create the node_modules/.gemini folder, which is crucial for telemetry and skill management. If, for any reason, this folder is not created automatically, you can create it manually:

mkdir -p node_modules/.gemini

(Note: mkdir -p works on Linux/macOS. For Windows, mkdir node_modules\.gemini or New-Item -ItemType Directory -Force -Path "node_modules/.gemini")

Troubleshooting: pnpm and gk command

When using pnpm, you might encounter issues where the gk command is not automatically registered in your path, or its peer dependencies are not correctly linked. This is often due to pnpm's strict linking model.

To ensure gk works as expected, add the following lines to your project's .npmrc file (create it if it doesn't exist):

shamefully-hoist=true
auto-install-peers=true

GeminiKit CLI (gk)

The package includes a dedicated CLI tool gk for managing the GeminiKit environment.

# Setup the environment (install Bun, Gemini CLI, and configure .gemini)
npx gk setup

# Check the health of the installation
npx gk doctor

# View and manage telemetry logs
npx gk log

# Show version information
npx gk --version

# Show help
npx gk --help

Usage

Gemini CLI Skills

Once installed and setup, you can use the Gemini CLI with the provided skills.

# Start a planning session
gemini plan "Implement a new authentication system"

# Perform deep research
gemini research "Current state of quantum computing"

# Automate a browser task
gemini chrome "Take a screenshot of google.com"

# Process media (video/audio/images)
gemini media "Optimize this video for web"

# Analyze screen recordings or videos
gemini use-mcp "Analyze this screen recording for UI elements"

# Solve a blocking problem
gemini solve "I'm stuck on the database schema design"

# Review code or get feedback
gemini review "Please review my recent changes in src/auth"

# Generate a presentation
gemini use-mcp "Create a slide deck about Node.js performance"

Direct Script Usage

You can also run the underlying scripts directly using bun:

# Run the sequential thinking processor
bun .gemini/skills/sequential-thinking/scripts/process-thought.ts --thought "Initial analysis" --number 1 --total 5

# Run a browser automation script
bun .gemini/skills/chrome-devtools/scripts/navigate.ts --url https://example.com

# Run media conversion
bun .gemini/skills/media-processing/scripts/convert.ts --input video.mkv --preset web

Development

To contribute or modify this kit:

1. Clone the repository.
2. Run bun install to install dependencies.
3. Make changes to skills in .gemini/skills.
4. Run bun tsc to verify type safety.

License

ISC