🤖 nGPT, A Swiss army knife for LLMs: A fast, lightweight CLI and interactive chat tool that brings the power of any OpenAI-compatible LLM (OpenAI, Ollama, Groq, Claude, Gemini, etc.) straight to your terminal. rewrite texts or refine code, craft git commit messages, generate and run OS-aware shell commands, seamlessly switching between single-shot prompts and an interactive chat when you need back-and-forth assistance.
- ✅ Versatile: Powerful and easy-to-use CLI tool for various AI tasks
- 🪶 Lightweight: Minimal dependencies with everything you need included
- 🔄 API Flexibility: Works with OpenAI, Ollama, Groq, Claude, Gemini, and any OpenAI-compatible endpoint
- 💬 Interactive Chat: Continuous conversation with memory in modern UI
- 📊 Streaming Responses: Real-time output for better user experience
- 🔍 Web Search: Enhance any model with contextual information from the web, using advanced content extraction to identify the most relevant information from web pages
- 📥 Stdin Processing: Process piped content by using
{}placeholder in prompts - 🎨 Markdown Rendering: Beautiful formatting of markdown and code with syntax highlighting
- ⚡ Real-time Markdown: Stream responses with live updating syntax highlighting and formatting
- ⚙️ Multiple Configurations: Cross-platform config system supporting different profiles
- 💻 Shell Command Generation: OS-aware command execution
- 🧠 Text Rewriting: Improve text quality while maintaining original tone and meaning
- 🧩 Clean Code Generation: Output code without markdown or explanations
- 📝 Rich Multiline Editor: Interactive multiline text input with syntax highlighting and intuitive controls
- 📑 Git Commit Messages: AI-powered generation of conventional, detailed commit messages from git diffs
- 🎭 System Prompts: Customize model behavior with custom system prompts
- 🤖 Custom Roles: Create and use reusable AI roles for specialized tasks
- 📃 Conversation Logging: Save your conversations to text files for later reference
- 🔌 Modular Architecture: Well-structured codebase with clean separation of concerns
- 🔄 Provider Switching: Easily switch between different LLM providers with a single parameter
- 🚀 Performance Optimized: Fast response times and minimal resource usage
See the Feature Overview for more details.
# Installation with pip
pip install ngpt
# Or install with uv (faster installation)
uv pip install ngpt
# Or install globally as a CLI tool (recommended for command-line usage)
uv tool install ngpt
# Arch Linux: install from AUR
paru -S ngptRequires Python 3.8 or newer.
For detailed installation instructions, see the Installation Guide.
# Chat with default settings
ngpt "Tell me about quantum computing"
# Alternatively, run as a Python module
python -m ngpt "Tell me about quantum computing"
# Start an interactive chat session with conversation memory
ngpt -i
# Return response without streaming
ngpt --no-stream "Tell me about quantum computing"
# Generate code
ngpt --code "function to calculate the Fibonacci sequence"
# Generate code with syntax highlighting
ngpt --code --prettify "function to calculate the Fibonacci sequence"
# Generate code with real-time syntax highlighting
ngpt --code --stream-prettify "function to calculate the Fibonacci sequence"
# Generate and execute shell commands
ngpt --shell "list all files in the current directory"
# Read from stdin and use the content in your prompt
echo "What is this text about?" | ngpt --pipe "Analyze the following text: {}"
# Using here-string (<<<) for quick single-line input
ngpt --pipe {} <<< "What is the best way to learn shell redirects?"
# Using standard input redirection to process file contents
ngpt --pipe "summarise {}" < README.md
# Using here-document (<<EOF) for multiline input
ngpt --pipe {} << EOF
What is the best way to learn Golang?
Provide simple hello world example.
EOF
# Create a custom role for specialized tasks
ngpt --role-config create json_generator
# Use a custom role for specific tasks
ngpt --role json_generator "Generate user data with name, email, and address"
# Use a role from the Role Gallery (first create it, then use it)
ngpt --role-config create sql_expert
# Paste the SQL Expert role from https://nazdridoy.github.io/ngpt/examples/role-gallery/
ngpt --role sql_expert "Write a query to find all users who made a purchase in the last 30 days"
# Rewrite text to improve quality while preserving tone and meaning
echo "your text" | ngpt -r
# Rewrite text from a command-line argument
ngpt -r "your text to rewrite"
# Rewrite text from a file
cat file.txt | ngpt -r
# Generate AI-powered git commit messages for staged changes
ngpt -g
# Generate commit message from staged changes with a context directive
ngpt -g --preprompt "type:feat"
# Process large diffs in chunks with recursive analysis
ngpt -g --rec-chunk
# Process a diff file instead of staged changes
ngpt -g --diff /path/to/changes.diff
# Use piped diff content for commit message generation
git diff HEAD~1 | ngpt -g --pipe
# Generate a commit message with logging for debugging
ngpt -g --log commit_log.txt
# Use interactive multiline editor to enter text to rewrite
ngpt -r
# Display markdown responses with beautiful formatting
ngpt --prettify "Explain markdown syntax with examples"
# Display markdown responses with real-time formatting
ngpt --stream-prettify "Explain markdown syntax with examples"
# Use a specific markdown renderer
ngpt --prettify --renderer=rich "Create a markdown table"
# Use multiline editor for complex prompts
ngpt --text
# Use custom system prompt
ngpt --preprompt "You are a Linux expert" "How do I find large files?"
# Log your conversation to a file
ngpt --interactive --log conversation.log
# Create a temporary log file automatically
ngpt --log "Tell me about quantum computing"
# Process text from stdin using the {} placeholder
cat README.md | ngpt --pipe "Summarize this document: {}"
# Use different model providers by specifying the provider name
ngpt --provider Groq "Explain quantum computing"
# Compare outputs from different providers
ngpt --provider OpenAI "Explain quantum physics" > openai_response.txt
ngpt --provider Ollama "Explain quantum physics" > ollama_response.txt
# Show all API configurations
ngpt --show-config --all
# List available models for the active configuration
ngpt --list-models
# List models for a specific configuration (index)
ngpt --list-models --config-index 1
# List models for a specific configuration (provider)
ngpt --list-models --provider Gemini
# With custom options
ngpt --api-key your-key --base-url http://your-endpoint --model your-model "Hello"
# Enable web search capability to enhance prompts with web information
ngpt --web-search "What's the latest news about AI?"
# Generate and execute shell commands (using -s or --shell flag)
# OS-aware: generates appropriate commands for Windows, macOS, or Linux
ngpt -s "list all files in current directory"
# On Windows generates: dir
# On Linux/macOS generates: ls -la
# Generate code (using -c or --code flag)
ngpt -c "create a python function that calculates fibonacci numbers"
# Use multiline text editor for complex prompts (using -t or --text flag)
ngpt -tFor more examples and detailed usage, visit the CLI Usage Guide.
❯ ngpt -h
usage: ngpt [-h] [-v] [--language LANGUAGE] [--config [CONFIG]] [--config-index CONFIG_INDEX] [--provider PROVIDER]
[--remove] [--show-config] [--all] [--list-models] [--list-renderers] [--cli-config [COMMAND ...]]
[--role-config [ACTION ...]] [--api-key API_KEY] [--base-url BASE_URL] [--model MODEL] [--web-search]
[--pipe] [--temperature TEMPERATURE] [--top_p TOP_P] [--max_tokens MAX_TOKENS] [--log [FILE]]
[--preprompt PREPROMPT | --role ROLE] [--no-stream | --prettify | --stream-prettify]
[--renderer {auto,rich,glow}] [--rec-chunk] [--diff [FILE]] [--chunk-size CHUNK_SIZE]
[--analyses-chunk-size ANALYSES_CHUNK_SIZE] [--max-msg-lines MAX_MSG_LINES]
[--max-recursion-depth MAX_RECURSION_DEPTH] [-i | -s | -c | -t | -r | -g]
[prompt]
nGPT - Interact with AI language models via OpenAI-compatible APIs
positional arguments::
[PROMPT] The prompt to send
options::
-h, --help show this help message and exit
-v, --version Show version information and exit
--language LANGUAGE Programming language to generate code in (for code mode)
Configuration Options::
--config [CONFIG] Path to a custom config file or, if no value provided, enter interactive configuration mode to create a new config
--config-index CONFIG_INDEX Index of the configuration to use or edit (default: 0)
--provider PROVIDER Provider name to identify the configuration to use
--remove Remove the configuration at the specified index (requires --config and --config-index or --provider)
--show-config Show the current configuration(s) and exit
--all Show details for all configurations (requires --show-config)
--list-models List all available models for the current configuration and exit
--list-renderers Show available markdown renderers for use with --prettify
--cli-config [COMMAND ...] Manage CLI configuration (set, get, unset, list, help)
--role-config [ACTION ...] Manage custom roles (help, create, show, edit, list, remove) [role_name]
Global Options::
--api-key API_KEY API key for the service
--base-url BASE_URL Base URL for the API
--model MODEL Model to use
--web-search Enable web search capability using DuckDuckGo to enhance prompts with relevant information
--pipe Read from stdin and use content with prompt. Use {} in prompt as placeholder for stdin content. Can be used with any mode option except --text and --interactive
--temperature TEMPERATURE Set temperature (controls randomness, default: 0.7)
--top_p TOP_P Set top_p (controls diversity, default: 1.0)
--max_tokens MAX_TOKENS Set max response length in tokens
--log [FILE] Set filepath to log conversation to, or create a temporary log file if no path provided
--preprompt PREPROMPT Set custom system prompt to control AI behavior
--role ROLE Use a predefined role to set system prompt (mutually exclusive with --preprompt)
--renderer {auto,rich,glow} Select which markdown renderer to use with --prettify or --stream-prettify (auto, rich, or glow)
Output Display Options (mutually exclusive)::
--no-stream Return the whole response without streaming or formatting
--prettify Render complete response with markdown and code formatting (non-streaming)
--stream-prettify Stream response with real-time markdown rendering (default)
Git Commit Message Options::
--rec-chunk Process large diffs in chunks with recursive analysis if needed
--diff [FILE] Use diff from specified file instead of staged changes. If used without a path, uses the path from CLI config.
--chunk-size CHUNK_SIZE Number of lines per chunk when chunking is enabled (default: 200)
--analyses-chunk-size ANALYSES_CHUNK_SIZE Number of lines per chunk when recursively chunking analyses (default: 200)
--max-msg-lines MAX_MSG_LINES Maximum number of lines in commit message before condensing (default: 20)
--max-recursion-depth MAX_RECURSION_DEPTH Maximum recursion depth for commit message condensing (default: 3)
Modes (mutually exclusive)::
-i, --interactive Start an interactive chat session
-s, --shell Generate and execute shell commands
-c, --code Generate code
-t, --text Enter multi-line text input (submit with Ctrl+D)
-r, --rewrite Rewrite text from stdin to be more natural while preserving tone and meaning
-g, --gitcommsg Generate AI-powered git commit messages from staged changes or diff file
Note: For better visualization of conventional commit messages on GitHub, you can use the GitHub Commit Labels userscript, which adds colorful labels to your commits.
For a complete reference of all available options, detailed CLI examples and usage information, see the CLI Usage Guide.
Comprehensive documentation, including usage guides and examples, is available at:
https://nazdridoy.github.io/ngpt/
Key documentation sections:
- Installation Guide
- CLI Usage Guide
- Configuration Guide
- Custom Roles Guide
- Role Gallery
- Examples & Tutorials
- Git Commit Message Guide
- Create an account at OpenAI
- Navigate to API keys: https://platform.openai.com/api-keys
- Click "Create new secret key" and copy your API key
- Configure nGPT with your key:
ngpt --config # Enter provider: OpenAI # Enter API key: your-openai-api-key # Enter base URL: https://api.openai.com/v1/ # Enter model: gpt-3.5-turbo (or other model)
- Create or use an existing Google account
- Go to Google AI Studio
- Navigate to API keys in the left sidebar (or visit https://aistudio.google.com/app/apikey)
- Create an API key and copy it
- Configure nGPT with your key:
ngpt --config # Enter provider: Gemini # Enter API key: your-gemini-api-key # Enter base URL: https://generativelanguage.googleapis.com/v1beta/openai # Enter model: gemini-2.0-flash
For more detailed information, refer to the API Key Setup documentation.
NGPT offers a CLI configuration system that allows you to set default values for command-line options. This is especially useful when you:
- Repeatedly use the same provider or model
- Have preferred settings for specific tasks
- Want to create different workflows based on context
For example, setting your preferred language for code generation or temperature value means you won't have to specify these parameters each time:
❯ ngpt --cli-config help
CLI Configuration Help:
Command syntax:
ngpt --cli-config help - Show this help message
ngpt --cli-config set OPTION VALUE - Set a default value for OPTION
ngpt --cli-config get OPTION - Get the current value of OPTION
ngpt --cli-config get - Show all CLI configuration settings
ngpt --cli-config unset OPTION - Remove OPTION from configuration
ngpt --cli-config list - List all available options
Available options:
General options (all modes):
config-index - int (default: 0) [exclusive with: provider]
log - str
max_tokens - int
no-stream - bool (default: False) [exclusive with: prettify, stream-prettify]
preprompt - str
prettify - bool (default: False) [exclusive with: no-stream, stream-prettify]
provider - str [exclusive with: config-index]
renderer - str (default: auto)
stream-prettify - bool (default: False) [exclusive with: no-stream, prettify]
temperature - float (default: 0.7)
top_p - float (default: 1.0)
web-search - bool (default: False)
Options for Code generation mode:
language - str (default: python)
Options for Git commit message mode:
analyses-chunk-size - int (default: 200)
chunk-size - int (default: 200)
diff - str
max-msg-lines - int (default: 20)
max-recursion-depth - int (default: 3)
rec-chunk - bool (default: False)
Example usage:
ngpt --cli-config set language java - Set default language to java for code generation
ngpt --cli-config set provider Gemini - Set Gemini as your default provider
ngpt --cli-config set temperature 0.9 - Set default temperature to 0.9
ngpt --cli-config set no-stream true - Disable streaming by default
ngpt --cli-config set recursive-chunk true - Enable recursive chunking for git commit messages
ngpt --cli-config set diff /path/to/file.diff - Set default diff file for git commit messages
ngpt --cli-config get temperature - Check the current temperature setting
ngpt --cli-config get - Show all current CLI settings
ngpt --cli-config unset language - Remove language setting
Notes:
- CLI configuration is stored in:
• Linux: ~/.config/ngpt/ngpt-cli.conf
• macOS: ~/Library/Application Support/ngpt/ngpt-cli.conf
• Windows: %APPDATA%\ngpt\ngpt-cli.conf
- Settings are applied based on context (e.g., language only applies to code generation mode)
- Command-line arguments always override CLI configuration
- Some options are mutually exclusive and will not be applied together
For more details, see the CLI Configuration Guide.
The --config option without arguments enters interactive configuration mode, allowing you to add or edit configurations:
# Add a new configuration
ngpt --config
# Edit an existing configuration at index 1
ngpt --config --config-index 1
# Edit an existing configuration by provider name
ngpt --config --provider Gemini
# Remove a configuration at index 2
ngpt --config --remove --config-index 2
# Remove a configuration by provider name
ngpt --config --remove --provider Gemini
# Use a specific configuration by provider name
ngpt --provider OpenAI "Tell me about quantum computing"In interactive mode:
- When editing an existing configuration, press Enter to keep the current values
- When creating a new configuration, press Enter to use default values
- For security, your API key is not displayed when editing configurations
- When removing a configuration, you'll be asked to confirm before deletion
For more details on configuring nGPT, see the Configuration Guide.
nGPT uses a configuration file stored in the standard user config directory for your operating system:
- Linux:
~/.config/ngpt/ngpt.confor$XDG_CONFIG_HOME/ngpt/ngpt.conf - macOS:
~/Library/Application Support/ngpt/ngpt.conf - Windows:
%APPDATA%\ngpt\ngpt.conf
The configuration file uses a JSON list format, allowing you to store multiple configurations. You can select which configuration to use
6D40
with the --config-index argument (or by default, index 0 is used).
[
{
"api_key": "your-openai-api-key-here",
"base_url": "https://api.openai.com/v1/",
"provider": "OpenAI",
"model": "gpt-4o"
},
{
"api_key": "your-groq-api-key-here",
"base_url": "https://api.groq.com/openai/v1/",
"provider": "Groq",
"model": "llama3-70b-8192"
},
{
"api_key": "your-ollama-key-if-needed",
"base_url": "http://localhost:11434/v1/",
"provider": "Ollama-Local",
"model": "llama3"
}
]For details on the configuration file format and structure, see the Configuration Guide.
nGPT determines configuration values in the following order (highest priority first):
- Command line arguments (
--api-key,--base-url,--model, etc.) - Environment variables (
OPENAI_API_KEY,OPENAI_BASE_URL,OPENAI_MODEL) - CLI configuration file (
ngpt-cli.conf, managed with--cli-config) - Main configuration file
ngpt.conforcustom-config-file - Default values
Let's see nGPT in action! Here are some practical ways you can use it every day:
# Get a quick explanation
ngpt "Explain the difference between threads and processes in Python"
# Generate code with real-time syntax highlighting
ngpt --code --stream-prettify "Write a Python function to reverse a linked list"With the --code flag, nGPT gives you clean code without explanations or markdown, just what you need to copy and paste into your project. The --stream-prettify option shows real-time syntax highlighting as the code comes in.
# Let nGPT generate the correct command for your OS
ngpt --shell "list all files in the current directory including hidden ones"
# On Linux/macOS: ls -la
# On Windows: dir /aOne of my favorite features! No more Googling obscure command flags, nGPT generates the right command for your operating system. It'll even execute it for you if you approve.
# Pipe text to rewrite it (e.g., improve clarity)
echo "This is a rough draft of my email." | ngpt -r
# Summarize a file using the pipe placeholder
cat long-article.txt | ngpt --pipe "Summarize this document concisely: {}"The text rewriting feature is perfect for quickly improving documentation, emails, or reports. And with pipe placeholders, you can feed in content from files or other commands.
# Stage your changes
git add .
# Let nGPT generate a conventional commit message based on the diff
ngpt -g
# Generate git commit message from a diff file
ngpt -g --diff changes.diffThis is a huge time-saver. nGPT analyzes your git diff and generates a properly formatted conventional commit message that actually describes what you changed. No more staring at the blank commit message prompt!
# Create a specialized role for JSON generation
ngpt --role-config create json_generator
# Use the custom role to generate structured data
ngpt --role json_generator "Generate random user profile data"{
"id": "a1b2c3d4-e5f6-7890-1234-567890abcdef",
"firstName": "Aurora",
"lastName": "Reynolds",
"email": "aurora.reynolds@example.com",
"phone": "+1-555-0101",
"address": {
"street": "123 Main St",
"city": "Anytown",
"state": "CA",
"zipCode": "90210"
},
"birthDate": "1990-07-15",
"registrationDate": "2022-01-20",
"isActive": true,
"roles": [
"user",
"premium"
]
}Custom roles allow you to create reusable AI personas for consistent responses across various prompts. For more details, see the Custom Roles Guide and check out the Role Gallery for ready-to-use roles.
# Ask questions that require up-to-date information
ngpt --web-search "What's the latest news about AI regulation?"The --web-search flag lets nGPT consult the web for recent information, making it useful for questions about current events or topics that might have changed since the AI's training data cutoff.
Let's look at how nGPT can fit into your everyday workflow with some practical examples:
As a developer, I use nGPT throughout my day:
Morning code review:
# Get explanations of complex code
git show | ngpt --pipe "Explain what this code change does and any potential issues: {}"Debugging help:
# Help understand a cryptic error message
npm run build 2>&1 | grep Error | ngpt --pipe "What does this error mean and how can I fix it: {}"Documentation generation:
# Generate JSDoc comments for functions
cat src/utils.js | ngpt --pipe "Write proper JSDoc comments for these functions: {}"Commit messages:
# After finishing a feature
git add .
ngpt -gFor content creators and writers:
Overcoming writer's block:
ngpt "Give me 5 different angles to approach an article about sustainable technology"Editing assistance:
cat draft.md | ngpt -rcurl -s https://example.com/research-paper.html | ngpt --pipe "Summarize the key findings from this research: {}"For sysadmins and DevOps folks:
Generating complex commands:
ngpt -s "find all log files larger than 100MB that haven't been modified in the last 30 days"Creating configuration files*:
ngpt --code "Create a Docker Compose file for a Redis, PostgreSQL, and Node.js application"Troubleshooting systems:
dmesg | tail -50 | ngpt --pipe "Explain what might be causing the issues based on these system logs: {}"We welcome contributions to nGPT! Whether it's bug fixes, feature additions, or documentation improvements, your help is appreciated.
To contribute:
- Fork the repository
- Create a feature branch:
git checkout -b feature/your-feature-name - Make your changes
- Commit with clear messages following conventional commit guidelines
- Push to your fork and submit a pull request
Please check the CONTRIBUTING.md file for detailed guidelines on code style, pull request process, and development setup.
This project is licensed under the MIT License. See the LICENSE file for details.