LLaMB - Command Line LLM Client

LLaMB is a command-line tool that allows you to interact with Large Language Models directly from your terminal. It features conversation history, file input/output, and support for multiple LLM providers.

Features

• 🚀 Fast - Get answers directly in your terminal without leaving your workflow
• 💬 Conversation history - Follow-up on previous questions with context
• 🔄 Continuous conversation - Interactive chat mode for follow-up questions
• 📄 File handling - Include file contents in your questions and save responses to files
• 🌐 URL content - Fetch URL content using Jina Reader to include in prompts
• 🔤 Smart file extensions - Automatically uses appropriate file extensions for code outputs
• 📝 Prompt templates - Save and reuse common prompts to streamline frequent tasks
• 🔄 Multiple providers - Support for OpenAI, Anthropic, Mistral, Ollama, and more
• 🔐 Secure - API keys stored securely in your system's credential manager
• 🖥️ Local models - Works with local models like Ollama and LM Studio
• 🎨 React-based UI - Clean, artifact-free terminal interface using ink
• 🧠 Smart error handling - Detects offline providers and offers alternatives
• 🛠️ Provider management - Edit, switch, and configure providers easily

Installation

npm install -g llamb

Or clone this repository and run:

npm install
npm run build
npm link

Setup

You can set your OpenAI API key in your environment:

export OPENAI_API_KEY=your_api_key_here

Or add a provider using the interactive command:

llamb provider:add

You can set your Jina Reader API key:

llamb jina:apikey

LLaMB supports many providers out of the box with pre-configured settings:

• OpenAI
• Anthropic
• Mistral AI
• OpenRouter
• Ollama (local)
• LM Studio (local)
• Other/Custom

Secure API Key Storage

LLaMB securely stores your API keys in your system's secure credential store:

• macOS: Keychain
• Windows: Credential Manager
• Linux: libsecret (used by GNOME/KDE)

This means your API keys are never stored in plain text in configuration files.

You can update API keys for existing providers:

llamb provider:apikey

If you install from source, you can install the man page with:

sudo npm run install-man

Documentation

LLaMB comes with a comprehensive man page that you can access after installation:

man llamb

Usage

Ask a question

llamb what is the best command to install docker on ubuntu

File and URL Operations

Include a file in your question

llamb -f script.js "Explain this code and suggest improvements"
llamb "Summarize this document" -f document.txt

Fetch URL content with Jina Reader

# Basic usage - fetch URL content and analyze it
llamb -j https://example.com "Explain this website"

# Combine with other options
llamb -j https://github.com/owner/repo/blob/main/README.md -p openai "What features does this project have?"

# Set or update the Jina Reader API key (optional)
llamb jina:apikey

# Test Jina Reader with a URL
llamb jina:test https://example.com

Save responses to files

You can save responses to files in several ways:

# Prompted for filename (default: llamb-response-<timestamp>.txt)
llamb "Generate a React component" -o

# Save to specific file
llamb "Generate a JSON configuration" -o config.json

# Overwrite existing files without prompting
llamb "Generate a CSS file" -o styles.css --overwrite

If a file already exists, you'll be asked whether to:

• Overwrite the existing file
• Generate a new filename automatically (e.g., file-1.txt)
• Cancel the save operation

Continuous Conversation Mode

LLaMB provides an interactive chat mode that allows you to have back-and-forth conversations with the model:

llamb -c "Tell me about TypeScript"

After receiving the initial response, you'll be prompted for follow-up questions:

🦙 Follow-up question (type /exit to end conversation):

You can type your follow-up questions, or use special commands:

• /exit or /quit - End the conversation and exit
• /clear - Clear conversation history
• /new - Start a new conversation
• /history - View conversation history
• /file - Attach a file to your next question
• /unfile - Remove the attached file
• /help - Show available commands

Conversation History

LLaMB keeps track of your conversation history to provide context for follow-up questions.

View conversation history

llamb /history
# or
llamb context:history

Clear conversation history

llamb /clear
# or
llamb context:clear

Start a new conversation

llamb /new
# or
llamb context:new

Ask without using conversation history

llamb -n "What is 2+2?"
# or
llamb --no-history "What is 2+2?"

Terminal-specific sessions

LLaMB maintains separate conversation histories for each terminal window. This means you can have different conversations going on in different terminal sessions.

To see information about your current terminal session:

llamb /debug
# or
llamb context:debug

Working with Models and Providers

List available models

llamb models
# or
llamb /models

List configured providers

llamb providers
# or
llamb /providers

Add a provider

llamb provider:add

Edit an existing provider

# Interactive mode
llamb provider:edit

# Non-interactive mode
llamb provider:edit --name openai --url https://api.openai.com/v1
llamb provider:edit --name openai --model gpt-4o
llamb provider:edit --name openai --key YOUR_API_KEY

Delete a provider

# Interactive mode - select provider and confirm
llamb provider:delete

# Non-interactive mode
llamb provider:delete --name openai

# Skip confirmation prompt
llamb provider:delete --name openai --force

Set the default provider

llamb provider:default

Change the default model for a provider

llamb model:default
# or
llamb /model

You can also change the model for a specific provider:

llamb model:default -p openai

Specify model or provider for a question

llamb -m gpt-4 -p openai how do I install nginx on Ubuntu?

Use a custom base URL for a single request

llamb -u http://localhost:8080/v1 what is the meaning of life?

Terminal UI Options

LLaMB provides different UI rendering options to handle different terminal environments:

Use ink-based UI (default)

LLaMB uses a React-based terminal UI powered by ink, which provides a clean, artifact-free interface with proper rendering:

# ink is the default, so no flag is needed
llamb "What is the history of the internet?"

# You can explicitly enable it with
llamb --ink "What is the history of the internet?"

Disable ink for traditional rendering

In some terminal environments, you might want to fall back to traditional rendering:

llamb --no-ink "What is the history of the internet?"

Use progress-only mode

If you're experiencing issues with scrollback artifacts in your terminal:

llamb --progress-only "What is the history of the internet?"

Configure terminal UI defaults

# Set ink UI as default (already the default)
llamb config:progress-mode --ink

# Use traditional rendering with real-time streaming
llamb config:progress-mode --disable

# Use progress-only mode (no streaming)
llamb config:progress-mode --enable

# View current settings
llamb config:progress-mode

Combine options for maximum flexibility

llamb -m llama2 -p ollama -f code.py -o analysis.md "Analyze this code"

Saved Prompt Templates

LLaMB allows you to save and reuse common prompts to streamline your workflow.

List all saved prompts

llamb prompt:list

Add a new prompt template

llamb prompt:add

This will prompt you for:

• A name for the prompt (e.g., "summarize")
• The prompt text (e.g., "Please summarize the following text in bullet points:")
• Optional description for your reference

Use a saved prompt

# Use with a file
llamb -t summarize -f document.txt

# Use with direct input
llamb -t code-review "function add(a, b) { return a + b; }"

# Combine with other options
llamb -t translate -p openai -m gpt-4 -o spanish.txt "Hello world"

Edit a saved prompt

llamb prompt:edit

Delete a prompt template

llamb prompt:delete

Show a prompt's details

llamb prompt:show summarize

Using placeholders in prompts

Prompt templates can include special placeholders that get replaced when used:

• {input} - Gets replaced with the command-line question text
• {file} - Gets replaced with file contents when using -f flag
• {filename} - Gets replaced with the filename when using -f flag

Example prompt template with placeholders:

Please analyze the following code from {filename}:

{file}

Focus on these aspects:
1. Performance
2. Security
3. Readability

Additional notes: {input}

When used as llamb -t analyze -f script.js "Focus on error handling", the placeholders will be replaced with the appropriate values.

Smart File Extensions

When saving responses to files, LLaMB automatically detects the content type and applies appropriate file extensions:

# When asking for code and saving with -o flag (without extension)
llamb "Write a Python function to calculate fibonacci" -o fibonacci
# Will save as fibonacci.py if response is only code

# When providing a filename without extension
llamb "Write a JavaScript React component" -o component
# Will save as component.jsx if response is purely JSX code

This feature:

• Only applies language extensions when the content is a pure code block
• Defaults to .txt for mixed content (explanations + code)
• Always respects explicitly provided extensions

License

MIT