gllm
gllm - Golang Command-Line LLM Companion
gllm is a powerful command-line tool designed to interact seamlessly with various Large Language Models (LLMs). Supports features like interactive chat, multi-turn conversations, file attachments, search integration, command agent, and extensive customization.
🚀 Features
- Flexible Model Selection: Easily configure and switch between different LLMs.
- Interactive Chat Mode: Start real-time conversations with AI models.
- Prompt Templates & System Prompts: Manage reusable prompts and instructions.
- Attachment Support: Process files and images as part of queries.
- Search Support: Using search engines, find relevant and latest information.
- Reading PDF Support: Google models support PDF processing (OpenAI compatibles only for text/image).
- Reasoning Support: Generate detailed explanations, logical breakdowns, and step-by-step analysis.
- Markdown Support: Format Markdown contents and make it easier to read.
- Multi-turn Chat: Engage in multiple rounds of conversation.
- Check out history and context: In chat mode or multi-turn chat, easily view the full context.
- Configuration Management: Customize model behavior and settings.
- Command Agent Mode: Let LLMs plan and execute commands with your confirmation.
- Version Control: Easily track and update your setup.
📌 Installation
# Install via package manager (if available)
brew tap activebook/gllm
brew install gllm
# Or manually build from source
git clone https://github.com/activebook/gllm.git
cd gllm
go build -o gllm
📦 Upgrade
brew tap activebook/gllm
brew upgrade gllm
🎯 Usage
🔹 Basic Commands
gllm "What is Go?" # Default model & system prompt
gllm "Where is the best place to visit in London?" -m gpt4o # Switch model
gllm "How to find a process and terminate it?" -t shellmate # Use shellmate prompt to specific shell question
gllm -s "Who's the POTUS right now? and check what's his latest tariff policy" -m gemini-pro -r 10 # Use Gemini model to search and set max references to 10
🔹 Attachments (Files, Image, Urls)
gllm "Summarize this" -a report.txt # Use file as input
gllm "Translate into English" -a image1.jpg # Use image as input and vision model
gllm "Summarize all:" -a "https://www.cnbc.com" -a "https://www.nytimes.com" # Attach multiple urls
gllm "Transcribe this audio" -a speech.mp3 # Use audil as input (only for gemini multimodal models)
🔍 Search & Vision
gllm "Who is the President of the United States right now?" --search # Use search to find latest news
gllm "Who is he/she in this photo? And what is his/her current title?" -s -a "face.png" --model gemini # Use vision model and search engine to find people in image
gllm "When was gpt4.5 released?" --search=google # Use specific search engine to find latest news
💬 Keep Conversations (Multi-turn chat)
gllm -s "Who's the POTUS right now?" -c abc # Start a conversation and retain the full context (last 10 messages)
gllm "Tell me again, who's the POTUS right now?" -c abc # Continue the default conversation
gllm "Let's talk about why we exist." -c newtalk # Start a new named conversation called 'newtalk'
gllm -s "Look up what famous people have said about this." -c newtalk # Continue the 'newtalk' conversation
gllm "Tell me more about his books." -c 1 # Continue the 'newtalk' conversation" with convo index (latest convo)
⚠️ Warning: If you’re using Gemini mode and an OpenAI-compatible model, keep in mind that they cannot be used within the same conversation.
These models handle chat messages differently, and mixing them will lead to unexpected behavior.
🖥️ » Interactive Chat (Available!)
gllm chat # Start chat with defaults
gllm chat -m gpt4o # Start chat with a specific model
gllm chat --sys-prompt coder # Use a named system prompt
gllm chat -c my_chat # Start a new chat session
gllm chat -c 1 # Follow a previous chat session with a convo index
gllm chat -s # Start a new conversation with search
🛠️ You can change setting in the chat session
gllm> /exit, /quit # Exit the chat session
gllm> /clear, /reset # Clear context
gllm> /help # Show available commands
gllm> /history /h [num] [chars] # Show recent conversation history (default: 20 messages, 200 chars)
gllm> /markdown, /mark [on|off|only] # Switch whether to render markdown or not
gllm> /system, /S [name|prompt] # change system prompt
gllm> /template, /t [name|tmpl] # change template
gllm> /search, /s [search_engine] # select a search engine to use
gllm> /reference. /r [num] # change link reference count
gllm> /attach, /a [filename] # Attach a file to the chat session
gllm> /detach, /d [filename|all] # Detach a file to the chat session
⚠️ Warning: You can’t switch models within the same conversation. Once you choose a model, you’ll need to stick with it throughout. Just like when using different models online, you can continue or change topics, you can do search and attach files, but the model type remains the same.
🔹 Prompt Templates
gllm --template coder # Use predefined coder prompt
gllm "Act as shell" --system-prompt "You are a Linux shell..."
gllm --system-prompt shell-assistant --template shellmate
🔹 Configuration Management
gllm config path # Show config file location
gllm config show # Display loaded configurations
🔹 Model Management
gllm model list # List available models
gllm model add --name gpt4 --key $API_KEY --model gpt-4o --temp 0.7
gllm model default gpt4 # Set default model
🔹 Template & System Prompt Management
gllm template list # List available templates
gllm template add coder "You are an expert Go programmer..."
gllm system add --name coder --content "You are an expert Go programmer..."
gllm system default coder # Set default system prompt
🔹 New update! & Search Engine Management
gllm search list # List available search engines
gllm search google --key $API_KEY --cx $SEARCH_ENGINE_ID # Use Google Search Engine
gllm search tavily --key $API_KEY # Use Tavily Search Engine
gllm search default [google,tavily] # Set default search engine
gllm search save [on|off] # Save search results on conversation history (careful! could induce token consumption. default: off)
🔹 New update! & Conversation Management
gllm convo list # list all conversations
gllm convo remove newtalk # remove a conversation
gllm convo remove "chat_*" # wildcard remove multiple conversations
gllm convo info newtalk # show a conversation in details
gllm convo info 1 # use index to view a conversation in details
gllm convo info newtalk -n 100 -c 300 # view history context (the lastest 100 messages, 300 characters each)
gllm convo clear # clear all conversations
🔹 New update! & Markdown output Management
gllm markdown on # enable markdown output which append the end of the streaming contents
gllm markdown off # disable markdown output
gllm markdown only # only render markdown output
🔹 New update! & Plugins Management
gllm plugin list # list all plugins, loaded add a checkmark before the name
gllm plugin load exec # load exec plugin, to execute the command
gllm plugin unload exec # unload exec plugin
🔹 Std input Support
cat script.py | gllm "Help me fix bugs in this python coding snippet:" -a - # Use std input as file attachment
cat image.png | gllm "What is this image about?" -a - # Use std input as image attachment
cat jap.txt | gllm "Translate all this into English" # Use std input as text input
cat report.txt | gllm "Summarize this"
echo "What is the capital of France?" | gllm # Use std input as text input
echo "Who's the POTUS right now?" | gllm -s # Use std input as search query
🔹 Python code execution (only for gemini2.0)
gllm --code "import time\\nstart = time.time()\\nis_prime = lambda n: n > 1 and all(n % d for d in range(2, int(n**0.5) + 1))\\nLIMIT = 100000\\ncount = sum(1 for x in range(LIMIT) if is_prime(x))\\nend = time.time()\\nprint(f'Number of primes less than {LIMIT}: {count}')\\nprint(f'Time taken: {end - start:.2f} seconds')" -m gemini2.5
🔹 Version Information
gllm version
gllm --version
🛠 Configuration
By default, gllm stores configurations in a user-specific directory. Use the config commands to manage settings.
default:
model: gpt4
system_prompt: coder
template: default
search: google
markdown: on
models:
- name: gpt4
endpoint: "https://api.openai.com"
key: "$OPENAI_KEY"
model: "gpt-4o"
temperature: 0.7
search_engines:
- google:
key: "$GOOGLE_API_KEY"
- tavily:
💡 Why gllm?
- Simplifies interaction with LLMs via CLI.
- Supports multiple models and configurations.
- Powerful customization with templates and prompts.
- Works with text, code, and image-based queries.
- Supports search with image and query.
- Check reasoning details.
- Easy to read markdown formatted output.
- Supports multiple conversations.
- Powerful and flexible Command Agent mode.
- Check out context details.
Start using gllm today and supercharge your command-line AI experience! 🚀
Project Features
This project includes various features to enhance usability and efficiency. Below is an overview of the key functionalities.
Installation & Usage
| Topic & Description | Screenshot |
|---|---|
| Install: Easily install using Homebrew |  |
| Upgrade: Upgrade to the latest version |  |
| How to Use: Simply run the command |  |
Core Functionalities
| Feature & Description | Screenshot |
|---|---|
| General Usage: Quick overview of core commands |  |
| Search Info: Perform smart, targeted searches |  |
| Search Thoroughly: Deep-dive web searches for comprehensive results |  |
| Configuration: Customize settings to fit your workflow |  |
| Reasoning: Leverage advanced reasoning capabilities |  |
Additional Features
| Feature & Description | Screenshot |
|---|---|
| Multi-Search: Run multiple searches in one command |  |
| Multi-Turn: Continue previous conversations seamlessly |  |
| PDF Reading (Gemini only): Extract and analyze PDF content |  |
| Code Execution (Gemini only): Execute Python code directly |  |
| Markdown Output: Generate clean, readable Markdown |  |
| One-Shot Agent: Run task in one shot |  |
Interactive Chat Features
| Feature & Description | Screenshot |
|---|---|
| Chat Mode: Enter an interactive chat session |  |
| Follow-Up: Pose follow-up questions in the same session |  |
| Chat History with Tool Calls: View your conversation context and tool usage |  |
| Chat History with Reasoning: Inspect past reasoning steps alongside your chat |  |
| Command Agent mode: Utilize the power of command-line agents |  |
| Multi-Commands In One-Shot: Run multiple commands in one shot |  |
For more details, using gllm --help to check.
🏗 Contributing
@cite { @author: Charles Liu @github: https://github.com/activebook @website: https://activebook.github.io }