LLM Chat CLI

A powerful, fast, and beautiful terminal-based chat application for interacting with multiple Large Language Model providers.

╔═══════════════════════════════════════╗
║     🤖 LLM Chat CLI v0.1.0        ║
╚═══════════════════════════════════════╝

✨ Features

Core Features

🚀 Multiple LLM Providers - Ollama, Together AI, Groq, SambaNova, Google Gemini
💬 Interactive Chat Mode - Full conversational context
🔧 Shell Mode - Pipe input for scripting and automation
⚡ Streaming Responses - Real-time output as models think
🔄 Model Switching - Switch models on the fly
📝 Multiple Output Formats - text, json, markdown, raw

Advanced Features

📊 Prompt Assessment - AI-powered analysis of your prompts (8 criteria)
✨ Auto-Improvement - Let AI help you write better prompts
📚 Prompt Engineering Guide - Built-in best practices
💾 Persistent History - All conversations saved and searchable
🔍 Search History - Find past conversations
📤 Export Conversations - Markdown, JSON, or plain text
📈 Usage Statistics - Track your conversations

Polish

🎨 Beautiful UI - Colors, emojis, clean layout
🎯 Smart Commands - Intuitive / commands
⚙️ Highly Configurable - CLI flags for everything
🔐 Privacy-Focused - Local-first, optional history saving

🚀 Quick Start

Installation

Option 1: Build from source (recommended)

# Clone the repository
git clone https://github.com/soyomarvaldezg/llm-chat.git
cd llm-chat

# Build
make build

# Or install to your PATH
make install

Option 2: Manual build

git clone https://github.com/soyomarvaldezg/llm-chat.git
cd llm-chat
go build -o llm-chat ./cmd/llm-chat

Setup

Configure at least one provider:

# For Ollama (local, free)
export OLLAMA_URL=http://localhost:11434

# For Groq (fast, free tier)
export GROQ_API_KEY=your-api-key

# For Together AI
export TOGETHER_API_KEY=your-api-key

# For SambaNova
export SAMBA_API_KEY=your-api-key

# For Google Gemini
export GEMINI_API_KEY=your-api-key

Basic Usage

# Start interactive chat
llm-chat

# Use specific provider
llm-chat -p groq

# Shell mode with piped input
cat myfile.py | llm-chat -s "explain this code"

# With assessment enabled
llm-chat --assess

📖 Usage Examples

Interactive Mode

# Basic chat
llm-chat

# With verbose metrics
llm-chat --verbose

# With prompt assessment
llm-chat --assess --auto-improve

# Using specific model
llm-chat --model qwen2.5-coder:7b-instruct-q4_K_M

Shell Mode (Piping)

# Explain code
cat main.go | llm-chat -s "explain this code"

# Generate commit message
git diff | llm-chat -s "write a concise commit message"

# Analyze data
cat data.csv | llm-chat -s "summarize this data"

# Debug errors
cat error.log | llm-chat -s "what's causing this error?"

# Code review
cat pr.diff | llm-chat -s "review this code for bugs and improvements"

# Generate documentation
cat *.go | llm-chat -s "create API documentation" -f markdown > docs.md

# Multiple providers
echo "Hello world" | llm-chat -p groq -s "translate to Spanish"
echo "Hola mundo" | llm-chat -p ollama -s "translate to French"

Advanced Examples

# Output formats
cat code.py | llm-chat -s "find bugs" -f json
cat code.py | llm-chat -s "document this" -f markdown > docs.md

# Temperature control
llm-chat --temperature 0.3  # More focused/deterministic
llm-chat --temperature 0.9  # More creative/diverse

# Token limits
llm-chat --max-tokens 500  # Short responses
llm-chat --max-tokens 4000 # Long responses

# Disable history for sensitive data
llm-chat --no-history

# Chain commands
cat file.txt | llm-chat -s "summarize" | llm-chat -s "translate to French"

# Verbose with metrics
cat large-file.txt | llm-chat -s "analyze" --verbose

🎯 Interactive Commands

Once in interactive mode, use these commands:

Basic Commands

/help - Show all commands
/clear - Clear the screen
/reset - Start fresh conversation
/exit or /quit - Exit

Provider & Model Management

/providers - List all available providers
/models - List models for current provider
/switch - Switch to different model

History Management

/history - Show current conversation
/saved - Show recent saved conversations
/search - Search through history
/export - Export current conversation
/stats - Show usage statistics

Prompt Engineering

/assess - Toggle prompt assessment
/guide - Show prompt engineering guide
/improve <prompt> - Get AI help improving a prompt

⚙️ Configuration

Environment Variables

Ollama

export OLLAMA_URL=http://localhost:11434        # Default
export OLLAMA_MODEL=llama3:8b-instruct-q4_K_M   # Your model

Together AI

export TOGETHER_API_KEY=your-api-key
export TOGETHER_MODEL=llama-70b-free            # Options: llama-70b, llama-70b-free, deepseek, qwen-72b

Groq

export GROQ_API_KEY=your-api-key
export GROQ_MODEL=llama-70b                     # Options: llama-70b, llama-8b, mixtral, gemma-7b

SambaNova

export SAMBA_API_KEY=your-api-key
export SAMBA_MODEL=llama-70b                    # Options: llama-70b, llama-8b, qwen-72b

Google Gemini

export GEMINI_API_KEY=your-api-key
export GEMINI_MODEL=flash-lite                  # Options: flash, flash-lite, pro

CLI Flags

-p, --provider string       LLM provider (default "ollama")
-v, --verbose              Show detailed metrics
-t, --temperature float    Temperature 0.0-2.0 (default 0.7)
-m, --max-tokens int       Maximum tokens (default 4000)
    --model string         Specific model to use
-s, --shell string         Shell mode with prompt
-f, --format string        Output format: text, json, markdown, raw
-a, --assess              Enable prompt assessment
    --auto-improve        Auto-offer prompt improvements
    --no-history          Don't save conversation
-h, --help                Show help

📊 Prompt Assessment

The built-in prompt assessment analyzes your prompts on 8 criteria:

Clarity - How clear and understandable
Specificity - How specific and detailed
Context - Background information provided
Structure - Organization and formatting
Constraints - Limitations specified
Output Format - Desired format specified
Role/Persona - Role definition
Examples - Examples provided

Usage

# Enable assessment
llm-chat --assess

# With auto-improvement
llm-chat --assess --auto-improve

# In chat, toggle with
/assess

# Improve a specific prompt
/improve explain python decorators

💾 History Management

All conversations are automatically saved to ~/.llm-chat/history.json

Commands

/saved        # View recent conversations
/search       # Search through history
/export       # Export current chat
/stats        # View statistics

Export Formats

/export
# Choose: markdown, json, or txt

Disable History

llm-chat --no-history  # For sensitive conversations

🏗️ Project Structure

llm-chat/
├── cmd/
│   └── llm-chat/
│       └── main.go              # Entry point
├── internal/
│   ├── providers/               # LLM provider implementations
│   │   ├── provider.go         # Interface
│   │   ├── ollama.go
│   │   ├── together.go
│   │   ├── groq.go
│   │   ├── samba.go
│   │   └── gemini.go
│   ├── registry/               # Provider registry
│   │   └── registry.go
│   ├── chat/                   # Chat session logic
│   │   ├── session.go
│   │   └── shell.go
│   ├── assessment/             # Prompt assessment
│   │   ├── analyzer.go
│   │   └── improver.go
│   ├── history/                # History management
│   │   └── manager.go
│   ├── ui/                     # Terminal UI
│   │   ├── display.go
│   │   └── markdown.go
│   └── config/                 # Configuration
│       └── config.go
├── pkg/
│   └── models/                 # Shared data models
│       └── message.go
├── Makefile
├── go.mod
└── README.md

🤝 Contributing

Contributions are welcome! To add a new provider:

Create internal/providers/yourprovider.go
Implement the Provider interface
Register in cmd/llm-chat/main.go
Add environment variables to README

📝 Tips & Tricks

Writing Better Prompts

Use /guide to see the full prompt engineering guide. Quick tips:

Be specific - "Explain Python decorators" → "Explain Python decorators with 3 code examples"
Add context - Include why you need it
Set constraints - "Keep under 200 words"
Specify format - "Format as a numbered list"
Define role - "As an expert Python developer..."
Use examples - Show what you want

Workflow Examples

Code Review Workflow

git diff | llm-chat -s "review for bugs, security, performance" > review.md

Documentation Generation

cat src/*.go | llm-chat -s "generate API docs" -f markdown > API.md

Data Analysis

cat data.csv | llm-chat -s "analyze and visualize key trends" --verbose

Learning Assistant

llm-chat --assess --auto-improve  # Interactive learning with feedback

🐛 Troubleshooting

Provider not available

# Check your setup
llm-chat

# You'll see which providers are configured
# Set the required environment variables

Ollama connection issues

# Check Ollama is running
curl http://localhost:11434

# Start Ollama
ollama serve

History file location

~/.llm-chat/history.json

📄 License

MIT License - See LICENSE file for details

🙏 Acknowledgments

Built with Cobra for CLI
Uses go-openai for OpenAI-compatible APIs
Integrates with Ollama for local models
Colors by fatih/color

⭐ Star History

If you find this useful, consider giving it a star on GitHub!

Happy chatting! 🤖✨

Name		Name	Last commit message	Last commit date
Latest commit History 5 Commits
internal		internal
pkg/models		pkg/models
.gitignore		.gitignore
Makefile		Makefile
README.md		README.md
go.mod		go.mod
go.sum		go.sum

soyomarvaldezg/llm-chat

Folders and files

Latest commit

History

Repository files navigation