chat-agent

Chat-Agent CLI

A command-line interface for interacting with LLM agents, featuring flexible model configurations and MCP support.

Features

• Multi-provider support: Configure different LLM providers (DeepSeek, OpenAI, etc.)
• Preset chat configurations: Define reusable chat presets with custom system prompts
• MCP integration: Connect to Model Context Protocol for extended capabilities
• Simple configuration: YAML-based configuration with sensible defaults

Quick Start

1. Create configuration file at ~/.chat-agent/config.yml:

# Model provider configuration
providers:
  deepseek:
    type: deepseek
    base_url: https://api.deepseek.com
    api_key: your-api-key-here

# Model configuration
models:
  deepseek-chat:
    provider: deepseek
    model: deepseek-chat
    temperature: 0.8

# Chat preset configuration
chats:
  default:
    model: deepseek-chat
    desc: "A friendly assistant"
    system: "You are a helpful assistant."

1. Start a chat session:

chat-agent --chat default

Usage

# Basic usage with default chat preset
chat-agent

# Specify a chat preset
chat-agent --chat default

# Custom welcome message
chat-agent --welcome "Hello! How can I help you today?"

# Enable debug mode
chat-agent --debug

# Specify custom config file
chat-agent --config /path/to/config.yml

# One-time task (non-interactive)
chat-agent --once "List files in current directory"

# Show help
chat-agent --help

# Show version
chat-agent --version

Interactive Commands

When in chat mode, you can use the following commands:

• /help or /h - Show help message
• /history or /i - Get conversation history
• /clear or /c - Clear conversation context
• /tools or /l - List loaded tools
• /t cmd - Execute local command (e.g., /t ls -la)
• /exit or /q - Exit program

Building from Source

Prerequisites

• Go 1.25 or higher
• Git

Build Commands

# Build for current platform
make build

# Build for all platforms (Linux, macOS, Windows)
make build-all

# Build for Linux only (x86_64 and ARM64)
make build-linux

# Build for macOS only (x86_64 and ARM64)
make build-darwin

# Create release archives
make release

# Run tests
make test

# Run linter
make lint

# Clean build artifacts
make clean

# Install to /usr/local/bin
make install

Cross-Compilation

The Makefile supports cross-compilation for:

• Linux: x86_64 (amd64), ARM64
• macOS: x86_64 (amd64), ARM64 (Apple Silicon)
• Windows: x86_64 (amd64)

Release Process

Creating a Release

1. Update version using the release script:

   ./scripts/release.sh
   

2. Or manually create a git tag:

   git tag -a v1.0.0 -m "Release v1.0.0"
   git push origin v1.0.0
   

3. GitHub Actions will automatically:

   • Build binaries for all platforms
   • Create release archives (.tar.gz for Unix, .zip for Windows)
   • Publish to GitHub Releases

Version Management

• Check current version: ./scripts/version.sh
• View available binaries: ls -la dist/

Configuration

Providers

Configure LLM providers with API keys and base URLs:

providers:
  deepseek:
    type: deepseek
    base_url: https://api.deepseek.com
    api_key: sk-your-api-key

Models

Define model configurations with temperature and other parameters:

models:
  deepseek-chat:
    provider: deepseek
    model: deepseek-chat
    temperature: 0.8
  deepseek-reasoner:
    provider: deepseek
    model: deepseek-reasoner
    temperature: 0.8

Chat Presets

Create reusable chat configurations with Go template support:

chats:
  default:
    model: deepseek-chat
    desc: "A friendly assistant"
    system: "You are a helpful assistant."
  reasoner:
    model: deepseek-reasoner
    desc: "Reasoning-focused assistant"
    system: "You are a reasoning assistant. Think step by step."
  template-example:
    model: deepseek-chat
    desc: "Assistant with template variables"
    system: |
      You are a helpful assistant.
      
      Current working directory: {{.Cwd}}
      Today's date: {{.Date}}
      Current time: {{.Now.Format "2006-01-02 15:04:05"}}
      
      Please help the user with tasks in the current directory.

Available template variables:

• {{.Cwd}} - Current working directory
• {{.Date}} - Today's date in YYYY-MM-DD format
• {{.Now}} - Current time (time.Time object, can be formatted)
  • Example: {{.Now.Format "2006-01-02 15:04:05"}}
• {{.User}} - Current username
• {{.Home}} - User's home directory
• {{env "VAR_NAME"}} - Access environment variables

MCP Servers

Integrate with Model Context Protocol servers:

mcp_servers:
  web_search:
    type: sse
    url: "https://your-mcp-host"

# Add MCP to chat presets
chats:
  assistant-with-search:
    model: deepseek-chat
    system: "You are a helpful assistant with web search capabilities."
    mcp_servers:
      - web_search