Chat Command Usage¶

The cliver chat command provides an interactive interface for communicating with various large language models. This guide covers all the features and options available when using the chat functionality.

Basic Usage¶

To start a simple chat session with the default model:

cliver chat

This will open an interactive session using your configured default LLM provider.

You can also pass a query directly (CLIver routes bare text to chat automatically):

cliver "What is the capital of China?"

Selecting Different Models¶

DeepSeek Models¶

To chat with DeepSeek models:

# Use DeepSeek-R1, note that you need to configure the model first
cliver chat --model deepseek-r1

QWen3 Models¶

To chat with QWen3 coder model:

# Use Qwen3-Coder, note that you need to configure the model first
cliver chat --model qwen3-coder

Chat Configuration Options¶

Temperature Control and others¶

Control the creativity of the model's responses:

Without query specified, it starts an interactive session with the options specified as the default value.

# More creative responses (higher temperature)
cliver chat --model deepseek-r1 --temperature 0.9

# More deterministic responses (lower temperature)
cliver chat --model deepseek-r1 --temperature 0.2

# Set max tokens for response
cliver chat --max-tokens 1024

# Set top_p parameter for sampling
cliver chat --top-p 0.9

# Set frequency penalty
cliver chat --frequency-penalty 0.5

System Prompt¶

Set a system prompt to guide the model's behavior:

NOTE: the system message will be appended to the builtin system message if specified.

cliver chat --system-message "You are a helpful assistant that responds in a professional manner."

Advanced Chat Features¶

Using MCP Servers¶

As long as MCP servers are configured, all tools will be included by default.

You can filter the tools using --included-tools option:

cliver chat --included-tools "*time"

Using Skills¶

CLIver has an LLM-driven skill system. During a chat session, the LLM can discover and activate skills using the builtin skill tool. Skills are defined as SKILL.md files in .cliver/skills/ (project) or ~/.config/cliver/skills/ (global).

See Skills for details on creating and using skills.

File Integration¶

Work with files directly in the chat:

# Include a file in your message
cliver chat "Can you summarize this document?" --file /path/to/document.txt

# Process multiple files
cliver chat "Compare these two files" --file /path/to/file1.txt --file /path/to/file2.txt

Examples¶

Example 1: Professional Assistant Session¶

cliver chat \
  --model qwen3-coder \
  --system-message "You are a professional technical assistant. Provide concise, accurate answers with examples when possible." \
  --temperature 0.3

Example 2: Creative Writing Assistant¶

cliver chat \
  --model deepseek-r1 \
  --system-message "Help me brainstorm creative writing ideas. Be imaginative and provide detailed suggestions." \
  --temperature 0.8

Example 3: Code Review Session¶

cliver chat \
  --model qwen3-coder \
  --system-message "Review this code for best practices, security issues, and potential improvements." \
  --file /path/to/code.py

Next Steps¶

After mastering the chat command, learn about Skills for specialized task activation, Memory & Identity for agent personalization, or check out Workflows to automate complex multi-step operations.