Skip to content

Command Line Interface (CLI)

QuantaLogic provides a powerful command-line interface for running AI agents and executing tasks directly from your terminal.

Installation

You can install QuantaLogic using any of these methods:

Bash
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
# Via pip
pip install quantalogic

# Via pipx (recommended for CLI tools)
pipx install quantalogic

# From source
git clone https://github.com/quantalogic/quantalogic.git
cd quantalogic
python -m venv .venv
source ./venv/bin/activate 
poetry install

Basic Usage

Bash
1
quantalogic [OPTIONS] COMMAND [ARGS]...

Global Options

Option Description Default
--version Show version information -
--model-name Specify the model (litellm format, e.g., "openrouter/deepseek/deepseek-chat") -
--log Set logging level (info/debug/warning) info
--verbose Enable verbose output False
--mode Agent mode (code/search/full) code
--vision-model-name Specify the vision model (litellm format, e.g., "openrouter/A/gpt-4o-mini") -
--max-iterations Maximum iterations for task solving 30
--help Show help message and exit -

Available Modes

  • code: Full coding capabilities with advanced reasoning
  • basic: Simple task execution without additional features
  • interpreter: Interactive REPL mode for dynamic interaction
  • full: All features enabled including advanced tools
  • code-basic: Basic coding features without advanced capabilities
  • search: Web search capabilities for information gathering
  • search-full: Enhanced search features with comprehensive analysis

Commands

task

Execute a task with the QuantaLogic AI Assistant:

Bash
1
quantalogic task [OPTIONS] [TASK]

Task-Specific Options

Option Description Default
--file Path to task file -
--model-name Specify the model (litellm format) -
--verbose Enable verbose output False
--mode Agent mode (code/search/full) -
--log Set logging level (info/debug/warning) -
--vision-model-name Specify the vision model (litellm format) -
--max-iterations Maximum iterations for task solving 30
--no-stream Disable streaming output False
--help Show help message and exit -

Examples

Basic Task Execution

Bash
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
# Simple code generation
quantalogic --mode code "Create a Python function that calculates factorial"

# Using a task file
quantalogic task --file path/to/task.txt

# Interactive mode with custom model
quantalogic --mode interpreter --model-name "openai/gpt-4" "Explain quantum computing"

# Debugging with verbose output and no streaming
quantalogic task --log debug --verbose --no-stream "Debug this Python script"

Best Practices

  1. Start Simple:
  2. Begin with --mode basic for straightforward tasks

  3. Gradually increase complexity as needed

  4. Debugging:

  5. Use --log debug for troubleshooting
  6. Enable --verbose for detailed execution information

  7. Disable streaming with --no-stream when needed for clearer output

  8. Model Selection:

  9. Choose models based on task complexity

  10. Consider using specialized models for specific tasks (e.g., vision models for image analysis)

  11. Task Management:

  12. Use task files for complex or repetitive tasks
  13. Adjust --max-iterations based on task complexity
  14. Break down complex tasks into smaller subtasks

Security Considerations

  • API keys should be set via environment variables (litellm will use env vars by default)
  • Code execution is sandboxed by default for security
  • Always review generated code before execution
  • Use appropriate permissions when working with file system operations

Error Handling

  • The CLI provides detailed error messages for common issues
  • Check logs with --log debug for troubleshooting
  • Use --verbose for additional context when errors occur

Requirements

  • Python 3.12+
  • Docker (optional, required for code execution tools)
  • Internet connection for model API access