fixed Build LLM integration layer for command interpretation

[Sahilbhatane]

Build LLM Integration Layer for Command Interpretation

Fixes #1

Summary

Implements the LLM integration layer for command interpretation, enabling Cortex to convert natural language requests into safe, validated bash commands using Claude and OpenAI GPT-4 APIs.

What Was Implemented

Core Features

• Multi-Provider Support: Seamless integration with both Claude (Anthropic) and OpenAI GPT-4 APIs
• Command Interpretation: Converts natural language input (e.g., "install docker with nvidia support") into executable bash command sequences
• Safety Validation: Built-in dangerous command filtering to prevent destructive operations
• Context-Aware Parsing: Optional system context parameter for more intelligent command generation
• Error Handling: Comprehensive error handling for API failures and invalid responses

Key Components

LLM/interpreter.py (Main Module)

• CommandInterpreter class with support for both API providers
• Intelligent prompt engineering with system-level instructions
• JSON-based command parsing with markdown code block support
• Command validation with dangerous pattern detection
• Context-aware command generation

LLM/test_interpreter.py (Test Suite)

• 23 comprehensive unit tests
• Mocked API calls for reliable testing
• Tests for validation, parsing, error handling, and both API providers
• 95%+ test coverage (exceeds the 80% requirement)

LLM/requirements.txt

• Minimal dependencies: openai>=1.0.0 and anthropic>=0.18.0

Requirements Met

Requirement	Status	Implementation
Support Claude API and OpenAI GPT-4	done	Both providers supported via APIProvider enum
Input: Natural language string	done	parse() method accepts user intent strings
Output: List of safe bash commands	done	Returns List[str] of validated commands
Prompt templates for different command types	done	System prompt with clear rules and examples
Error handling for API failures	done	Try-catch blocks with descriptive RuntimeError messages
Unit tests with 80%+ coverage	done	23 tests with 95%+ coverage

Security Features

Dangerous Pattern Detection:

• Filters rm -rf / commands
• Blocks disk formatting operations (mkfs., dd if=)
• Prevents fork bombs
• Can be disabled with validate=False parameter for advanced users

API Safety:

• Low temperature (0.3) for predictable outputs
• System prompt enforces safe command generation
• JSON-structured responses for reliable parsing

Usage Examples

from LLM.interpreter import CommandInterpreter

# Using OpenAI GPT-4
interpreter = CommandInterpreter(api_key="sk-...", provider="openai")
commands = interpreter.parse("install docker with nvidia support")
# Returns: ["sudo apt update", "sudo apt install -y docker.io", 
#           "sudo apt install -y nvidia-docker2", "sudo systemctl restart docker"]

# Using Claude
interpreter = CommandInterpreter(api_key="sk-ant-...", provider="claude")
commands = interpreter.parse("install python 3.11 and setup virtual environment")

# With system context
interpreter.parse_with_context(
    "optimize this server for machine learning",
    system_info={"os": "ubuntu", "version": "22.04", "gpu": "RTX 4090"}
)

[mikejmorgan-ai]

Outstanding implementation @Sahilbhatane!

✅ Both Claude and GPT-4 support
✅ Safety validation included
✅ 23 unit tests with 80%+ coverage
✅ Excellent documentation

Merging now. $200 bounty paid within 3 days.

Contact: mike@cortexlinux.com or Discord for payment details.

Thank you for this solid foundation! 🚀