Code Generation Guide

Generate code with Codestral specialist model.

Overview

Zorora includes dedicated code generation capabilities using the Codestral specialist model. This allows you to generate code, refactor existing code, and implement algorithms with high-quality output.

Using Code Generation

Terminal Interface

Automatic Detection:

[1] ⚙ > Write a Python function to validate email addresses

The system automatically detects code generation intent and routes to Codestral.

Force Code Generation:

[2] ⚙ > /code create a REST API endpoint for user authentication

Code Generation Examples

Functions:

/code write a function to parse JSON files with error handling

Classes:

/code create a User class with authentication methods

Scripts:

/code write a script to backup files to S3

Refactoring:

/code refactor this function to use async/await

Codestral Models

Local Model

Model: qwen/qwen3-vl-4b

Characteristics:

• Fast generation (10-30 seconds)
• Basic code quality
• No API costs
• Privacy-preserving

When to use:

• Quick code snippets
• Simple functions
• Prototyping

HuggingFace Model

Model: Qwen2.5-Coder-32B-Instruct

Characteristics:

• High quality code (60-90 seconds)
• Better understanding of complex requirements
• Requires HF token
• API costs apply

When to use:

• Complex implementations
• Production code
• Algorithm implementation

Code Generation Workflow

Step 1: Request Code

/code write a function to validate email addresses using regex

Step 2: Review Generated Code

Codestral generates:

• Complete, executable code
• Comments and documentation
• Error handling
• Example usage

Step 3: Save Code (Optional)

> /save email_validator.py
Saved to: email_validator.py

Best Practices

Writing Code Prompts

Be Specific:

• “Write a Python function to validate email addresses using regex” ✓
• “Email validator” ✗

Include Requirements:

• “Create a REST API endpoint with authentication and rate limiting” ✓
• “API endpoint” ✗

Specify Language/Framework:

• “Write a Flask route for user registration” ✓
• “User registration” ✗

Code Quality

Review Generated Code:

• Check for errors
• Verify logic
• Test functionality
• Review security

Iterate:

• Refine prompts for better results
• Ask for improvements
• Request specific patterns

Saving Code

Use Descriptive Filenames:

/save email_validator.py
/save user_auth_api.py
/save data_processor.py

Organize Files:

• Save to appropriate directories
• Use version control
• Document purpose

Advanced Usage

Code Refactoring

/code refactor this function to use async/await and add error handling

Algorithm Implementation

/code implement quicksort algorithm in Python with detailed comments

Testing Code

/code write unit tests for the email validator function

Documentation

/code add docstrings and type hints to this function

Integration with Development Workflow

Using /develop

For multi-file changes, use /develop:

/develop Add user authentication to my Flask app

This workflow:

1. Explores codebase
2. Plans changes
3. Gets approval
4. Executes with Codestral
5. Lints and validates

Combining Commands

# Research then code
/search latest Python async patterns
/code implement async pattern from research

# Code then develop
/code create user model
/develop add authentication using this model

Troubleshooting

Code Generation Fails

Problem: Codestral model not responding

Solution:

• Check model endpoint is configured
• Verify API key (for HF endpoint)
• Check network connectivity
• Review model availability

Low Quality Code

Problem: Generated code is poor quality

Solution:

• Use HF 32B model for better quality
• Be more specific in prompts
• Break down complex requests
• Iterate and refine

Code Not Saving

Problem: /save command not working

Solution:

• Check file permissions
• Verify directory exists
• Check for filename conflicts
• Review error messages

See Also

• Development Workflow - Multi-step development
• Terminal REPL - Command-line interface
• Slash Commands - Command reference
• Configuration - Configure Codestral model