Cerebras Code Generator Skill

Description

Generates code using Z.ai glm-4.6 model for fast test and code generation. Use this for rapid iteration when generating tests, boilerplate, and repetitive code patterns.

When to Use

• ✅ Test generation - unit tests, integration tests, test fixtures
• ✅ Boilerplate code - CRUD operations, API endpoints, data models
• ✅ Repetitive patterns - similar components, migration scripts
• ✅ Bulk file creation - generating multiple similar files quickly
• ❌ NOT for complex architecture, security-critical code, or nuanced logic

Configuration

# Required environment variables
export ZAI_API_KEY="your-api-key"  # or CEREBRAS_API_KEY for legacy
export ZAI_MODEL="glm-4.6"  # Fast, cost-effective model

# Optional settings
export CEREBRAS_BASE_URL="https://api.cerebras.ai/v1"
export CONTEXT_DB_PATH="./.claude/skills/cfn-cerebras-code-generator/contexts.db"

Usage

# Basic code generation
./generate-code.sh \
  --file-path "/path/to/file.ext" \
  --prompt "Create a REST API endpoint" \
  --context-files "src/models.py,src/utils.py"

# With explicit model
./generate-code.sh \
  --model "llama-3.1-70b" \
  --file-path "/path/to/file.py" \
  --prompt "Implement authentication middleware"

Implementation Details

Context Tracking

• Stores generation history in SQLite database
• Tracks what worked and what didn't
• Maintains conversation context
• Provides examples of successful patterns

OpenAI Compatibility

• Uses OpenAI-compatible request/response format
• Supports streaming responses
• Handles token limits and rate limiting
• Automatic retry logic

Features

• ✅ Visual diff generation
• ✅ Context file inclusion
• ✅ Error handling and validation
• ✅ Generation history tracking
• ✅ Success pattern learning
• ✅ Multiple model support