An intelligent Python code generation system featuring two distinct implementation approaches, each demonstrating different architectural patterns for AI-driven development workflows.

This project showcases two complementary approaches to AI-powered code generation:

A graph-based workflow system built with Apache Burr that follows a predefined state machine pattern. The workflow is explicitly structured with clear transitions and deterministic paths through code generation, validation, and improvement cycles.

Key Characteristics:

• Predefined workflow graph with explicit state transitions
• Deterministic flow through generation → validation → improvement
• Structured retry logic with configurable thresholds
• Comprehensive quality gates enforced at each step
• Workflow visualisation and state tracking

A model-driven agent system built with Strands SDK where the AI foundation model dynamically determines the workflow flow. The agent autonomously decides when to generate code, validate results, and iterate based on contextual understanding.

Key Characteristics:

• AI-driven decision making with dynamic workflow paths
• Contextual flow control determined by the foundation model
• Adaptive iteration strategies based on real-time assessment
• Intelligent tool orchestration guided by AI reasoning
• Emergent workflow patterns that adapt to task complexity

# System Requirements
# - Python >= 3.12
# - AWS Account with Bedrock access
# - Claude 3.5 Sonnet model enabled

# Project Setup
git clone <repository-url>
cd ai-python-coding-agent

# Initialize project with uv
uv init

# Configure AWS credentials
aws configure sso
aws sso login

# Install dependencies
uv sync

# Activate virtual environment
source .venv/bin/activate

cd 01_ai_workflow

# Run predefined tasks
uv run 01_ai_workflow.py simple    # Basic factorial function
uv run 01_ai_workflow.py moderate  # CSV analysis with statistics
uv run 01_ai_workflow.py complex   # Multi-threaded web scraper
uv run 01_ai_workflow.py all       # All tasks sequentially

# Test the workflow framework
uv run 00_test.py

cd 02_ai_agent

# Run interactive agent sessions
uv run ai_agent_strands.py

# Test with predefined scenarios
uv run 02_ai_agent_test.py

• Python 3.12+ - Modern Python with latest features
• AWS Bedrock - Claude 3.5 Sonnet foundation model
• Instructor - Structured output from language models
• Pydantic - Data validation and type safety

• Apache Burr (01_ai_workflow) - Workflow orchestration and state management
• Strands SDK (02_ai_agent) - Agent framework and tool orchestration

• uv - Fast Python package manager
• pytest - Testing framework
• boto3 - AWS SDK for Python

Each implementation includes comprehensive documentation:

Detailed guide covering:

• Apache Burr workflow architecture
• State machine design patterns
• Quality threshold configuration
• Workflow visualisation
• Performance optimisation
• Custom task creation

Complete reference including:

• Strands SDK agent architecture
• AI-driven tool orchestration
• Iterative improvement patterns
• Session management
• Agent reasoning and decision making
• Advanced configuration options

# Configure AWS SSO (recommended)
aws configure sso --profile your-profile-name
aws sso login --profile your-profile-name

# Set environment variables
export AWS_PROFILE=your-profile-name
export AWS_REGION=us-east-1  # Your preferred region

Create a .env file in the project root:

# AWS Configuration
AWS_REGION=us-east-1
AWS_PROFILE=your-profile-name

# Model Configuration
MODEL=anthropic.claude-3-5-sonnet-20241022-v2:0

# Optional: Workflow Configuration
MAX_RETRIES=5
WARNING_THRESHOLD=5
ENABLE_CODE_EXECUTION=true

• Standardised processes requiring consistent quality gates
• Regulatory compliance with auditable workflow steps
• Team environments where workflow transparency is crucial
• Production systems needing predictable execution paths
• Performance-critical applications requiring optimised flows

• Research and exploration with unknown problem complexity
• Dynamic requirements that change during execution
• Creative problem solving requiring adaptive strategies
• Prototype development with evolving specifications
• Complex reasoning tasks benefiting from AI autonomy

Note: Agent implementation typically uses more tokens due to reasoning overhead but provides more adaptive solutions.

1. AWS Authentication

   # Verify AWS configuration
   aws sts get-caller-identity
   aws bedrock list-foundation-models --region us-east-1

2. Python Environment

   # Check Python version
   python --version  # Should be 3.12+
   
   # Verify dependencies
   uv tree

3. Model Access

   # Test Bedrock access
   aws bedrock get-foundation-model \
     --model-id anthropic.claude-3-5-sonnet-20241022-v2:0 \
     --region us-east-1

• Workflow Issues: See 01_ai_workflow troubleshooting
• Agent Issues: See 02_ai_agent troubleshooting
• General Setup: Check AWS Bedrock service availability in your region

• Workflow contributions: Focus on graph structure, state management, and deterministic flows
• Agent contributions: Enhance reasoning capabilities, tool orchestration, and adaptive behaviours
• Shared components: Improve AWS integration, model interactions, and quality validation

Choose your approach: Structured workflows or AI-driven agents - both paths lead to intelligent code generation! 🚀