- What it is: A robust AI framework demonstrating high-precision persona control and model-agnostic architecture.
- Core Skills: Advanced Python, LLM-Ops, System Integration, and Adapter-Pattern Architecture.
- Seniority: Built by a PLM Expert with 15+ years of Enterprise experience, bridging the gap between legacy systems and AI.
Giulia is a technical reference implementation for transitioning Enterprise PLM expertise into AI Engineering. This project solves the "Vendor Lock-in" problem by using a decoupled architecture, allowing businesses to switch between AI providers (Gemini, OpenAI, Llama) without rewriting their core logic.
Business Relevance: Companies need AI that follows strict corporate guidelines. Giulia proves that LLMs can be steered with 100% adherence to bilingual protocols and length constraints.
- System Architecture: Provider & Adapter Patterns for model-agnosticism.
- Data Persistence: Normalized, vendor-neutral JSON state management.
- AI Steerability: Precise control of tone, language, and logic through structured prompting.
- Modern Tooling: High-performance dependency management via
uv.
- Decoupled AI Core: Implementation of an
AIModelInterface. Switch LLM providers via CLI without touching the orchestrator logic. - Universal History Manager: Automatically normalizes vendor-specific data (e.g., Gemini Content Objects) into a standard schema for long-term auditability.
- Hierarchical Prompting: Managed asset structure (
core,tasks,library) to optimize LLM performance while keeping code clean.
Instead of a generic chatbot, Giulia acts as a High-Context Executive Assistant:
- Steerability Proof: Demonstrates how to maintain a sophisticated, charismatic, yet strictly professional tone over long conversations.
- Operational Constraints: Hard-enforced 60-word limits and bilingual automatic detection.
- Enterprise Identity: She is programmed to recognize and augment the user's specific background in PLM and System Architecture.
- Install uv:
curl -LsSf https://astral.sh/uv/install.sh | sh - Setup Environment:
git clone [https://github.com/RaptileBytez/giulia-ai.git](https://github.com/RaptileBytez/giulia-ai.git) cd giulia-ai uv sync echo "GEMINI_API_KEY=your_key_here" > .env
uv run main.py # Launch Standard Session
uv run main.py --mock # Developer Test Mode (Zero Cost)├── data/
│ ├── chat_history/ # Session JSON files (Git-ignored)
│ └── logs/ # Application and API logs (Git-ignored)
├── prompts/ # New Hierarchical Structure
│ ├── core/ # Identity (Giulia persona, wrappers)
│ ├── tasks/ # Active production prompts (categorized)
│ └── library/ # Research & model-optimized assets (OpenAI, etc.)
├── utils/
│ ├── ai/ # AI Core Subpackage
│ │ ├── __init__.py # Central exports
│ │ ├── model_interface.py # Abstract base classes
│ │ ├── model_provider.py # Gemini & Mock implementations
│ │ ├── prompt_loader.py # Path-based templating engine
│ │ └── history_manager.py # Persistence logic
│ └── logger.py # Unified logging system
├── chatbot.py # Refactored Orchestrator using Interface
└── main.py # Entry point with argparse support
Behind the creative label Raptile Bytez is a seasoned Enterprise Systems Expert and PLM Consultant with 15+ years of experience in the Oracle Agile e6 ecosystem.
With a degree in Business Information Systems (2009), I focus on bringing "Industrial-Grade" stability to the world of AI Engineering. I build systems that don't just "talk," but integrate into complex professional workflows.
- Enterprise: Oracle Agile e6, PLM Architecture, System Integration.
- AI & Automation: LLM Orchestration (Gemini, OpenAI), Prompt Engineering, Python.
- Philosophy: Clean Code, Modular Design, and Stateless Architecture.
I am currently building my network in the AI space. Whether you are an AI enthusiast, a fellow developer, or a recruiter looking for a consultant with both business logic and AI-coding skills — let's connect!
MIT License - See LICENSE for details.