DataPizza AI: the Italian open-source framework for controllable GenAI systems

In the expanding GenAI ecosystem, developers need tools that allow them to build reliable AI systems without losing transparency.

That’s exactly what DataPizza AI aims to provide — an Italian open-source framework built in Python for creating agents, multi-agent systems, and RAG pipelines with clarity and control.

---

Philosophy and Core Ideas

Its guiding principle is simple: Less abstraction, more control.

Instead of adding complexity, DataPizza AI relies on three core principles:

• API-first design — consistent, predictable interfaces for every component.
• Observable by design — native tracing with OpenTelemetry to understand exactly what happens under the hood.
• Provider agnostic — instantly swap between OpenAI, Google Gemini, Anthropic, Mistral, or Azure without touching your logic.

---

Key Features

• Multimodal processing: support for images, PDFs, audio, and DOCX files.
• Tool-based automation: expose any Python function as a callable tool using the @tool decorator.
• Structured JSON output: perfect for automation and decision-based workflows.
• Advanced tracing: ContextTracing tracks execution flow, latency, and token usage.
• Full RAG pipeline: ingestion, embedding, vector storage (Qdrant), and built-in rerankers.

---

Why It Matters

With DataPizza AI, teams can move from prototype to production seamlessly.

The framework keeps the codebase clean, observable, and easy to debug, while staying compatible with any major AI provider.

It’s a clear proof that Italian engineering can deliver world-class open-source AI frameworks.

---

Final Thoughts

DataPizza AI stands as a minimal yet powerful toolkit for developers building next-generation intelligent systems.

Less opacity, more insight, full control.