Datapizza AI

Build reliable Gen AI solutions without overhead

[Datapizza AI @datapizza_ai]

Datapizza AI v0.1.0 is live 🔥

Agents are now a first-class component — rebuilt from the ground up: new runner architecture, handoffs, hooks, structured outputs, runtime memory, agent descriptions, and agent-as-tool interoperability.

Also shipping:

→ Persistent MCP sessions + Anthropic client memory bug fix
→ WatsonX, Mistral embedder, and latest Gemini models support
→ Redis cache security fix (now JSON-based)
→ Bug fixes, reliability improvements, expanded test coverage

Docs: https://t.co/RJV7imtT29
GITHUB: https://t.co/dojOf1sOFq

[Datapizza AI @datapizza_ai]

We ran experiments to answer: Is Contextual Retrieval still worth it in 2026? Tested on 3 datasets, with/without rerankers (Cohere v4 Pro, Qwen3 8B).

Here's what we found → https://t.co/qYftRIbCmv

[Datapizza AI @datapizza_ai]

We’re rolling out a new YouTube playlist with step-by-step tutorials for our open-source framework — and it’s just getting started. 🚀

We’re releasing videos gradually, covering everything from setup to advanced workflows, so you can dive deeper into the architecture as the series evolves.

👉 Start here: https://t.co/gH0p8JhdV5

#opensource #developers #AI #tutorials

[Datapizza AI @datapizza_ai]

Here are the latest updates from our R&D Lab 🚀

Right now we’re exploring advanced RAG techniques — and for proper evaluation, we need datasets that reflect real-world difficulty.

We tested several public datasets:

📍LegalBench

📍MultiHop-RAG

📍LoCoMo

These have been super helpful starting points for evaluation.

As we applied them to our specific use cases, we realized we needed something more tailored to the GenAI RAG challenges we're focusing on — particularly around domain-specific knowledge and reasoning chains that match our clients' real-world scenarios.

So we built a framework to generate custom evaluation datasets that fit our needs.

We now have two internal domain-heavy evaluation datasets + a public one based on the DnD SRD 5.2.1 that we're sharing with the community.

This is just an initial step, but we're excited about where it's headed. We broke down our approach here:

🔗 Blog post https://t.co/OOYnjntpuW
🔗 GitHub repo https://t.co/sclO2dH4LZ
🔗 Dataset on Hugging Face https://t.co/SK2v9t8tMW

Would love to hear your thoughts, feedback, or ideas on how to improve this!
#RAG #dataset #evaluation #GenAI #framework

[Datapizza AI @datapizza_ai]

🚀 Release v0.0.9 is out!

Here’s what’s new in our open-source framework 👇

New features:

- MCP SUPPORT

Fix:

- Added metadata args on parse method (azure and docling),
- Enhanced OpenAIClient init arguments,

DOCS: https://t.co/gcO9M4Lf5t

GITHUB: https://t.co/dojOf1sOFq

#datapizzaai #opensource #MCP

[Datapizza AI @datapizza_ai]

🚀 Release v0.0.7 is out!

Here’s what’s new in our open-source framework 👇

🔧 Features
- Docling parser: extended to all supported files + flexible OCR configuration

- Bedrock client: now with async support via aioboto3 (and full test coverage)

- New tools
→ web_fetch
→ FileSystem
→ SQLDatabase

- Milvus vectorstore added

Agents: smoother usage & improved UX

🧩 Fixes
- https://t.co/6d0ryt0hCw: accepts list of file paths

- Bedrock async methods: safer error handling

- Node splitter: handles empty cases

- Tool converter & EasyOCR dependency fixes

- Agents: clear user input after first step

- OpenAI client compatibility update (>=2)

💡 Focus of this release: performance, async readiness, and better developer experience.

👉 Full changelog: https://t.co/dojOf1sOFq

#datapizzaai #framework #opensource

[Datapizza AI @datapizza_ai]

Our AI framework is live on Product Hunt! 🚀

A few days ago we launched our open-source framework, Datapizza AI.

And starting today, you can also find it on Product Hunt!

For those who haven’t heard of it yet, Product Hunt is a global showcase for new tech launches: every day users discover, vote for, and discuss new products in the comments.

If you’d like to learn more about our framework, check out our Product Hunt page! Here’s the link 👉 https://t.co/CrFh4Lv5VW

And if you want to support us, give us your feedback! 🙏

#datapizzaai #OpenSource #madeinitaly

[Alessandro Risaro @risooo2]

When people get what Datapizza AI is about without us explaining — that’s the best part.

[Datapizza AI @datapizza_ai]

@sisco404 crazy

[Datapizza AI @datapizza_ai]

Datapizza AI 🤝 Italy

[Pierpaolo D’Odorico 🇮🇹 @pierdd_]

From 3 people in a small room to a 60+ team building AI systems for some of the biggest companies in Italy.

And today - Datapizza AI is now Open Source. 🇮🇹

[Datapizza AI @datapizza_ai]

Today marks the birth of Datapizza AI: the open-source framework for GenAI 🌏

100% Open Source. 100% Made in Italy.

But what exactly is Datapizza AI?

It’s our framework for building multi-agent AI systems and production-ready RAG solutions — and we’ve decided to release it open source for everyone.

The project starts from a simple idea.

We want to give developers and companies full control over their own AI solutions.

No tech lock-in. No paywalls. No “slide deck” fluff.

We want to use Italy as a starting point to make our voice heard in the European Generative AI landscape.

Inside the repo, you’ll find:

📌 Ultra-fast quick start guide

📌 Full RAG system examples

📌 Templates for collaborative AI agents

Leave us a star and tell us what you think!

Any feedback is welcome — it’ll help us understand how to make it even better. 🙏

Try and Star on GitHub - https://t.co/F2u9YcPdrP

#AI #Italy #aimadeinitaly #madeinitaly

[Datapizza AI @datapizza_ai]

Our main mission has always been to make Italy a leading player in tech 🇮🇹

And on Monday, October 13th, we’ll announce something big that will take us even further in this direction 🙌

Take a look here to stay updated 👉 https://t.co/IE6erxrtYv