Open Source
Datapizza AI
Build reliable GenAI solutions
Datapizza AI provides clear interfaces and predictable behavior for agents and RAG. End-to-end visibility and reliable orchestration keep engineers in control from PoC to scale
Created by the AI Engineering team behind Datapizza.
Powering +50 GenAI solutions in production.
Built by Engineers, trusted by Engineers
"Well-built and designed from scratch and it shows. The OpenTelemetry integration is invaluable for observability, and the ability to create custom modules is fantastic."
Enterprise
Robert Timis, Freelance
"The docs are well written, and the modular pipelines are impressive. It's a solid framework pragmatic and clearly designed by people who build real things."
Inferentia
Simone Rizzo, AI Engineer
"Datapizza AI is a compelling, easy-to-use framework that stands out among dev-friendly AI agent libraries."
Wurth Italia
AI & Data Science Team
"Your caching system and RAG utilities are really well explained and easy to use. Great job!"
Bloomberg (ex)
Caleb Gucciardi, Software Engineer
"If I had to design a framework from scratch, I’d build abstractions exactly like this. It’s clean, composable, and production-oriented"
LlamaIndex
Clelia Astra Bertelli, Python Open Source Engineer
"Migration from other frameworks is quick, and the overall developer experience is excellent."
Wurth Italia
AI & Data Science Team
How to use it
Start building reliable agents
Step 1
Check your Python version
Make sure you have Python installed on your system.
Step 2
Install Datpizza AI framework
pip install -U datapizza-aiStep 3
Define your agents
Start defining your agents now
RAG
From ingestion to evaluation
Configurable ingestion chain
Document parsing, splitting, embedding (with efficient batching), and storage in a vector database (e.g., Qdrant).
Built-in reranking
Add a reranker (e.g., Cohere) to boost relevance before generation.
Chunk transformation
Enrich retrieved snippets (titles, URLs, metadata) to provide higher-quality context to the generator.
Monitoring
OpenTelemetry tracing & rich logging
Standards-based instrumentation
Fully compatible with existing OpenTelemetry tooling.
Client I/O tracing
Optional toggle to log inputs, outputs, and in-memory context passed to model calls.
Custom spans
Trace fine-grained phases and sub-steps to pinpoint bottlenecks.
Easy exporters
Ship traces to Zipkin or any OTLP-compatible backend.
Proven in production
Built With It, Shipped With It
These are real solutions developed with the framework and deployed in enterprise contexts to power daily operations, automate processes, and deliver insights faster.
Agent
Booking assistant
Agent managing support queries across WhatsApp, with full customer context.
Agent
Speech-to-SQL
Enables natural language queries over databases. Manager-friendly, no SQL knowledge required.
Agent
Compliance
Automatically reviews policy documents and flags inconsistencies based on company standards.
Agent
Knowledge Assistant
Enterprise-grade assistant answering on complex documentation via structured retrieval.
Less abstraction, more control
A framework that keeps your agents predictable, your debugging fast, and your code trusted in production
API-first
Clear lifecycle hooks, sync/async and streaming out of the box.
Composable
Reusable blocks, declarative configuration, easy overrides
Observable
Tracing, logging, and actionable metrics.
Vendor-Agnostic
Swap models, rerankers, and vector DBs without rewiring business logic.
"A framework made to be easy to learn, easy to maintain and ready for production"
Datapizza Engineering Team
Don’t miss what’s next
We’re building fast. Get notified when we ship something cool.
Frequently Asked Questions
Built by Datapizza, the AI transformation company
This framework is developed and maintained by our AI Engineering team. We are a company at the forefront of Generative AI, AI Engineering and production-grade Gen AI solutions for enterprise.