Datapizza AI is a Python-based framework to build Gen AI solutions with multi-provider support, document processing, and observable tooling. It emphasizes API-first design, composable components, and vendor-agnostic integration.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Datapizza AI is a Python framework for building Gen AI solutions. It targets speed and reliability, offering API-first design, vendor-agnostic model support, and built-in observability features. It provides tools for multi-provider access (OpenAI, Google Gemini, Anthropic, Mistral, Azure) and integrations for documents, web search, and custom tools.
How it works
The project presents modular components with reusable blocks and declarative configuration. It includes document processing (PDF, DOCX, images), smart chunking, embedding, and a reranking option. It supports observability through OpenTelemetry tracing and client I/O tracing. It emphasizes swapping providers without changing business logic and a migration-friendly interface.
Getting started
Install core framework:
pip install datapizza-ai
Install provider-specific clients (optional):
pip install datapizza-ai-clients-openai
pip install datapizza-ai-clients-google
pip install datapizza-ai-clients-anthropic
Example client invocation:
from datapizza.clients.openai import OpenAIClient
client = OpenAIClient(api_key="YOUR_API_KEY")
result = client.invoke("Hi, how are u?")
print(result.text)
Example agent usage (from README):
from datapizza.agents import Agent
from datapizza.clients.openai import OpenAIClient
from datapizza.tools import tool
@tool
def get_weather(city: str) -> str:
return f"The weather in {city} is sunny"
client = OpenAIClient(api_key="YOUR_API_KEY")
agent = Agent(name="assistant", client=client, tools=[get_weather])
response = agent.run("What is the weather in Rome?")
Recent releases
Latest releases include:
- v0.1.0 (2026-03-13): Added structured output support for agents, agent hooks, agent handoff, runtime memory, and description support.
- v0.0.9 (2025-11-04): Integrated Model Context Protocol (MCP) tool call and parser enhancements for Azure and Docling parsers.
- v0.0.7 (2025-10-29): Docling parser expanded to all supported files with flexible OCR config; Bedrock async support via aioboto3; new tools like web_fetch, FileSystem, SQLD.
- v0.0.2 (2025-10-13): First release.
Traction
GitHub stars: 2231 (no daily/weekly changes provided in FACTS).
Behind the repo
Not provided in the facts.
Caveats
License: MIT. Created 2025-09-01. Last push 2026-05-19. Language: Python. Open issues: 29. Forks: 139. Repo topics include agent, ai, genai, llm, python.