BetterAI documentation
Multiple AI models, one unified back end.
On this page
Where intelligence finds purpose.
BetterAI is a modular AI back end written in Python that puts OpenAI, Anthropic, Google Gemini and Groq behind the same interface. On top of it run FastAPI and Streamlit, which expose the platform in production.
Start here
With the API running, the interactive endpoint docs are at http://localhost:8000/docs.
Today the platform does the following:
- Document parsing: turns
txt,md,pdfanddocxfiles into structured data, following the caller’s schema. - Deep research: deep web search through Tavily, returned as markdown context.
- Vector store: ingestion, semantic search and deletion on Pinecone.
- Local embeddings: in-memory chunking and retrieval with FAISS.
- Image generation: the Da-Vinci service generates and edits images from text and visual references.
Project creator
Enzo SchitiniData Scientist & Data Analyst • AI Engineer @stefanini • SQL
In the AI projects I have built, I noticed a clear gap: there were no solid pillars to support a reliable AI backend. That is what motivated me to create an efficient, secure and scalable architecture designed for companies. My mission is to simplify the complex and deliver real structure to those who want to innovate with AI.
If the project is useful to you, let’s talk.
Explore by area
IntroductionInstallation, dependencies and license. From clone to first request.AgentsThe platform’s agents built on Agno: database and model gateway utilities, and AgnoUI.Streamlit ApplicationsAcquarello and Content Generator, the two applications that run on the platform.APIThe Web Service Network: the FastAPI service in production, its routes and the CURL compiler.ModulesThe platform’s AI services: agents, document parsing, deep research and images.Internal ServicesThe infrastructure underneath: tracing, databases, storage, vector store, embeddings, token cost and utilities.