Skip to main content
6 richieste/minuto per IP - gratis, no login richiesto.
UsernameLe risposte AI possono contenere errori - verifica sempre le fonti citate. I contenuti si basano sulla knowledge base del sito (articoli e podcast KB).

Example Questions

Write a question here below to start the chat with the knowledge base.

0/1000

Come utilizzare Ask Federico - Chat RAG sulla Knowledge Base

Write a question or choose an example

Enter a free question on our knowledge base (articles and podcasts) or click on one of the example questions to get started quickly.

System RAG retrieves relevant sources

The question is compared through vector search with the indexed knowledge base (blog articles + podcast transcripts), retrieving similar steps.

Ella generates a grounded response

Model Ollama on-prem (llama3.1:8b) composes responses based solely on retrieved sources, reducing the risk of hallucinations compared to a generic chatbot.

Check cited sources

Expand the "Cited Sources" section under each response to view original extracts, similarity percentages, and related articles with direct links.

Suggerimenti

  • Ask specific questions (e.g. "how does X work in Angular 21") instead of general ones: retrieval works better with targeted queries.
  • Check sources cited before considering a response definitive: RAG reduces hallucinations but not entirely eliminates them.
  • Chat history remains only in your browser (local storage): use "Clear History" to start over without leaving a trace.

Domande frequenti

What is a RAG (Retrieval-Augmented Generation) system?

Be concise — keep similar length. A technique that combines a retrieval engine with a generative linguistic model. Instead of responding solely from the model's memory, the system retrieves relevant documents from its knowledge base first and uses them as context to generate a "grounded" response, anchored in verifiable sources.

Are my data/questions being sent to external cloud services?

No: Our AI model (Ollama llama3.1:8b) runs on-prem on Hetzner VPS without any third-party cloud provider (no OpenAI, Anthropic or Cloudflare AI). Questions remain on the author's infrastructure, in line with a privacy-first approach.

What does the "AI offline - only sources" badge mean?

If Ollama service is not reachable, the system degrades gracefully: still shows relevant sources found by targeted search, but without linguistic model response; no automatic fallback to cloud providers.

What content is the knowledge base trained on/indexed for?

Include articles from Federico Calò's blog and selected podcast transcripts (Lex Fridman, Huberman Lab, Y Combinator, Tim Ferriss, Marcello Ascani, etc.). It is not trained on this data; the documents are indexed and retrieved upon request.

Why is there a frequency limit (rate limit) on questions?

Rate limit protects the on-prem infrastructure (CPU/RAM of the VPS running Ollama) from excessive use and ensures reasonable response times for all users. The exact limit is indicated in the info bar above the chat.