1. Hardware & OS

• Windows 10/11 con GPU NVIDIA RTX (es. 3060): già ok secondo quanto hai scritto.

• Driver NVIDIA aggiornati: https://www.nvidia.it/Download/index.aspx

2. Linguaggi & SDK

• Python 3.10+: https://www.python.org/downloads/

• Node.js: https://nodejs.org/en/download/

3. Ollama – Motore LLM Locale

• Ollama (ultima release per Windows):
  Download: https://ollama.com/download
  Repository ufficiale: https://github.com/ollama/ollama
  Guida e documentazione: https://ollama.com/docs
  Libreria modelli: https://ollama.com/library

4. Interfaccia Web UI

• Hollama (Web UI minimale):
  Repo ufficiale: https://github.com/fmaclen/hollama
  Download e istruzioni: https://github.com/fmaclen/hollama/releases
  Info tecnica, showcase & demo: https://madewithsvelte.com/hollama

• Open WebUI (alternativa ricca & plugin-friendly):
  Repo ufficiale: https://github.com/open-webui/open-webui
  Download: https://github.com/open-webui/open-webui/releases

5. Modelli LLM consigliati

• Elenco modelli uncensored, coding e professionali da Ollama:

  • CodeLlama (13B/34B): https://ollama.com/library/codellama

  • Deepseek Coder (33B): https://ollama.com/library/deepseek-coder

  • Qwen2.5-Coder (32B): https://ollama.com/library/qwen2.5-coder

  • Dolphin Mixtral/Hermes 3/Llama3 Abliterated (uncensored/legal): https://ollama.com/library/mixtral, https://ollama.com/library/hermes, https://ollama.com/library/llama3

  • Elenco generale: https://ollama.com/library

6. Toolkit RAG, Preprocessing & Database (se vuoi integrare documenti)

• LangChain (Python):
  Repo ufficiale: https://github.com/langchain-ai/langchain
  Guide: https://blog.langchain.dev/tutorial

• FAISS: https://github.com/facebookresearch/faiss

• Qdrant (alternativa): https://qdrant.tech/

• PyMuPDF (estrazione PDF): https://pymupdf.readthedocs.io/en/latest/

7. Siti di riferimento e tutorial

• Guida step-by-step Ollama: https://shiftasia.com/community/how-to-build-a-local-ai-chatbot-with-ollama-a-step-by-step-guide/

• Tutorial Hollama: https://madewithsvelte.com/hollama

• Guida integrazione RAG legale: https://www.giuseppebriganti.it/wordpress/2025/05/ia-e-avvocati-9-motore-rag-base/

• Elenco modelli uncensored: https://merlio.app/blog/top-uncensored-llms

• Documentazione Python API per Ollama: https://github.com/ollama/ollama-python

Tutto quello che serve è scaricabile e documentato ai link qui sopra.
Procedura consigliata:
Scarica i tool (Ollama, Hollama, modelli), installa Python/Node.js, configura la Web UI, scegli i modelli LLM e (facoltativo) prepara l’ambiente RAG/documentale se intendi integrare dataset.