LLM to access the internet and local files and Learn Locally Only.

[markcrobinson1955]

Hi. Newby here. A few Questions:

1. I've set up and downloaded Llama 3 Instruct.
   When I try to access the internet, or local files, I get:
   "As a conversational AI model, I don't have direct access to the internet."
   Is there a different model that I should be using?

2. I want to train my LLM on my personal preferences, but I want all my training to stay local. If I train my model to answer questions the way I would answer them, I obviously want to keep that to myself.

3. I'm unable to upload documents.
   Under "Local Document Collections" I can browse to a folder containing documents, but can't seem to read them.

Thanks in Advance.

[cebtenzzre]

We should really make an FAQ, because questions like this come up a lot.

What an LLM in GPT4All can do:

• Read your question as text
• Use additional textual information from .txt and .pdf files in LocalDocs collections that you have added, and only the information that appears in the "Context" at the end of its response (which is retrieved as a separate step by a different kind of model called an embedding model)
• Generate a response to your question, as text

What an LLM in GPT4All cannot do:

• Be "trained" or "learn" anything over time. Its parameters are constant and never modified, any additional information must be part of the current conversation somehow - as a prompt, response, system prompt, or included LocalDocs context. When you start a new chat, you still have your system prompt, prompt template, and the availability of your LocalDocs collections, but all previous context is lost.
• Interact with your computer in arbitrary ways
• "See" the files on your computer, even if they are in a LocalDocs collection. It only knows about specific parts of the files in your LocalDocs collection that an embedding model found to be relevant based on their local content.
• Make requests to internet resources in order to directly retrieve their content
• See or hear audio/visual content
• Generate audio/visual content