Show HN: SiteGPT – Create ChatGPT-like chatbots trained on your website content

Post by AnthonyMouse » Sun, Apr 02, 2023, 2:44 AM UTC

BoorishBears wrote: post #58

AnthonyMouse wrote: post #40

Earlier quoted context omitted.

Is that (i.e. GPT) not still a service? What people want is something they can run on their own hardware without sending their queries to some third party service which is doing who knows what with them. This is already possible if you want to mess around with green code that isn't in system repositories yet and buy expensive hardware to make it fast, but you can imagine why some people don't have the time or money f…

I mean you don't need to use GPT, it's just if you wanted to build the product in OP (ChatGPT tuned for your site) you would. Question answering can be tackled by smaller models that run on CPUs: https://huggingface.co/tasks/question-answering And if it's strictly for personal use there's always the chat-tuned stuff being built on top of LLaMA like Alpaca > waiting for Intel or AMD to realize Intel and AMD just got t…

Post by BoorishBears » Sun, Apr 02, 2023, 3:34 AM UTC

AnthonyMouse wrote: post #61

BoorishBears wrote: post #58

Earlier quoted context omitted.

I mean you don't need to use GPT, it's just if you wanted to build the product in OP (ChatGPT tuned for your site) you would. Question answering can be tackled by smaller models that run on CPUs: https://huggingface.co/tasks/question-answering And if it's strictly for personal use there's always the chat-tuned stuff being built on top of LLaMA like Alpaca > waiting for Intel or AMD to realize Intel and AMD just got t…

> I mean you don't need to use GPT, it's just if you wanted to build the product in OP (ChatGPT tuned for your site) you would. Hence the demand for something else. > Intel and AMD just got their lunch eaten by Apple Silicon which did exactly that, so I'm sure they're working on it Apple's GPU doesn't benchmark much different than competing iGPUs for gaming. It may be that the only thing stopping anyone from running…

Post by sergiotapia » Sun, Apr 02, 2023, 4:27 AM UTC

yawnxyz wrote: post #64

Ok is this really "training" a new model on the data? Or are you breaking it down into vector embeddings, and then using an embeddings search against the relevant content and then passing that into the context window of the OpenAI API? This is cool, thanks for putting it together — but I think we as a group of designers and engineers should get our language right. If we mean creating embeddings — we should say it, si…

    embedding = OpenAI.generate_embedding(some_question)

    embedding_matches = pinecone.query(embedding)
    context_strings = embedding_matches.context_strings

    OpenAI.chat(some_question + context_strings)

    """
    This is my user question, how old is James bond?
 
    using this context answer this question:

    {{from doc: james bon is 19 years old}}
    """

Post by number6 » Sun, Apr 02, 2023, 5:26 AM UTC

abxytg wrote: post #45

OpenAI provides a demo to do this in their docs: https://platform.openai.com/docs/tutorials/web-qa-embeddings IME doing this task the scraping isn't easy to generalize. the embed / chat part is honestly low hanging fruit on top of the openAI api. if you're capable of scraping the content you want to do this with I'd say whip it up yourself. its a 15m project.

Post by mpaepper » Sun, Apr 02, 2023, 6:53 AM UTC

Fatboyrunning wrote: post #5

Sorry but I am not going to pay for this before trying it out on my content. Surely there can be a test function within the website which allows me to see what a user would experience?