Live data from Hacker News

Show HN: Customizable, embeddable Chat GPT based on your own documents

libraria.dev

51–60 of 135 posts

Re: Show HN: Customizable, embeddable Chat GPT based on your own documents

#52
post #27
post #23

Testing it, nice. I tried to import a .pdf but I got an error message about not having enough "credits" https://i.imgur.com/7za9FEn.png Also, the google drive integration is not working (yet). Can you give us a glimpse of what's coming down the road? How can I get more credits and when will Gdrive integration will be available? Also: pricing?

Hi! Strange, that google drive integration should work. I'll take a look and update you right here when it does. Pricing: Feel free to send me an email bea@bogpad.io, I'll send you some credits to play around with. I'm still trying to figure out pricing!

https://i.imgur.com/OB5yLvc.png

Re: Show HN: Customizable, embeddable Chat GPT based on your own documents

#53
I'm wondering if this is designed only for in-house use, or if it's something intended to be public-facing. If a company were to embed this on their website, say, to answer questions about products for sale...

1. Would there be any way to restrict the conversation with the customer to questions related to the products?

2. What strategies could be used to prevent malicious overuse / spamming / flooding the bot to cost the company money?

Re: Show HN: Customizable, embeddable Chat GPT based on your own documents

#55
post #31

I asked Dumbledore something from the books[0] and got something somewhat innovative, somewhat wrong: "How can I turn a staircase into a slide?" > To turn a staircase into a slide, you can use the spell 'Tergeo' to remove any dirt or grime from the stairs, and then use the spell 'Lubricus' to make the stairs slippery. Lubricus is an incantation, which can be cast with a wand and by reciting the word 'Lubricus' while…

and got something somewhat innovative, somewhat wrong

That's not surprising, it's just par for the course with LLMs.

Re: Show HN: Customizable, embeddable Chat GPT based on your own documents

#57
This is amazing!

I would love to have an AWS expert handy at all times, so I tried to upload all AWS documentation using this: https://docs.aws.amazon.com/sitemap_index.xml. I can no longer use the site, so I suspect that busted something.

In hindsight, that was not cool and I'm sorry about it.

Re: Show HN: Customizable, embeddable Chat GPT based on your own documents

#59
post #30
post #25

A few things: 1) I went through Stripe checkout to upgrade to the $10/month plan, but it's still showing me as on the free plan on the billing page. 2) I guess related to 1), but I want to show my business partner the results of a quick dump of a PDF plus scrape of our website; it's not clear how to supply him with the public chat/bot URL. 3) 'Last scraped' always shows 'invalid date'. Feel free to reach out to me di…

Thank you! Looking you up right now!

Love the approach and amazed by your packaging it looks slick and mature even for such a novel tech stack. But my understanding is that all the OpenAI models (3.5-Turbo et al) have a non compete terms of use clause so wondering how you approach this ? I can see the case that you are a complementary service increasing paid usage of OpenAI, but, I can also see them saying hey we want to be the ones to do that with enhanced features you are competing against. Perhaps just a long term risk to live with if you get past the short term market traction test ? Asking because I face a similar set of questions for a different set of reasons.

Re: Show HN: Customizable, embeddable Chat GPT based on your own documents

#60
Congrats on the launch! I'm testing it now and are excited about it. I've tried few similar solutions in the past few days but they were buggy and not as feature rich as your solution. I do have a few questions:

- What's the differences between Assistant and Chat mode?

- If I have GPT4 access is there anyway I can insert my API key and use it with your app? I assume you're using 3.5 turbo.

- What toolchain are you using for splitting / chunking?

Post reply on HN