Basically I'm attempting to call the OpenAI's new ...
# 🌎general
f
Basically I'm attempting to call the OpenAI's new Assistant Api with the chatbot so that the LLM handles the conversations on a specific dataset.
c
That would be a question for my colleague @famous-jewelry-85388 🙂
w
Sorry to barge in, @little-france-93089 and @early-boots-14521 I believe did that already🙂
c
This is good, then they might be able to help!
f
@wide-oyster-38514 can you share where he did it? just for future readers?
@future-rocket-81226 are you facing errors or challenges? you can achieve this using an execute code card and calling OpenAi's API is using axios code
@crooked-van-25152 thanks Sabrina 🙂
f
Well I want the whole conversation to be handled by the Assistant API.
f
So you want the bot to literally act like chatGPT?
and you want OpenAI to handle everything related to it?
f
Yes. I mainly want to use Botpress to connect to twilio, zapier, etc.
f
Understood, why don't you use the AI Generate?
f
And for a UI. But everything else I want to use Open AI Assistant.
f
or do you want specific model?
f
I like retrival, web browsing capabilities, and model flexibility.
f
Understood, then you can use the execute code card, make API calls to OpenAI's API, just don't forget to also send the conversation history details
f
Well assistants API has a thing called threads, wouldn't it be stored in a thread?
f
Threads are is to make the GPT act like conversation. So each thread is a conversation.
f
Correct
f
Then just call the OpenAI's call, and in theory the thread should handle the rest
f
sweet
f
but that's theoritical, I didn't try the threads yet
not sure I have access even 😄
f
would i need to host the assistant on an EC2 instance?
f
I am sorry I am not that expert with the new GPTs, I believe it is better to ask that question on their Discord channel or their community
However I am interested in their answer 🙂
w
Hey @future-rocket-81226 you have all the great ideas. When Flowly and hf_david will be online, perhaps they can answer your questions. I believe they have connected GPTs with Botpress.
f
Got it. Thank you so much!
l
Wow, i am late to this conversation
f
Yes! where have you been man 😛
l
Glad to help! However, this API call is still in BETA form, not from Botpress...just ChatGPTs...GPTs are still in BETA scale. It literally isnt worth the effort if you are trying to scourge an LLM or Assistant at the moment. The hallucinations are absurd. To be clear, and integration with a chatbot is worth it, but the trigger quesitons will not reach the interface of Botpress. (not a BP issue) If i am being 100% transparent...creating a personal assistant, training tool, LLM, is just as feasible in BP than it is in GPT. Most importantly...you have control parameters for your Bot. (I am not a botbassador) so this endorsemen literally comes through Trial and Error with around 50 GPTs and successes in over 100 real-world chatbot integrations. To be a straight shooter to those curious, GPTs are not viable yet. Their use case can only be defined for one user. If you guys are still interested, I can host a googlemeets where we sit down and you can step-by-step follow my code lines, replit source codes, and API calls.
Short term, GPTs are a cool idea. but please, im begging..dont let guys on YT or AIAA source networks sell you ideas that Ai driven chatbots are dead, or that training tools are dead, or that employee vetting tools are dead. They arent. Whether it be for personal hobbyists or full blown entreprenuers, this business has never been more alive. Marketing is the issue. And the point I am getting at, with proper strategies and bot building methodologies, ChatGPTs cant compete. Why? They are too Open Source with too many Open control measures. This is the Ai that takes jobs. Not facilitates humans to work more efficiently. Truth be told, BP makes GPTs look obsolete far into the future. Well, I know for my portfolio atleast. Sorry for the rant, this is just a personal take, an opinion. But one from two years worth of industry experience across multiple open-source platforms. GPTs are for the lazy. Custom software is for the driven.
I about to get banned from this discord 😅
Do you use replit?
f
Heyo!
Interesting take.
Well I really like the way that the Open AI assistant works and you can use retrival.
I'd ideally want OpenAI to handle the conversations.
I know of it but I don't use it.
l
I don't disagree! I really dont. If you dont mind me asking...what use case is the specifically for?
f
Ofcourse I don't mind.
Well I want to build a chatbot that is a customer support representative (duh) but is also proficient on driving conversions when a customer lands on the web page.
For example if someone is interested in remodeling their bathroom then it'll give them information while also converting them into appointments or leads.
l
Do you want this bot to be able to live-time scrape said webpage and promote conversation variables based off of user navigation?
f
No just based off user input.
So for example you're checking out my Home Remodling website and you're interested in learning more or it's after hours.
It'll provide that support and answer questions while also simeltaneously drive you to create a lead. (phone number, name, and email).
l
I am tracking! I have a few with clients just like that. So you want to take previous response variables, drive Ai Task driven responses to products available on the site, then talk them down a funnel into a sale?
f
Well get their contact information and set an appointment while also assisting current customers.
I don't like the if this then that model. I like the generative side of language and how it's more natural.
l
Okay okay, I am tracking. I really dont want to come off as a jerk...but why hard focus on GPTs when hallucinations and GPT-4 API calls are limited within responses...when BP provides AI Tasks & Ai Generate Text/Transitions?
f
No you're not coming off as a jerk at all. I don't know too much about how BP works to be fair. I should've maybe done some homework on my end.
l
I can literally sit down with you show you..how to create a natural LM with these tasks using a solid chunked KB, Ai Tasks and Fallbacks to use previous variables as control measures
f
Yes! Are you avail now?
l
Thats what this forum is for man!
I am not available currently, I have a call at five, DM me to set some time aside.
f
I sent you a FR.
Just DMed you. This is awesome! Thank you so much
l
Perfect. I hope I helped. Just a quick recap and reiteration of what I am trying to say...GPTs hallucinate. If you were to bring this to market just as is. I mean...just as a GPTs integration to BP or other open source...it isnt trustworthy enough to place out into the world for consistent clientelle trusts.
f
And by the way I didn't come here from youtube 😛 I've just been having this idea for a while and wanted to provide AI solutions to business readily available on their website as an iFrame.
I think that chatbots today are super if this than that. And I think that's pretty lame.
l
Love that! Please check the basics out too first before diving into complexities. It will help so greatly leading into the future. When you consult with clients...remember, you are the expert!
f
Well, I guess I will mark this thread as solved for now 😄
thanks @little-france-93089 😄
f
Yes. Do you see my DM?
f
good luck @future-rocket-81226
f
Thank you i'll be around.
f
and happy bot building 🚀
2 Views