How can we make a AI-chat application with Claude API and bubble? In the sense, it should include the context of the previous messages before generating the new response.
Yes, you can do this 100%. We may not have any cloud-specific videos, but the way that you generate a list of messages in JSON is the same with OpenAI as it is with Cloud. Here's a link to an OpenAI video of ours demonstrating this: czcams.com/video/Oqs9xgR-MBM/video.htmlsi=72vvbOiNkYl286PA&t=1146
@@planetnocode yep but here's the thing- even though I followed these tutorials to a T, it doesn't work. Well- it does work for the first 2 inputs from the user, we get the contextual response from Claude, but then suddenly the order in which it received these messages changes. For eg:- role: user, content: 5 countries in Asia Role: assistant, content: Here are 5 countries in Asia (proceeds to list them with 1. Russia, 2. China, 3....) Role: user, content: 2nd option Role: assistant, content: Great! Russia is the largest country by size in... would you like to know more about it's culture, tourism, flora and fauna.... Role: assistant, content: The tourism in Russia has been... Role: user, content: tourism And then, since the order of the last two messages have been switched, it gives an HTTP 400 error since it expected the last message on the list to be that of an Assistant's, not user's. Any way you can guide me on solving this at the earliest? Thanks a bunch!
@@planetnocode Thanks for your response- but here's the thing: If I follow these tutorials, the order of the JSON messages gets switched in the conversation and raises an HTTP 400 error, even though the first part is able to contain the context. It goes like:- User: 5 countries in Asia Assistant: Here are 5 countries in Asia... 1. India, 2. Russia, 3. China... User: 2nd option Assistant: Great! Russia is the largest ... would you like to know more about it's population, tourism, flora? Assistant: The population of Russia.. User: population The above happens with it's JSON script format. Is there any way to tackle this issue?
Heya, sorry although we used to offer dev agency style work with clients we no longer do so. Good places to look for a Bubble dev include UpWork and Contra as Bubble have recently accounted collaborations with their marketplaces.
first, thanks for this tutorial video. I was building assistant with open ai Assistants and as you know, its able to search files and save instructions but ngl its too expensive..im curious about something.. is it possible to upload files or make the bot able to search files? like for knowledge bases. so in summary is it possible to build some bot like gpt assistants with claude?
Great question. There are three ways we'd approaching this. 1. The OpenAI Assistant allows you to provide a small number of PDFs as part of the knowledge base. 2. You can use a third-party service like Voiceflow to manage the knowledge base. 3. The easiest option, without incurring much additional cost, would be to use a plugin to extract text from PDFs you upload to your Bubble app. Then, you can put all of that text in the context window of your first message or prompt that you send off to the AI. However, with this method, be aware that token usage can skyrocket.
Is there a tutorial somewhere in building a custom chatbot with own data and utilizing Anthropic similar to how Chatbase does it?
found one yet?
How can we make a AI-chat application with Claude API and bubble? In the sense, it should include the context of the previous messages before generating the new response.
Yes, you can do this 100%. We may not have any cloud-specific videos, but the way that you generate a list of messages in JSON is the same with OpenAI as it is with Cloud. Here's a link to an OpenAI video of ours demonstrating this: czcams.com/video/Oqs9xgR-MBM/video.htmlsi=72vvbOiNkYl286PA&t=1146
@@planetnocode yep but here's the thing- even though I followed these tutorials to a T, it doesn't work. Well- it does work for the first 2 inputs from the user, we get the contextual response from Claude, but then suddenly the order in which it received these messages changes. For eg:-
role: user, content: 5 countries in Asia
Role: assistant, content: Here are 5 countries in Asia (proceeds to list them with 1. Russia, 2. China, 3....)
Role: user, content: 2nd option
Role: assistant, content: Great! Russia is the largest country by size in... would you like to know more about it's culture, tourism, flora and fauna....
Role: assistant, content: The tourism in Russia has been...
Role: user, content: tourism
And then, since the order of the last two messages have been switched, it gives an HTTP 400 error since it expected the last message on the list to be that of an Assistant's, not user's.
Any way you can guide me on solving this at the earliest? Thanks a bunch!
@@planetnocode Thanks for your response- but here's the thing:
If I follow these tutorials, the order of the JSON messages gets switched in the conversation and raises an HTTP 400 error, even though the first part is able to contain the context.
It goes like:-
User: 5 countries in Asia
Assistant: Here are 5 countries in Asia... 1. India, 2. Russia, 3. China...
User: 2nd option
Assistant: Great! Russia is the largest ... would you like to know more about it's population, tourism, flora?
Assistant: The population of Russia..
User: population
The above happens with it's JSON script format. Is there any way to tackle this issue?
Do you do custom builds?
Heya, sorry although we used to offer dev agency style work with clients we no longer do so. Good places to look for a Bubble dev include UpWork and Contra as Bubble have recently accounted collaborations with their marketplaces.
No, sorry, we don't offer developer services.Contra has a good list of Bubble developers: contra.com/bubble
So A.I. goes out with a fizzle. No code is for consumers, we're developers. Clown.
🤡 👍
first, thanks for this tutorial video. I was building assistant with open ai Assistants and as you know, its able to search files and save instructions but ngl its too expensive..im curious about something.. is it possible to upload files or make the bot able to search files? like for knowledge bases. so in summary is it possible to build some bot like gpt assistants with claude?
Great question. There are three ways we'd approaching this.
1. The OpenAI Assistant allows you to provide a small number of PDFs as part of the knowledge base.
2. You can use a third-party service like Voiceflow to manage the knowledge base.
3. The easiest option, without incurring much additional cost, would be to use a plugin to extract text from PDFs you upload to your Bubble app. Then, you can put all of that text in the context window of your first message or prompt that you send off to the AI. However, with this method, be aware that token usage can skyrocket.